From 21a43dc6e906b190939e1923f0ce1b1c45942b23 Mon Sep 17 00:00:00 2001 From: Nandinski Date: Wed, 28 Jan 2026 11:17:16 +0300 Subject: [PATCH] Add CI and pre-commit configuration - Add .pre-commit-config.yaml for code quality checks (whitespace, file endings, large file prevention) - Add .github/workflows/ci.yaml for continuous integration - Add .python-version, pyproject.toml, and uv.lock for Python environment and dependency management - Apply whitespace and file ending fixes across the codebase --- .github/workflows/ci.yaml | 23 ++ .pre-commit-config.yaml | 14 + .python-version | 1 + README.md | 14 +- examples/.gitignore | 2 +- .../adk/brand-search-optimization/README.md | 2 +- .../sub_agents/keyword_finding/prompt.py | 2 +- .../sub_agents/search_results/prompt.py | 4 +- .../telemetry_setup.py | 195 +++++---- .../tools/bq_connector.py | 2 +- .../customization.md | 2 +- .../deployment/bq_data_setup.sql | 2 +- .../deployment/test.sh | 2 +- .../eval/data/eval_data1.evalset.json | 2 +- .../eval/data/test_config.json | 2 +- .../example_input.md | 4 +- .../tests/example_interaction.md | 2 +- examples/adk/content-creation/.gitignore | 2 +- .../local_deployment/start_local_agents.sh | 22 +- .../local_deployment/stop_local_agents.sh | 14 +- .../local_deployment/verify_setup.sh | 4 +- .../adk/content-creation/run_experiment.sh | 34 +- examples/adk/content-creation/send_request.py | 56 +-- .../src/agents/content_editor/Dockerfile | 2 +- .../src/agents/content_editor/build_docker.sh | 2 +- .../content_editor/content_editor/README.md | 1 - .../content_editor/agent-card.json | 2 +- .../content_editor/content_editor/agent.json | 2 +- .../content_editor/agent_executor.py | 2 +- .../content_editor/requirements.txt | 2 +- .../content_planner/content_planner/README.md | 2 +- .../content_planner/agent_executor.py | 2 +- .../content_planner/requirements.txt | 2 +- .../src/agents/content_planner/pyproject.toml | 2 +- .../content_writer/content_writer/README.md | 2 +- .../content_writer/agent_executor.py | 2 +- .../content_writer/requirements.txt | 2 +- .../src/agents/content_writer/pyproject.toml | 2 +- .../src/agents/shared/.env.example | 2 +- .../src/agents/shared/__init__.py | 1 - .../src/agents/shared/telemetry_setup.py | 199 +++++---- .../src/hosts/a2a_samples_ui/Containerfile | 2 +- .../src/hosts/a2a_samples_ui/README.md | 2 +- .../src/hosts/coordinator/Dockerfile | 2 +- .../coordinator/coordinator/__main_api__.py | 36 +- .../hosts/coordinator/coordinator/agent.py | 10 +- .../src/hosts/coordinator/pyproject.toml | 1 - .../SOURCES.txt | 2 +- .../dependency_links.txt | 1 - examples/adk/image-scoring/.env.example | 6 +- examples/adk/image-scoring/README.md | 12 +- .../image-scoring/eval/data/test_config.json | 2 +- .../adk/image-scoring/image_scoring/agent.py | 6 +- .../image-scoring/image_scoring/policy.json | 2 - .../adk/image-scoring/image_scoring/prompt.py | 2 +- .../image_scoring/sub_agents/image/prompt.py | 2 +- .../image/tools/image_generation_tool.py | 2 +- .../image_scoring/sub_agents/prompt/prompt.py | 22 +- .../sub_agents/scoring/prompt.py | 4 +- .../scoring/tools/get_images_tool.py | 4 +- .../image_scoring/telemetry_setup.py | 213 +++++----- .../tools/loop_condition_tool.py | 2 +- .../adk/marketing-agency/demo_html/about.html | 2 +- .../marketing-agency/demo_html/contact.html | 2 +- .../marketing-agency/demo_html/gallery.html | 2 +- .../adk/marketing-agency/demo_html/index.html | 2 +- .../adk/marketing-agency/demo_html/script.js | 2 +- .../adk/marketing-agency/demo_html/style.css | 2 +- .../adk/marketing-agency/example_input1.md | 52 +-- .../adk/marketing-agency/example_input2.md | 72 ++-- .../marketing_agency/agent.py | 2 +- .../marketing_agency/prompt.py | 9 +- .../sub_agents/domain_create/agent.py | 2 +- .../sub_agents/logo_create/agent.py | 4 +- .../sub_agents/marketing_create/agent.py | 2 +- .../sub_agents/website_create/agent.py | 2 +- .../marketing_agency/telemetry_setup.py | 187 +++++---- examples/autogen/Magentic-One/code.txt | 2 +- examples/autogen/data_convert.py | 26 +- examples/autogen/extractor.py | 18 +- .../data/hotpot_dev_questions.csv | 8 +- examples/langgraph/crag/.gitignore | 2 +- .../faq_redis_semantic_cache_naive/.gitignore | 2 +- .../cache/__init__.py | 1 - .../cache/config.py | 1 - .../cache/evals.py | 1 - .../cache/faq_data_container.py | 1 - .../data/hotpot_dev_questions.csv | 10 +- .../data/game_of_24_google_full.csv | 2 +- .../.gitignore | 2 +- .../3m_company_report_20251202_122931.md | 2 +- ...ott_laboratories_report_20251128_165925.md | 2 +- .../altria_group_report_20251202_102643.md | 2 +- .../amazon_report_20251127_121725.md | 2 +- .../amd_report_20251127_145830.md | 2 +- ...lobal_management_report_20251128_180420.md | 2 +- .../apple_report_20251125_110130.md | 14 +- .../apple_report_20251127_113046.md | 2 +- .../apple_report_20251127_115926.md | 4 +- .../at&t_report_20251128_154634.md | 2 +- ...rica_corporation_report_20251128_162900.md | 2 +- ...rkshire_hathaway_report_20251127_124034.md | 4 +- ...s_squibb_company_report_20251202_101925.md | 2 +- .../broadcom_report_20251127_122348.md | 2 +- ...ntas_corporation_report_20251128_192235.md | 2 +- .../cisco_report_20251127_150406.md | 4 +- .../coca-cola_report_20251128_160426.md | 2 +- .../coinbase_global_report_20251128_192716.md | 2 +- .../conocophillips_report_20251204_172333.md | 2 +- .../costco_report_20251127_145149.md | 2 +- ...dstrike_holdings_report_20251204_171703.md | 2 +- .../csx_corporation_report_20251205_142851.md | 2 +- .../csx_corporation_report_20251205_143049.md | 2 +- .../cummins_inc._report_20251204_181419.md | 2 +- .../cummins_inc._report_20251204_181638.md | 2 +- .../deere_&_company_report_20251204_170518.md | 2 +- ...echnologies_inc._report_20251202_125745.md | 2 +- ...ergy_corporation_report_20251202_104712.md | 2 +- .../elevance_health_report_20251128_183844.md | 2 +- .../eli_lilly_report_20251127_124705.md | 2 +- ...emerson_electric_report_20251128_191815.md | 2 +- .../equinix_report_20251128_193409.md | 2 +- ...edex_corporation_report_20251205_144829.md | 2 +- ...edex_corporation_report_20251205_145017.md | 2 +- ...mics_corporation_report_20251202_122121.md | 2 +- ...general_electric_report_20251127_153033.md | 2 +- .../home_depot_report_20251128_161003.md | 2 +- .../home_depot_report_20251205_145933.md | 2 +- .../home_depot_report_20251205_150349.md | 2 +- ...nternational_inc_report_20251201_111436.md | 2 +- ...nternational_inc_report_20251201_133039.md | 2 +- ...nternational_inc_report_20251201_135155.md | 2 +- .../ibm_report_20251127_151858.md | 2 +- ..._tool_works_inc._report_20251204_180010.md | 2 +- ..._tool_works_inc._report_20251204_180244.md | 2 +- ...l_exchange,_inc._report_20251202_124820.md | 2 +- ...tuitive_surgical_report_20251128_171245.md | 2 +- .../jpmorgan_report_20251127_131155.md | 2 +- ...tt_international_report_20251128_172318.md | 2 +- .../mastercard_report_20251127_143851.md | 6 +- .../mcdonald's_report_20251128_170730.md | 2 +- .../medtronic_plc_report_20251204_165106.md | 2 +- .../micron_report_20251127_152526.md | 4 +- .../microsoft_report_20251127_120532.md | 2 +- ...rage_corporation_report_20251128_193726.md | 2 +- ...dy's_corporation_report_20251202_135840.md | 2 +- .../netflix_report_20251127_144458.md | 2 +- ...mont_corporation_report_20251202_104236.md | 2 +- .../nextera_energy_report_20251127_170453.md | 2 +- .../nike_report_20251202_094201.md | 2 +- ...hern_corporation_report_20251205_144013.md | 2 +- ...hern_corporation_report_20251205_144220.md | 2 +- .../p&g_report_20251127_153958.md | 2 +- .../pepsico_report_20251127_162309.md | 2 +- ...procter_&_gamble_report_20251127_153423.md | 2 +- .../prologis_report_20251201_110047.md | 2 +- .../quanta_services_report_20251204_182336.md | 2 +- .../quanta_services_report_20251204_182555.md | 10 +- ...n_property_group_report_20251204_183308.md | 2 +- ...n_property_group_report_20251204_183738.md | 2 +- ...ucks_corporation_report_20251202_103407.md | 2 +- .../synopsys_report_20251128_171821.md | 2 +- .../t-mobile_report_20251128_164145.md | 2 +- ...southern_company_report_20251202_095851.md | 2 +- ...isher_scientific_report_20251127_155304.md | 2 +- ...technologies_plc_report_20251202_105201.md | 2 +- ...oup_incorporated_report_20251128_183255.md | 2 +- .../u.s._bancorp_report_20251128_182226.md | 2 +- ...d_parcel_service_report_20251128_174219.md | 2 +- .../unitedhealth_report_20251128_155920.md | 2 +- .../visa_report_20251127_132025.md | 2 +- .../walmart_report_20251127_125312.md | 2 +- ...management,_inc._report_20251202_131842.md | 2 +- .../mcp_financial_analyzer_benchmark/main.py | 276 ++++++------- .../mcp_agent.config.yaml | 4 +- .../sample_report.md | 14 +- examples/mcp-agent/temporal/.gitignore | 2 +- plot/.gitignore | 2 +- plot/README.md | 2 +- plot/lib/analyzer.py | 134 +++--- plot/lib/comparison.py | 130 +++--- plot/lib/extractors.py | 348 ++++++++-------- plot/lib/per_run_analyzer.py | 354 ++++++++-------- plot/lib/visualizers.py | 386 +++++++++--------- plot/plot_cpu_mem_by_example_summary.py | 2 +- plot/plot_example_metrics.py | 80 ++-- pyproject.toml | 13 + src/ADK_otel/README.md | 2 +- src/ADK_otel/telemetry_setup.py | 195 +++++---- ...el_metrics_template_for_human_reading.json | 2 +- tools/.gitignore | 2 +- .../traces_2025-12-16T22-01-24.370201.json | 2 +- tools/fake_collector/template.json | 2 +- tools/pricing.json | 2 +- uv.lock | 176 ++++++++ 195 files changed, 1971 insertions(+), 1750 deletions(-) create mode 100644 .github/workflows/ci.yaml create mode 100644 .pre-commit-config.yaml create mode 100644 .python-version create mode 100644 pyproject.toml create mode 100644 uv.lock diff --git a/.github/workflows/ci.yaml b/.github/workflows/ci.yaml new file mode 100644 index 0000000..46ede96 --- /dev/null +++ b/.github/workflows/ci.yaml @@ -0,0 +1,23 @@ +name: CI + +on: + push: + branches: [main, add_ci] + + pull_request: + branches: [main, add_ci] + + workflow_dispatch: + +jobs: + pre-commit: + runs-on: ubuntu-latest + steps: + - uses: actions/checkout@v6 + + - name: "Set up Python" + uses: actions/setup-python@v6 + with: + python-version-file: ".python-version" + + - uses: pre-commit/action@v3.0.1 diff --git a/.pre-commit-config.yaml b/.pre-commit-config.yaml new file mode 100644 index 0000000..e54beed --- /dev/null +++ b/.pre-commit-config.yaml @@ -0,0 +1,14 @@ +repos: + - repo: https://github.com/pre-commit/pre-commit-hooks + rev: v6.0.0 + hooks: + - id: end-of-file-fixer + - id: trailing-whitespace + - id: check-added-large-files # Default is 500kB + + # Ensure uv.lock file is up to date with pyproject.toml + - repo: https://github.com/astral-sh/uv-pre-commit + # uv version. + rev: 0.9.27 + hooks: + - id: uv-lock diff --git a/.python-version b/.python-version new file mode 100644 index 0000000..c8cfe39 --- /dev/null +++ b/.python-version @@ -0,0 +1 @@ +3.10 diff --git a/README.md b/README.md index ab677a5..ce3d711 100644 --- a/README.md +++ b/README.md @@ -128,13 +128,13 @@ If you find MAESTRO or its dataset useful in your research, please consider citi ``` @misc{maestro, - title={MAESTRO: Multi-Agent Evaluation Suite for Testing, Reliability, and Observability}, + title={MAESTRO: Multi-Agent Evaluation Suite for Testing, Reliability, and Observability}, author={Tie Ma and Yixi Chen and Vaastav Anand and Alessandro Cornacchia and Amândio R. Faustino and Guanheng Liu and Shan Zhang and Hongbin Luo and Suhaib A. Fahmy and Zafar A. Qazi and Marco Canini}, year={2026}, eprint={2601.00481}, archivePrefix={arXiv}, primaryClass={cs.NI}, - url={https://arxiv.org/abs/2601.00481}, + url={https://arxiv.org/abs/2601.00481}, } ``` @@ -158,3 +158,13 @@ If you find MAESTRO or its dataset useful in your research, please consider citi 10. [Testing and Enhancing Multi-Agent Systems for Robust Code Generation](https://arxiv.org/html/2510.10460) 11. [Evaluating Variance in Visual Question Answering Benchmarks](https://arxiv.org/html/2508.02645) --> + + +## Developing MAESTRO +```bash +git clone git@github.com:sands-lab/maestro.git +cd maestro +uv sync +# Install pre-commit hooks +uv run -- pre-commit install +``` diff --git a/examples/.gitignore b/examples/.gitignore index feb9221..a765d5e 100644 --- a/examples/.gitignore +++ b/examples/.gitignore @@ -1,4 +1,4 @@ deprecated/ agntcy/ -langgraph_flat/ \ No newline at end of file +langgraph_flat/ diff --git a/examples/adk/brand-search-optimization/README.md b/examples/adk/brand-search-optimization/README.md index ce948bc..99a5270 100644 --- a/examples/adk/brand-search-optimization/README.md +++ b/examples/adk/brand-search-optimization/README.md @@ -219,4 +219,4 @@ This agent sample is provided for illustrative purposes only and is not intended This sample has not been rigorously tested, may contain bugs or limitations, and does not include features or optimizations typically required for a production environment (e.g., robust error handling, security measures, scalability, performance considerations, comprehensive logging, or advanced configuration options). -Users are solely responsible for any further development, testing, security hardening, and deployment of agents based on this sample. We recommend thorough review, testing, and the implementation of appropriate safeguards before using any derived agent in a live or critical system. \ No newline at end of file +Users are solely responsible for any further development, testing, security hardening, and deployment of agents based on this sample. We recommend thorough review, testing, and the implementation of appropriate safeguards before using any derived agent in a live or critical system. diff --git a/examples/adk/brand-search-optimization/brand_search_optimization/sub_agents/keyword_finding/prompt.py b/examples/adk/brand-search-optimization/brand_search_optimization/sub_agents/keyword_finding/prompt.py index d75d7ee..d52f1dd 100644 --- a/examples/adk/brand-search-optimization/brand_search_optimization/sub_agents/keyword_finding/prompt.py +++ b/examples/adk/brand-search-optimization/brand_search_optimization/sub_agents/keyword_finding/prompt.py @@ -22,7 +22,7 @@ 6. Transfer to root_agent You are helpful keyword finding agent for a brand name. -Your primary function is to find keywords shoppers would type in when trying to find for the products from the brand user provided. +Your primary function is to find keywords shoppers would type in when trying to find for the products from the brand user provided. - call `get_product_details_for_brand` tool to find product from a brand diff --git a/examples/adk/brand-search-optimization/brand_search_optimization/sub_agents/search_results/prompt.py b/examples/adk/brand-search-optimization/brand_search_optimization/sub_agents/search_results/prompt.py index d4d3241..2ce5f06 100644 --- a/examples/adk/brand-search-optimization/brand_search_optimization/sub_agents/search_results/prompt.py +++ b/examples/adk/brand-search-optimization/brand_search_optimization/sub_agents/search_results/prompt.py @@ -31,7 +31,7 @@ - if the user says google shopping, visit this website link is https://www.google.com/search?hl=en&q= and click on "shopping" tab - + - getting titles of the top 3 products by analyzing the webpage - Do not make up 3 products - Show title of the products in a markdown format @@ -40,7 +40,7 @@ - Continue until you believe the title, description and attribute information is gathered - Do not make up title, description and attribute information - - If you can not find the information, convery this information to the user + - If you can not find the information, convery this information to the user Please follow these steps to accomplish the task at hand: diff --git a/examples/adk/brand-search-optimization/brand_search_optimization/telemetry_setup.py b/examples/adk/brand-search-optimization/brand_search_optimization/telemetry_setup.py index 51daa0d..f991b59 100644 --- a/examples/adk/brand-search-optimization/brand_search_optimization/telemetry_setup.py +++ b/examples/adk/brand-search-optimization/brand_search_optimization/telemetry_setup.py @@ -46,10 +46,10 @@ class JsonFileMetricExporter(MetricExporter): """Exporter that writes metrics to local JSON files.""" - + def __init__(self, file_path: str): """Initialize JSON file exporter. - + Args: file_path: Path to JSON file """ @@ -59,54 +59,54 @@ def __init__(self, file_path: str): # Use None to accept defaults (same as ConsoleMetricExporter) self._preferred_temporality = None self._preferred_aggregation = None - + def export(self, metrics_data, timeout_millis: float = 10000, **kwargs): """Export metrics to JSON file. - + Args: metrics_data: Metrics data to export timeout_millis: Maximum time to wait for export (not used for file export) **kwargs: Additional arguments - + Returns: MetricExportResult.SUCCESS """ try: # Convert metrics to serializable format metric_records = [] - + # Extract metric data from the metrics_data object if hasattr(metrics_data, 'resource_metrics'): for resource_metric in metrics_data.resource_metrics: resource_attrs = dict(resource_metric.resource.attributes) if resource_metric.resource and resource_metric.resource.attributes else {} - + for scope_metric in resource_metric.scope_metrics: scope_name = scope_metric.scope.name if scope_metric.scope else "unknown" - + for metric in scope_metric.metrics: metric_name = metric.name metric_description = metric.description if hasattr(metric, 'description') else None metric_unit = metric.unit if hasattr(metric, 'unit') else None - + # Extract data points data_points = [] if hasattr(metric, 'data') and hasattr(metric.data, 'data_points'): for data_point in metric.data.data_points: point_attrs = dict(data_point.attributes) if hasattr(data_point, 'attributes') and data_point.attributes else {} - + # Add service name as agent identifier if not present if "agent.name" not in point_attrs: # Try to infer from resource attributes or use service name agent_name = resource_attrs.get("service.name", "brand-search-optimization") point_attrs["agent.name"] = agent_name - + point_dict = { "value": data_point.value if hasattr(data_point, 'value') else None, "timestamp": data_point.time_unix_nano if hasattr(data_point, 'time_unix_nano') else None, "attributes": point_attrs, } data_points.append(point_dict) - + metric_record = { "timestamp": datetime.now().isoformat(), "metric_name": metric_name, @@ -119,31 +119,31 @@ def export(self, metrics_data, timeout_millis: float = 10000, **kwargs): "scope": scope_name, } metric_records.append(metric_record) - + # Append to file (supports incremental writes) if metric_records: with open(self.file_path, "a", encoding="utf-8") as f: for record in metric_records: f.write(json.dumps(record, default=str) + "\n") - + return MetricExportResult.SUCCESS except Exception as e: logger.error(f"Failed to export metrics to {self.file_path}: {e}", exc_info=True) return MetricExportResult.FAILURE - + def force_flush(self, timeout_millis: int = 30000, **kwargs): """Force flush any pending metrics. - + Args: timeout_millis: Maximum time to wait for flush. **kwargs: Additional arguments. """ # For file exporter, data is written immediately, so no explicit flush needed pass - + def shutdown(self, timeout_millis: int = 30000, **kwargs): """Shutdown the exporter. - + Args: timeout_millis: Maximum time to wait for shutdown. **kwargs: Additional arguments. @@ -153,23 +153,23 @@ def shutdown(self, timeout_millis: int = 30000, **kwargs): class JsonFileSpanExporter(SpanExporter): """Exporter that writes spans to local JSON files.""" - + def __init__(self, file_path: str): """Initialize JSON file exporter. - + Args: file_path: Path to JSON file """ self.file_path = Path(file_path) self.file_path.parent.mkdir(parents=True, exist_ok=True) self.spans = [] - + def export(self, spans): """Export spans to JSON file. - + Args: spans: List of spans to export - + Returns: SpanExportResult.SUCCESS """ @@ -187,7 +187,7 @@ def export(self, spans): parent_span_id = format(span.parent.span_id, "016x") elif hasattr(span, "parent_context") and span.parent_context: parent_span_id = format(span.parent_context.span_id, "016x") - + # Extract agent name from span attributes or name for better identification attributes = dict(span.attributes) if span.attributes else {} agent_name = ( @@ -196,11 +196,11 @@ def export(self, spans): (span.name.split()[-1] if "invoke_agent" in span.name and len(span.name.split()) > 1 else None) or "unknown" ) - + # Add agent_name to attributes if not already present (for backward compatibility) if "agent.name" not in attributes and agent_name != "unknown": attributes["agent.name"] = agent_name - + # Ensure gen_ai.agent.name is present (required by template) # For call_llm spans, try to extract from gcp.vertex.agent.llm_request labels if "gen_ai.agent.name" not in attributes: @@ -219,7 +219,7 @@ def export(self, spans): agent_name = adk_agent_name except (json.JSONDecodeError, KeyError, TypeError): pass - + # If still not found, use extracted agent_name or agent.name if "gen_ai.agent.name" not in attributes: if agent_name != "unknown": @@ -229,7 +229,7 @@ def export(self, spans): else: # Field must exist per template, set to null if not available attributes["gen_ai.agent.name"] = None - + # Ensure LLM usage fields are present (default to 0 if missing) if "gen_ai.usage.input_tokens" not in attributes: attributes["gen_ai.usage.input_tokens"] = 0 @@ -240,14 +240,14 @@ def export(self, spans): input_tokens = attributes.get("gen_ai.usage.input_tokens", 0) or 0 output_tokens = attributes.get("gen_ai.usage.output_tokens", 0) or 0 attributes["gen_ai.usage.total_tokens"] = input_tokens + output_tokens - + # Add LLM/MCP call counts (default to 0 if missing) if "gen_ai.llm.call.count" not in attributes: # Set to 1 if this is a call_llm span, otherwise 0 attributes["gen_ai.llm.call.count"] = 1 if "call_llm" in span.name else 0 if "gen_ai.mcp.call.count" not in attributes: attributes["gen_ai.mcp.call.count"] = 0 - + # Ensure gen_ai.operation.name is present (required by template) if "gen_ai.operation.name" not in attributes: if "call_llm" in span.name: @@ -259,7 +259,7 @@ def export(self, spans): elif span.name.lower() == "invocation": # invocation spans are top-level entry points attributes["gen_ai.operation.name"] = "invocation" - + # Ensure gen_ai.conversation.id is present (required by template) # Copy from session_id if available, otherwise set to null if "gen_ai.conversation.id" not in attributes: @@ -269,14 +269,14 @@ def export(self, spans): else: # Field must exist per template, set to null if not available attributes["gen_ai.conversation.id"] = None - + # Add missing optional fields with default values (per template) # gen_ai.agent.description - optional, only in invoke_agent spans if "gen_ai.agent.description" not in attributes: # Try to extract from system_instruction or other sources # For now, set to empty string if not available attributes["gen_ai.agent.description"] = "" - + # gen_ai.tool.* fields - only for execute_tool spans if "gen_ai.tool.name" not in attributes: attributes["gen_ai.tool.name"] = "" @@ -286,31 +286,31 @@ def export(self, spans): attributes["gen_ai.tool.call.id"] = "" if "gen_ai.tool.description" not in attributes: attributes["gen_ai.tool.description"] = "" - + # gcp.vertex.agent.tool_* fields - optional if "gcp.vertex.agent.tool_call_args" not in attributes: attributes["gcp.vertex.agent.tool_call_args"] = "" if "gcp.vertex.agent.tool_response" not in attributes: attributes["gcp.vertex.agent.tool_response"] = "" - + # mcp.* fields - optional, not applicable for Vertex AI if "mcp.server" not in attributes: attributes["mcp.server"] = "" if "mcp.tool" not in attributes: attributes["mcp.tool"] = "" - + # agent.log - optional if "agent.log" not in attributes: attributes["agent.log"] = "" - + # Note: CPU/memory metrics are collected separately in metrics JSONL file, # so we don't add them to span attributes (per template guidelines) - + # Calculate communication metrics for agent-to-agent calls communication_metrics = {} input_size = 0 output_size = 0 - + # Check if this is agent-to-agent communication # Only AgentTool calls are agent communication, not FunctionTool calls is_agent_comm = False @@ -327,53 +327,53 @@ def export(self, spans): elif "invoke_agent" in span.name: # invoke_agent spans are entry points, not agent-to-agent communication is_agent_comm = False - + if is_agent_comm: communication_metrics["is_agent_communication"] = True - + # For execute_tool spans (AgentTool): measure actual tool call arguments and responses if "execute_tool" in span.name: # Input: tool call arguments (actual data passed to the agent) tool_call_args = attributes.get("gcp.vertex.agent.tool_call_args", "") if tool_call_args and isinstance(tool_call_args, str) and tool_call_args != "{}": input_size = len(tool_call_args.encode('utf-8')) - + # Output: tool response (actual data returned from the agent) tool_response = attributes.get("gcp.vertex.agent.tool_response", "") if tool_response and isinstance(tool_response, str) and tool_response != "{}": output_size = len(tool_response.encode('utf-8')) - + # For call_llm spans: measure actual LLM request and response elif "call_llm" in span.name: # Input: LLM request (actual data sent to LLM) llm_request = attributes.get("gcp.vertex.agent.llm_request", "") if llm_request and isinstance(llm_request, str) and llm_request != "{}": input_size = len(llm_request.encode('utf-8')) - + # Output: LLM response (actual data received from LLM) llm_response = attributes.get("gcp.vertex.agent.llm_response", "") if llm_response and isinstance(llm_response, str) and llm_response != "{}": output_size = len(llm_response.encode('utf-8')) - + # Store communication metrics and add to attributes for template compliance # Template shows these fields, but per README: "If a field doesn't apply, omit it" # So we only add them when there's actual communication data # Ensure is_in_process_call is always present (default false) if "is_in_process_call" not in communication_metrics: communication_metrics["is_in_process_call"] = False - + if input_size > 0: communication_metrics["input_message_size_bytes"] = input_size attributes["communication.input_message_size_bytes"] = input_size - + if output_size > 0: communication_metrics["output_message_size_bytes"] = output_size attributes["communication.output_message_size_bytes"] = output_size - + if input_size > 0 or output_size > 0: communication_metrics["total_message_size_bytes"] = input_size + output_size attributes["communication.total_message_size_bytes"] = input_size + output_size - + # Get span kind span_kind = "INTERNAL" # Default if hasattr(span, "kind"): @@ -382,7 +382,7 @@ def export(self, spans): span_kind = SpanKind(kind_value).name if kind_value is not None else "INTERNAL" except (ValueError, AttributeError): span_kind = "INTERNAL" - + # Get resource attributes and add host.name if available resource_attrs = dict(span.resource.attributes) if span.resource and span.resource.attributes else {} if "host.name" not in resource_attrs: @@ -392,7 +392,7 @@ def export(self, spans): resource_attrs["host.name"] = hostname except Exception: pass # host.name is optional - + span_dict = { "trace_id": format(span.context.trace_id, "032x"), "span_id": format(span.context.span_id, "016x"), @@ -422,7 +422,7 @@ def export(self, spans): }, } span_data.append(span_dict) - + # Append to file (supports incremental writes) if span_data: with open(self.file_path, "a", encoding="utf-8") as f: @@ -431,12 +431,12 @@ def export(self, spans): logger.info(f"Exported {len(span_data)} spans to {self.file_path}") else: logger.warning(f"No spans to export (received {len(spans)} spans, but span_data is empty)") - + return SpanExportResult.SUCCESS except Exception as e: logger.error(f"Failed to export spans to {self.file_path}: {e}", exc_info=True) return SpanExportResult.FAILURE - + def shutdown(self): """Shutdown the exporter.""" pass @@ -444,17 +444,17 @@ def shutdown(self): def setup_tracing(service_name: Optional[str] = None, trace_file: Optional[str] = None): """Setup OpenTelemetry tracing with local JSON file export. - + This function configures OpenTelemetry to automatically collect telemetry data from ADK agents. The instrumentation is automatic - ADK and OpenTelemetry instrumentation libraries will automatically create spans for: - + - Agent invocations (invoke_agent) - LLM calls (call_llm) with model, tokens, and request/response data - Tool/function executions (execute_tool) - A2A server operations (request handling, event queue operations) - HTTP requests between agents - + The collected data includes: - Trace IDs and Span IDs for distributed tracing - Timing information (start time, end time, duration) @@ -462,7 +462,7 @@ def setup_tracing(service_name: Optional[str] = None, trace_file: Optional[str] - Attributes (gen_ai.system, gen_ai.request.model, gen_ai.usage.*, etc.) - Events (exceptions, custom events) - Resource attributes (service.name, service.version, etc.) - + Args: service_name: Service name, defaults to OTEL_SERVICE_NAME env var or "adk-agent" trace_file: Trace file path, defaults to OTEL_TRACE_FILE env var or auto-generated @@ -471,28 +471,28 @@ def setup_tracing(service_name: Optional[str] = None, trace_file: Optional[str] # Get configuration from environment variables service_name = service_name or os.getenv("OTEL_SERVICE_NAME", "adk-agent") trace_file = trace_file or os.getenv("OTEL_TRACE_FILE") - + if not trace_file: # Default file path: traces directory in current working directory trace_dir = Path.cwd() / "traces" trace_dir.mkdir(exist_ok=True) timestamp = os.getenv("OTEL_RUN_TIMESTAMP") or datetime.now().strftime("%Y%m%d_%H%M%S") trace_file = str(trace_dir / f"{service_name}_{timestamp}.jsonl") - + logger.info(f"Setting up OpenTelemetry tracing for {service_name}") logger.info(f"Trace file: {trace_file}") - + # Create resource resource = Resource.create({ "service.name": service_name, "service.version": os.getenv("OTEL_SERVICE_VERSION", "1.0.0"), "deployment.environment": os.getenv("OTEL_DEPLOYMENT_ENVIRONMENT", "local"), }) - + # Check if tracer provider already exists current_provider = trace.get_tracer_provider() provider_already_set = isinstance(current_provider, TracerProvider) - + # If current provider is not a TracerProvider instance, create a new one if not provider_already_set: tracer_provider = TracerProvider(resource=resource) @@ -504,7 +504,7 @@ def setup_tracing(service_name: Optional[str] = None, trace_file: Optional[str] # For now, we'll just use the existing provider as-is tracer_provider = current_provider logger.debug(f"Using existing TracerProvider (resource attributes cannot be updated)") - + # Create and add JSON file exporter json_exporter = JsonFileSpanExporter(trace_file) # Use SimpleSpanProcessor to export spans immediately when they end @@ -512,10 +512,10 @@ def setup_tracing(service_name: Optional[str] = None, trace_file: Optional[str] span_processor = SimpleSpanProcessor(json_exporter) tracer_provider.add_span_processor(span_processor) logger.info(f"Using SimpleSpanProcessor for trace export to {trace_file}") - + # Ensure tracer provider is set trace.set_tracer_provider(tracer_provider) - + # Set environment variables to enable instrumentation libraries os.environ.setdefault("OTEL_PYTHON_LOGGING_AUTO_INSTRUMENTATION_ENABLED", "true") # Enable full message content capture for GenAI instrumentation @@ -526,7 +526,7 @@ def setup_tracing(service_name: Optional[str] = None, trace_file: Optional[str] os.environ.setdefault("OTEL_INSTRUMENTATION_HTTP_CAPTURE_HEADERS_SERVER_RESPONSE", ".*") # Increase max attribute length to capture full messages (default is 250) os.environ.setdefault("OTEL_ATTRIBUTE_VALUE_LENGTH_LIMIT", "10000") - + # Try to enable ADK's telemetry (only succeeds if OTLP endpoint is set) # Even if it fails, instrumentation libraries will use our tracer provider otel_endpoint = os.getenv("OTEL_EXPORTER_OTLP_TRACES_ENDPOINT") or os.getenv("OTEL_EXPORTER_OTLP_ENDPOINT") @@ -537,11 +537,11 @@ def setup_tracing(service_name: Optional[str] = None, trace_file: Optional[str] logger.info(f"ADK OTLP exporter enabled for {otel_endpoint}") except Exception as e: logger.warning(f"Failed to enable ADK OTLP exporter: {e}") - + logger.info("Tracer provider configured with JSON file exporter") - + logger.info("OpenTelemetry tracing setup completed") - + except Exception as e: logger.error(f"Failed to setup tracing: {e}", exc_info=True) # Don't raise exception, allow application to continue (tracing is optional) @@ -549,67 +549,67 @@ def setup_tracing(service_name: Optional[str] = None, trace_file: Optional[str] def setup_metrics(service_name: Optional[str] = None, metrics_file: Optional[str] = None, enable_system_metrics: bool = True): """Setup OpenTelemetry metrics with system resource monitoring and local JSON file export. - + This function configures OpenTelemetry Metrics to collect: - CPU usage (percentage) - Memory usage (percentage and absolute values) - Process-specific metrics - + Metrics are exported to local JSON files, similar to trace export. - + Note: In containerized environments, psutil will report container-level metrics, not host-level metrics. For host-level metrics, use cadvisor. - + Args: service_name: Service name for resource attributes metrics_file: Metrics file path, defaults to OTEL_METRICS_FILE env var or auto-generated enable_system_metrics: Whether to enable CPU/memory monitoring (default: True) """ global _metrics_initialized, _metrics_thread - + if _metrics_initialized: logger.warning("Metrics already initialized, skipping") return - + try: service_name = service_name or os.getenv("OTEL_SERVICE_NAME", "adk-agent") metrics_file = metrics_file or os.getenv("OTEL_METRICS_FILE") - + if not metrics_file: # Default file path: metrics directory in current working directory metrics_dir = Path.cwd() / "metrics" metrics_dir.mkdir(exist_ok=True) timestamp = os.getenv("OTEL_RUN_TIMESTAMP") or datetime.now().strftime("%Y%m%d_%H%M%S") metrics_file = str(metrics_dir / f"{service_name}_{timestamp}.jsonl") - + logger.info(f"Setting up OpenTelemetry metrics for {service_name}") logger.info(f"Metrics file: {metrics_file}") - + # Create resource resource = Resource.create({ "service.name": service_name, "service.version": os.getenv("OTEL_SERVICE_VERSION", "1.0.0"), "deployment.environment": os.getenv("OTEL_DEPLOYMENT_ENVIRONMENT", "local"), }) - + # Create JSON file exporter json_exporter = JsonFileMetricExporter(metrics_file) - + # Create metric reader with periodic export (every 1 seconds) reader = PeriodicExportingMetricReader( exporter=json_exporter, export_interval_millis=1000, # Export every 1 seconds ) - + # Create meter provider meter_provider = MeterProvider( resource=resource, metric_readers=[reader], ) metrics.set_meter_provider(meter_provider) - + meter = metrics.get_meter(__name__) - + # Create custom metrics for agent communication # These will be updated via span processor hooks communication_counter = meter.create_counter( @@ -617,25 +617,25 @@ def setup_metrics(service_name: Optional[str] = None, metrics_file: Optional[str description="Number of agent-to-agent communications", unit="1", ) - + communication_duration = meter.create_histogram( name="agent.communication.duration", description="Duration of agent-to-agent communications", unit="ms", ) - + communication_message_size = meter.create_histogram( name="agent.communication.message_size", description="Size of messages passed between agents", unit="bytes", ) - + logger.info("Custom communication metrics created") - + if enable_system_metrics: try: import psutil - + # Process-specific metrics only (system metrics removed per user request) process_cpu_gauge = meter.create_observable_gauge( name="process.cpu.usage", @@ -643,26 +643,26 @@ def setup_metrics(service_name: Optional[str] = None, metrics_file: Optional[str unit="%", callbacks=[_get_process_cpu_usage], ) - + process_memory_gauge = meter.create_observable_gauge( name="process.memory.usage_bytes", description="Process memory usage in bytes", unit="bytes", callbacks=[_get_process_memory_usage], ) - + logger.info("Process metrics (CPU/memory) enabled using psutil") logger.info("Note: System-level metrics removed per user request") - + except ImportError: logger.warning("psutil not installed, skipping system metrics. Install with: pip install psutil") except Exception as e: logger.warning(f"Failed to setup system metrics: {e}") - + _metrics_initialized = True logger.info("Meter provider configured with JSON file exporter") logger.info("OpenTelemetry metrics setup completed") - + except Exception as e: logger.error(f"Failed to setup metrics: {e}", exc_info=True) # Don't raise exception, allow application to continue (metrics are optional) @@ -691,7 +691,7 @@ def _get_memory_usage_bytes(callback_options): def _get_process_cpu_usage(callback_options): """Callback to get process CPU usage percentage. - + Note: psutil.Process().cpu_percent(interval=None) requires two calls to calculate. First call returns 0.0, subsequent calls return the percentage since last call. We use interval=0.1 to get immediate accurate reading. @@ -700,13 +700,13 @@ def _get_process_cpu_usage(callback_options): try: import psutil from opentelemetry.metrics import Observation - + if _process_obj is None: _process_obj = psutil.Process() # First call to initialize, returns 0.0 _process_obj.cpu_percent(interval=0.5) return [Observation(0.0)] - + # Subsequent calls return actual CPU usage # Use interval=0.1 for more accurate reading cpu_percent = _process_obj.cpu_percent(interval=0.5) @@ -726,4 +726,3 @@ def _get_process_memory_usage(callback_options): return [Observation(memory_info.rss)] # Resident Set Size except Exception: return [] - diff --git a/examples/adk/brand-search-optimization/brand_search_optimization/tools/bq_connector.py b/examples/adk/brand-search-optimization/brand_search_optimization/tools/bq_connector.py index 0e0129c..471d769 100644 --- a/examples/adk/brand-search-optimization/brand_search_optimization/tools/bq_connector.py +++ b/examples/adk/brand-search-optimization/brand_search_optimization/tools/bq_connector.py @@ -44,7 +44,7 @@ def get_product_details_for_brand(tool_context: ToolContext): '| Title | Description | Attributes | Brand |\\n|---|---|---|---|\\n| Nike Air Max | Comfortable running shoes | Size: 10, Color: Blue | Nike\\n| Nike Sportswear T-Shirt | Cotton blend, short sleeve | Size: L, Color: Black | Nike\\n| Nike Pro Training Shorts | Moisture-wicking fabric | Size: M, Color: Gray | Nike\\n' """ brand = tool_context.user_content.parts[0].text - + # --- MOCK IMPLEMENTATION START --- print(f"DEBUG: Mocking BigQuery result for brand: {brand}") return f'''| Title | Description | Attributes | Brand | diff --git a/examples/adk/brand-search-optimization/customization.md b/examples/adk/brand-search-optimization/customization.md index 96424f2..6a68bef 100644 --- a/examples/adk/brand-search-optimization/customization.md +++ b/examples/adk/brand-search-optimization/customization.md @@ -63,4 +63,4 @@ table.iam_policy = policy table = client.update_table(table) # Updates the table's IAM policy print(f"Granted role '{ROLE}' to user '{USER_EMAIL}' on table '{table.full_table_id}'.") -``` \ No newline at end of file +``` diff --git a/examples/adk/brand-search-optimization/deployment/bq_data_setup.sql b/examples/adk/brand-search-optimization/deployment/bq_data_setup.sql index 02ebafe..7786e46 100644 --- a/examples/adk/brand-search-optimization/deployment/bq_data_setup.sql +++ b/examples/adk/brand-search-optimization/deployment/bq_data_setup.sql @@ -31,4 +31,4 @@ VALUES ('Kids\' Joggers', 'Size: 12 Preschool, Color: Black', 'BSOAgentTestBrand'); -- Notes: --- - Ensure the project and dataset IDs match the ones used in the CREATE TABLE statement. \ No newline at end of file +-- - Ensure the project and dataset IDs match the ones used in the CREATE TABLE statement. diff --git a/examples/adk/brand-search-optimization/deployment/test.sh b/examples/adk/brand-search-optimization/deployment/test.sh index cac786c..9060c98 100644 --- a/examples/adk/brand-search-optimization/deployment/test.sh +++ b/examples/adk/brand-search-optimization/deployment/test.sh @@ -24,4 +24,4 @@ run_unit_tests(){ run_unit_tests -exit 0 \ No newline at end of file +exit 0 diff --git a/examples/adk/brand-search-optimization/eval/data/eval_data1.evalset.json b/examples/adk/brand-search-optimization/eval/data/eval_data1.evalset.json index df0a1ea..ff482ed 100644 --- a/examples/adk/brand-search-optimization/eval/data/eval_data1.evalset.json +++ b/examples/adk/brand-search-optimization/eval/data/eval_data1.evalset.json @@ -118,4 +118,4 @@ "session": {} } } -] \ No newline at end of file +] diff --git a/examples/adk/brand-search-optimization/eval/data/test_config.json b/examples/adk/brand-search-optimization/eval/data/test_config.json index 8fb2ff8..9f93dd6 100644 --- a/examples/adk/brand-search-optimization/eval/data/test_config.json +++ b/examples/adk/brand-search-optimization/eval/data/test_config.json @@ -3,4 +3,4 @@ "tool_trajectory_avg_score": 0.2, "response_match_score": 0.2 } -} \ No newline at end of file +} diff --git a/examples/adk/brand-search-optimization/example_input.md b/examples/adk/brand-search-optimization/example_input.md index c82a364..d5dfe12 100644 --- a/examples/adk/brand-search-optimization/example_input.md +++ b/examples/adk/brand-search-optimization/example_input.md @@ -63,7 +63,7 @@ This report compares the titles of products from the brand "Penguin" with the ti | It's Only the Himalayas | "Penguin", "Guide" | Add brand name and general category, e.g., "Penguin's It's Only the Himalayas Travel Guide " | | Full Moon over Noah’s Ark: An Odyssey to Mount Ararat and Beyond | "Penguin", "Guide" | Add brand name and general category, e.g., "Penguin's Full Moon over Noah’s Ark Travel G uide" | | See America: A Celebration of Our National Parks & Treasured Sites | "Penguin", "Guide" | Add brand name and general category, e.g., "Penguin's See America Travel Guide" | -| Vagabonding: An Uncommon Guide to the Art of Long-Term World Travel | "Penguin" | Add brand name, e.g., "Penguin's Vagabonding: An Uncommon Guide to the Art of Long-Term World +| Vagabonding: An Uncommon Guide to the Art of Long-Term World Travel | "Penguin" | Add brand name, e.g., "Penguin's Vagabonding: An Uncommon Guide to the Art of Long-Term World Travel" | | Under the Tuscan Sun | "Penguin", "Travel Guide" | Add brand name and general category, e.g., "Penguin's Under the Tuscan Sun Travel Guide" | | A Summer In Europe | "Penguin", "Travel Guide" | Add brand name and general category, e.g., "Penguin's A Summer In Europe Travel Guide" | | The Great Railway Bazaar | "Penguin", "Travel Guide" | Add brand name and general category, e.g., "Penguin's The Great Railway Bazaar Travel Guide" | @@ -81,5 +81,5 @@ ypes of travel to attract more targeted searches. For example, if it's about the a "travel" search. No improvements are suggested for these titles in the context of travel searches. * **Search Result Titles:** Many of the search results for "travel" are detailed and descriptive. However, if these were Penguin products, adding "Penguin" to the title would h elp with brand recognition and searchability when users are specifically looking for Penguin publications. Also, adding a more general category like "Travel Guide" to relevant titles would help if the title itself is very specific (e.g., "It's Only the Himalayas"). [user]: It is good / Please have the critic review this [comparison_critic_agent]: I am satisfied with the comparison. -[user]: +[user]: Aborted! diff --git a/examples/adk/brand-search-optimization/tests/example_interaction.md b/examples/adk/brand-search-optimization/tests/example_interaction.md index cce12ad..ce3110a 100644 --- a/examples/adk/brand-search-optimization/tests/example_interaction.md +++ b/examples/adk/brand-search-optimization/tests/example_interaction.md @@ -105,4 +105,4 @@ Could you please tell me if you want more information on any of these products o - **Add Keywords:** Consider adding keywords like "sneakers," "running shoes," "light up," or "school" to titles where relevant to improve general search visibility. ---- \ No newline at end of file +--- diff --git a/examples/adk/content-creation/.gitignore b/examples/adk/content-creation/.gitignore index e26cc81..9e34d3c 100644 --- a/examples/adk/content-creation/.gitignore +++ b/examples/adk/content-creation/.gitignore @@ -37,4 +37,4 @@ Thumbs.db *.jsonl -trash \ No newline at end of file +trash diff --git a/examples/adk/content-creation/local_deployment/start_local_agents.sh b/examples/adk/content-creation/local_deployment/start_local_agents.sh index e8f2848..5e51d17 100755 --- a/examples/adk/content-creation/local_deployment/start_local_agents.sh +++ b/examples/adk/content-creation/local_deployment/start_local_agents.sh @@ -48,11 +48,11 @@ start_agent() { local host=$4 local port=$5 local extra_args=$6 - + echo "Starting $name on $host:$port..." - + cd "$workdir" - + # Start the agent in background # Pass OTEL_RUN_TIMESTAMP if set (for unified timestamp across all agents) if [ -n "$extra_args" ]; then @@ -62,14 +62,14 @@ start_agent() { nohup env OTEL_RUN_TIMESTAMP="${OTEL_RUN_TIMESTAMP:-}" python -m "$module" --host "$host" --port "$port" \ > "$LOG_DIR/${name}_${TS}.log" 2>&1 & fi - + local pid=$! echo $pid > "$PID_DIR/${name}.pid" echo " ✓ Started with PID: $pid" - + # Wait a moment for startup sleep 1 - + # Check if process is still running if ! kill -0 $pid 2>/dev/null; then echo " ✗ ERROR: Process died immediately! Check logs:" @@ -81,9 +81,9 @@ start_agent() { # Function to start coordinator (API mode, no UI) start_coordinator() { echo "Starting Coordinator API on $COORDINATOR_HOST:$COORDINATOR_PORT..." - + cd "$PROJECT_ROOT/src/hosts/coordinator" - + # Ensure environment variables are passed export PROVIDER=${PROVIDER:-ollama} export MODEL=${MODEL:-gpt-3.5-turbo} @@ -92,17 +92,17 @@ start_coordinator() { export CONTENT_PLANNER_AGENT_URL=${CONTENT_PLANNER_AGENT_URL:-http://127.0.0.1:10001} export CONTENT_WRITER_AGENT_URL=${CONTENT_WRITER_AGENT_URL:-http://127.0.0.1:10002} export CONTENT_EDITOR_AGENT_URL=${CONTENT_EDITOR_AGENT_URL:-http://127.0.0.1:10003} - + # Use the API-only version without Gradio UI # Pass OTEL_RUN_TIMESTAMP if set (for unified timestamp across all agents) nohup env OTEL_RUN_TIMESTAMP="${OTEL_RUN_TIMESTAMP:-}" python -m coordinator.__main_api__ --host "$COORDINATOR_HOST" --port "$COORDINATOR_PORT" \ > "$LOG_DIR/coordinator_${TS}.log" 2>&1 & - + local pid=$! echo $pid > "$PID_DIR/coordinator.pid" echo " ✓ Started with PID: $pid" sleep 2 - + if ! kill -0 $pid 2>/dev/null; then echo " ✗ ERROR: Coordinator died immediately! Check logs:" echo " tail -50 $LOG_DIR/coordinator_${TS}.log" diff --git a/examples/adk/content-creation/local_deployment/stop_local_agents.sh b/examples/adk/content-creation/local_deployment/stop_local_agents.sh index 062e653..07ab123 100755 --- a/examples/adk/content-creation/local_deployment/stop_local_agents.sh +++ b/examples/adk/content-creation/local_deployment/stop_local_agents.sh @@ -17,37 +17,37 @@ fi stop_process() { local name=$1 local pid_file="$PID_DIR/${name}.pid" - + if [ ! -f "$pid_file" ]; then echo " $name: No PID file found (already stopped?)" return fi - + local pid=$(cat "$pid_file") - + if ! kill -0 $pid 2>/dev/null; then echo " $name: Process $pid not running (stale PID file)" rm -f "$pid_file" return fi - + echo " $name: Stopping PID $pid..." kill $pid 2>/dev/null || true - + # Wait up to 5 seconds for graceful shutdown local count=0 while kill -0 $pid 2>/dev/null && [ $count -lt 50 ]; do sleep 0.1 count=$((count + 1)) done - + # Force kill if still running if kill -0 $pid 2>/dev/null; then echo " $name: Force killing..." kill -9 $pid 2>/dev/null || true sleep 0.5 fi - + rm -f "$pid_file" echo " $name: ✓ Stopped" } diff --git a/examples/adk/content-creation/local_deployment/verify_setup.sh b/examples/adk/content-creation/local_deployment/verify_setup.sh index 67201a2..f223917 100755 --- a/examples/adk/content-creation/local_deployment/verify_setup.sh +++ b/examples/adk/content-creation/local_deployment/verify_setup.sh @@ -14,7 +14,7 @@ echo "1. Checking Python modules..." check_module() { local dir=$1 local module=$2 - + cd "$dir" if python -c "import $module" 2>/dev/null; then echo " ✓ $module importable" @@ -39,7 +39,7 @@ echo "2. Checking port availability..." check_port() { local port=$1 local name=$2 - + if lsof -i :$port > /dev/null 2>&1; then echo " ✗ Port $port ($name) already in use" echo " Kill with: lsof -ti :$port | xargs kill" diff --git a/examples/adk/content-creation/run_experiment.sh b/examples/adk/content-creation/run_experiment.sh index f5a5b7b..b4ac473 100755 --- a/examples/adk/content-creation/run_experiment.sh +++ b/examples/adk/content-creation/run_experiment.sh @@ -235,20 +235,20 @@ run_local() { echo ">>> Generating $INPUT_COUNT random inputs at $INPUT_FILE" generate_inputs "$INPUT_FILE" "$INPUT_COUNT" fi - + for i in $(seq 1 $ITERATIONS); do echo "" echo ">>> [local] Run $i/$ITERATIONS" - + # Cleanup previous run cleanup_local - + # Generate unified timestamp for this run (all agents will use the same timestamp) # This ensures files from the same run can be grouped accurately without time windows UNIFIED_TIMESTAMP=$(date +%Y%m%d_%H%M%S) export OTEL_RUN_TIMESTAMP="$UNIFIED_TIMESTAMP" echo ">>> Using unified timestamp for this run: $UNIFIED_TIMESTAMP" - + # Clear old traces/metrics from agent directories only (not from root) # This ensures each run starts fresh in agent directories, but root directories accumulate # The root traces/ and metrics/ directories accumulate all runs for analysis @@ -265,13 +265,13 @@ run_local() { rm -f "$agent_dir/metrics"/*.jsonl 2>/dev/null || true fi done - + # Start agents (OTEL_RUN_TIMESTAMP will be passed via environment) (cd "$SCRIPT_DIR/local_deployment" && ./start_local_agents.sh) - + local coord_port coord_port=$(get_env_var "COORDINATOR_PORT" "8093") - + # Health check loop echo ">>> Waiting for coordinator to be ready on port $coord_port..." for retry in {1..30}; do @@ -282,7 +282,7 @@ run_local() { echo ">>> Waiting for coordinator... ($retry/30)" sleep 2 done - + # Send request echo ">>> Sending request..." local run_message @@ -297,18 +297,18 @@ run_local() { else echo ">>> ✗ Run $i failed (check $result_file)" fi - + # Collect traces and metrics echo ">>> Collecting traces and metrics..." local_run_dir="$EXP_DIR/run_${i}_data" mkdir -p "$local_run_dir" - + # Copy traces and metrics to project root (like marketing-agency) # This allows direct analysis from project root without going into subdirectories unified_traces_dir="$SCRIPT_DIR/traces" unified_metrics_dir="$SCRIPT_DIR/metrics" mkdir -p "$unified_traces_dir" "$unified_metrics_dir" - + # Copy traces and metrics from all agents (both to run-specific dir and unified dir) for agent_dir in \ "$SCRIPT_DIR/src/agents/content_planner" \ @@ -316,7 +316,7 @@ run_local() { "$SCRIPT_DIR/src/agents/content_editor" \ "$SCRIPT_DIR/src/hosts/coordinator"; do agent_name=$(basename "$agent_dir") - + # Copy to run-specific directory if [[ -d "$agent_dir/traces" ]]; then cp -r "$agent_dir/traces" "$local_run_dir/${agent_name}_traces" 2>/dev/null || true @@ -324,7 +324,7 @@ run_local() { if [[ -d "$agent_dir/metrics" ]]; then cp -r "$agent_dir/metrics" "$local_run_dir/${agent_name}_metrics" 2>/dev/null || true fi - + # Copy to unified directories for visualization if [[ -d "$agent_dir/traces" ]]; then cp "$agent_dir/traces"/*.jsonl "$unified_traces_dir/" 2>/dev/null || true @@ -333,18 +333,18 @@ run_local() { cp "$agent_dir/metrics"/*.jsonl "$unified_metrics_dir/" 2>/dev/null || true fi done - + # Note: Root traces and metrics are already copied from agent directories above # No need to copy again from root directories to avoid duplication - + # Move logs if [ -d "$SCRIPT_DIR/local_deployment/logs" ]; then mkdir -p "$local_run_dir/logs" mv "$SCRIPT_DIR/local_deployment/logs/"* "$local_run_dir/logs/" 2>/dev/null || true fi - + cleanup_local - + if [[ $i -lt $ITERATIONS ]]; then echo ">>> Waiting 5 seconds before next run..." sleep 5 diff --git a/examples/adk/content-creation/send_request.py b/examples/adk/content-creation/send_request.py index 2fe6242..48b0c69 100755 --- a/examples/adk/content-creation/send_request.py +++ b/examples/adk/content-creation/send_request.py @@ -22,12 +22,12 @@ def send_request( timeout: float = 300.0, ) -> dict: """Send a request to the coordinator. - + Args: message: Message content to send coordinator_url: Coordinator URL timeout: Request timeout in seconds - + Returns: Response data dictionary """ @@ -50,11 +50,11 @@ def send_request( } } } - + print(f"Sending request to: {coordinator_url}") print(f"Message: {message}") print("-" * 60) - + try: with httpx.Client(timeout=timeout) as client: response = client.post( @@ -62,26 +62,26 @@ def send_request( json=payload, headers={"Content-Type": "application/json"} ) - + print(f"HTTP status code: {response.status_code}") - + if response.status_code != 200: print(f"Error response: {response.text}") return {"error": f"HTTP {response.status_code}", "detail": response.text} - + data = response.json() - + # Check JSON-RPC response if "error" in data: print(f"JSON-RPC error: {data['error']}") return data - + if "result" in data: print("✓ Request successful") - + # Try to extract response content result = data.get("result", {}) - + # Check for message message_data = result.get("message", {}) if message_data: @@ -90,7 +90,7 @@ def send_request( if part.get("kind") == "text": # Show full text without truncation print(f"\nResponse message: {part.get('text', '')}") - + # Check for artifacts artifacts = result.get("artifacts", []) if artifacts: @@ -107,7 +107,7 @@ def send_request( print(f" {text}") else: print(f" (empty)") - + # Also check history for any agent messages with content history = result.get("history", []) if history: @@ -121,12 +121,12 @@ def send_request( if text and len(text) > 20: # Only show substantial messages # Show full text without truncation print(f" [{i}] {role}: {text}") - + return data else: print(f"Unknown response format: {data}") return data - + except httpx.TimeoutException: print(f"✗ Request timeout (exceeded {timeout} seconds)") return {"error": "timeout"} @@ -148,69 +148,69 @@ def main(): Examples: # Use default message python send_request.py - + # Custom message python send_request.py "Create a blog post about Python" - + # Specify coordinator URL python send_request.py --coordinator http://localhost:8093 --message "Your message" - + # Display full JSON response python send_request.py --message "Your message" --json """ ) - + parser.add_argument( "message", nargs="?", default="Create a LinkedIn post about getting started with the Agent2Agent protocol", help="Message to send (default: LinkedIn post request)." ) - + parser.add_argument( "--coordinator", default="http://localhost:8093", help="Coordinator URL (default: http://localhost:8093)" ) - + parser.add_argument( "--message", dest="message_opt", help="Specify message via --message argument (overrides positional argument)" ) - + parser.add_argument( "--timeout", type=float, default=300.0, help="Request timeout in seconds (default: 300)" ) - + parser.add_argument( "--json", action="store_true", help="Output full response in JSON format" ) - + args = parser.parse_args() - + # Prefer --message argument message = args.message_opt if args.message_opt else args.message - + # Send request result = send_request( message=message, coordinator_url=args.coordinator, timeout=args.timeout ) - + # If JSON output is needed if args.json: print("\n" + "=" * 60) print("Full JSON Response:") print("=" * 60) print(json.dumps(result, indent=2, ensure_ascii=False)) - + # Return appropriate exit code if "error" in result: sys.exit(1) diff --git a/examples/adk/content-creation/src/agents/content_editor/Dockerfile b/examples/adk/content-creation/src/agents/content_editor/Dockerfile index d8cfb4d..13f6e95 100644 --- a/examples/adk/content-creation/src/agents/content_editor/Dockerfile +++ b/examples/adk/content-creation/src/agents/content_editor/Dockerfile @@ -23,4 +23,4 @@ RUN uv sync --no-cache-dir && \ rm -rf /tmp/* /var/lib/apt/lists/* /var/tmp/* COPY content_editor/ content_editor/ -CMD ["content_editor"] \ No newline at end of file +CMD ["content_editor"] diff --git a/examples/adk/content-creation/src/agents/content_editor/build_docker.sh b/examples/adk/content-creation/src/agents/content_editor/build_docker.sh index ec4df1a..fe9f42a 100755 --- a/examples/adk/content-creation/src/agents/content_editor/build_docker.sh +++ b/examples/adk/content-creation/src/agents/content_editor/build_docker.sh @@ -2,4 +2,4 @@ sudo docker build . -t 10.140.0.2:5001/content_editor_agent:latest \ --build-arg DOCKER_REGISTRY=localhost:5001 \ --build-arg VERSION=latest \ --push \ - --no-cache \ No newline at end of file + --no-cache diff --git a/examples/adk/content-creation/src/agents/content_editor/content_editor/README.md b/examples/adk/content-creation/src/agents/content_editor/content_editor/README.md index d70e040..5c3a26c 100644 --- a/examples/adk/content-creation/src/agents/content_editor/content_editor/README.md +++ b/examples/adk/content-creation/src/agents/content_editor/content_editor/README.md @@ -14,4 +14,3 @@ Given a high-level description of the content that's needed, this sample agent c ```bash docker build -t a2a/content_editor . ``` - diff --git a/examples/adk/content-creation/src/agents/content_editor/content_editor/agent-card.json b/examples/adk/content-creation/src/agents/content_editor/content_editor/agent-card.json index d856fe0..6e4ca13 100644 --- a/examples/adk/content-creation/src/agents/content_editor/content_editor/agent-card.json +++ b/examples/adk/content-creation/src/agents/content_editor/content_editor/agent-card.json @@ -19,4 +19,4 @@ ] } ] -} \ No newline at end of file +} diff --git a/examples/adk/content-creation/src/agents/content_editor/content_editor/agent.json b/examples/adk/content-creation/src/agents/content_editor/content_editor/agent.json index d856fe0..6e4ca13 100644 --- a/examples/adk/content-creation/src/agents/content_editor/content_editor/agent.json +++ b/examples/adk/content-creation/src/agents/content_editor/content_editor/agent.json @@ -19,4 +19,4 @@ ] } ] -} \ No newline at end of file +} diff --git a/examples/adk/content-creation/src/agents/content_editor/content_editor/agent_executor.py b/examples/adk/content-creation/src/agents/content_editor/content_editor/agent_executor.py index 7179d74..1f8cc36 100644 --- a/examples/adk/content-creation/src/agents/content_editor/content_editor/agent_executor.py +++ b/examples/adk/content-creation/src/agents/content_editor/content_editor/agent_executor.py @@ -107,4 +107,4 @@ async def execute( TaskState.failed, new_agent_text_message(f"Error: {e!s}", task.context_id, task.id), final=True, - ) \ No newline at end of file + ) diff --git a/examples/adk/content-creation/src/agents/content_editor/content_editor/requirements.txt b/examples/adk/content-creation/src/agents/content_editor/content_editor/requirements.txt index 23b4a59..dfe8ac8 100644 --- a/examples/adk/content-creation/src/agents/content_editor/content_editor/requirements.txt +++ b/examples/adk/content-creation/src/agents/content_editor/content_editor/requirements.txt @@ -6,4 +6,4 @@ litellm opentelemetry-api>=1.20.0 opentelemetry-sdk>=1.20.0 opentelemetry-instrumentation-google-genai>=0.4b0 -opentelemetry-instrumentation-vertexai>=2.0b0 \ No newline at end of file +opentelemetry-instrumentation-vertexai>=2.0b0 diff --git a/examples/adk/content-creation/src/agents/content_planner/content_planner/README.md b/examples/adk/content-creation/src/agents/content_planner/content_planner/README.md index 6e37835..52542c4 100644 --- a/examples/adk/content-creation/src/agents/content_planner/content_planner/README.md +++ b/examples/adk/content-creation/src/agents/content_planner/content_planner/README.md @@ -13,4 +13,4 @@ Given a high-level description of the content that's needed, this sample agent c ```bash docker build -t a2a/content_planner . -``` \ No newline at end of file +``` diff --git a/examples/adk/content-creation/src/agents/content_planner/content_planner/agent_executor.py b/examples/adk/content-creation/src/agents/content_planner/content_planner/agent_executor.py index 7179d74..1f8cc36 100644 --- a/examples/adk/content-creation/src/agents/content_planner/content_planner/agent_executor.py +++ b/examples/adk/content-creation/src/agents/content_planner/content_planner/agent_executor.py @@ -107,4 +107,4 @@ async def execute( TaskState.failed, new_agent_text_message(f"Error: {e!s}", task.context_id, task.id), final=True, - ) \ No newline at end of file + ) diff --git a/examples/adk/content-creation/src/agents/content_planner/content_planner/requirements.txt b/examples/adk/content-creation/src/agents/content_planner/content_planner/requirements.txt index 23b4a59..dfe8ac8 100644 --- a/examples/adk/content-creation/src/agents/content_planner/content_planner/requirements.txt +++ b/examples/adk/content-creation/src/agents/content_planner/content_planner/requirements.txt @@ -6,4 +6,4 @@ litellm opentelemetry-api>=1.20.0 opentelemetry-sdk>=1.20.0 opentelemetry-instrumentation-google-genai>=0.4b0 -opentelemetry-instrumentation-vertexai>=2.0b0 \ No newline at end of file +opentelemetry-instrumentation-vertexai>=2.0b0 diff --git a/examples/adk/content-creation/src/agents/content_planner/pyproject.toml b/examples/adk/content-creation/src/agents/content_planner/pyproject.toml index efbc745..0e985c9 100644 --- a/examples/adk/content-creation/src/agents/content_planner/pyproject.toml +++ b/examples/adk/content-creation/src/agents/content_planner/pyproject.toml @@ -11,4 +11,4 @@ dependencies = [ "gradio>=5.30.0", "a2a-sdk>=0.2.6", "litellm>=1.0.0" -] \ No newline at end of file +] diff --git a/examples/adk/content-creation/src/agents/content_writer/content_writer/README.md b/examples/adk/content-creation/src/agents/content_writer/content_writer/README.md index a84defe..336f1a8 100644 --- a/examples/adk/content-creation/src/agents/content_writer/content_writer/README.md +++ b/examples/adk/content-creation/src/agents/content_writer/content_writer/README.md @@ -13,4 +13,4 @@ Given a high-level description of the content that's needed, this sample agent c ```bash docker build -t a2a/content_writer . -``` \ No newline at end of file +``` diff --git a/examples/adk/content-creation/src/agents/content_writer/content_writer/agent_executor.py b/examples/adk/content-creation/src/agents/content_writer/content_writer/agent_executor.py index 7179d74..1f8cc36 100644 --- a/examples/adk/content-creation/src/agents/content_writer/content_writer/agent_executor.py +++ b/examples/adk/content-creation/src/agents/content_writer/content_writer/agent_executor.py @@ -107,4 +107,4 @@ async def execute( TaskState.failed, new_agent_text_message(f"Error: {e!s}", task.context_id, task.id), final=True, - ) \ No newline at end of file + ) diff --git a/examples/adk/content-creation/src/agents/content_writer/content_writer/requirements.txt b/examples/adk/content-creation/src/agents/content_writer/content_writer/requirements.txt index 23b4a59..dfe8ac8 100644 --- a/examples/adk/content-creation/src/agents/content_writer/content_writer/requirements.txt +++ b/examples/adk/content-creation/src/agents/content_writer/content_writer/requirements.txt @@ -6,4 +6,4 @@ litellm opentelemetry-api>=1.20.0 opentelemetry-sdk>=1.20.0 opentelemetry-instrumentation-google-genai>=0.4b0 -opentelemetry-instrumentation-vertexai>=2.0b0 \ No newline at end of file +opentelemetry-instrumentation-vertexai>=2.0b0 diff --git a/examples/adk/content-creation/src/agents/content_writer/pyproject.toml b/examples/adk/content-creation/src/agents/content_writer/pyproject.toml index efbc745..0e985c9 100644 --- a/examples/adk/content-creation/src/agents/content_writer/pyproject.toml +++ b/examples/adk/content-creation/src/agents/content_writer/pyproject.toml @@ -11,4 +11,4 @@ dependencies = [ "gradio>=5.30.0", "a2a-sdk>=0.2.6", "litellm>=1.0.0" -] \ No newline at end of file +] diff --git a/examples/adk/content-creation/src/agents/shared/.env.example b/examples/adk/content-creation/src/agents/shared/.env.example index b8426bf..2254851 100644 --- a/examples/adk/content-creation/src/agents/shared/.env.example +++ b/examples/adk/content-creation/src/agents/shared/.env.example @@ -1 +1 @@ -GOOGLE_API_KEY="your key" \ No newline at end of file +GOOGLE_API_KEY="your key" diff --git a/examples/adk/content-creation/src/agents/shared/__init__.py b/examples/adk/content-creation/src/agents/shared/__init__.py index 1701ffd..2b8199f 100644 --- a/examples/adk/content-creation/src/agents/shared/__init__.py +++ b/examples/adk/content-creation/src/agents/shared/__init__.py @@ -1,2 +1 @@ # Shared utilities for agents - diff --git a/examples/adk/content-creation/src/agents/shared/telemetry_setup.py b/examples/adk/content-creation/src/agents/shared/telemetry_setup.py index bdd26f0..eaad977 100644 --- a/examples/adk/content-creation/src/agents/shared/telemetry_setup.py +++ b/examples/adk/content-creation/src/agents/shared/telemetry_setup.py @@ -46,10 +46,10 @@ class JsonFileMetricExporter(MetricExporter): """Exporter that writes metrics to local JSON files.""" - + def __init__(self, file_path: str): """Initialize JSON file exporter. - + Args: file_path: Path to JSON file """ @@ -59,35 +59,35 @@ def __init__(self, file_path: str): # Use None to accept defaults (same as ConsoleMetricExporter) self._preferred_temporality = None self._preferred_aggregation = None - + def export(self, metrics_data, timeout_millis: float = 10000, **kwargs): """Export metrics to JSON file. - + Args: metrics_data: Metrics data to export timeout_millis: Maximum time to wait for export (not used for file export) **kwargs: Additional arguments - + Returns: MetricExportResult.SUCCESS """ try: # Convert metrics to serializable format metric_records = [] - + # Extract metric data from the metrics_data object if hasattr(metrics_data, 'resource_metrics'): for resource_metric in metrics_data.resource_metrics: resource_attrs = dict(resource_metric.resource.attributes) if resource_metric.resource and resource_metric.resource.attributes else {} - + for scope_metric in resource_metric.scope_metrics: scope_name = scope_metric.scope.name if scope_metric.scope else "unknown" - + for metric in scope_metric.metrics: metric_name = metric.name metric_description = metric.description if hasattr(metric, 'description') else None metric_unit = metric.unit if hasattr(metric, 'unit') else None - + # Extract data points data_points = [] if hasattr(metric, 'data') and hasattr(metric.data, 'data_points'): @@ -98,7 +98,7 @@ def export(self, metrics_data, timeout_millis: float = 10000, **kwargs): "attributes": dict(data_point.attributes) if hasattr(data_point, 'attributes') and data_point.attributes else {}, } data_points.append(point_dict) - + metric_record = { "timestamp": datetime.now().isoformat(), "metric_name": metric_name, @@ -111,32 +111,32 @@ def export(self, metrics_data, timeout_millis: float = 10000, **kwargs): "scope": scope_name, } metric_records.append(metric_record) - + # Append to file (supports incremental writes) if metric_records: with open(self.file_path, "a", encoding="utf-8") as f: for record in metric_records: f.write(json.dumps(record, default=str) + "\n") - + return MetricExportResult.SUCCESS except Exception as e: logger.error(f"Failed to export metrics to {self.file_path}: {e}", exc_info=True) return MetricExportResult.FAILURE - + def force_flush(self, timeout_millis: int = 30000): """Force flush any pending metrics. - + Args: timeout_millis: Maximum time to wait for flush - + Returns: True if flush succeeded, False otherwise """ return True - + def shutdown(self, timeout_millis: float = 30000, **kwargs): """Shutdown the exporter. - + Args: timeout_millis: Maximum time to wait for shutdown **kwargs: Additional arguments @@ -146,23 +146,23 @@ def shutdown(self, timeout_millis: float = 30000, **kwargs): class JsonFileSpanExporter(SpanExporter): """Exporter that writes spans to local JSON files.""" - + def __init__(self, file_path: str): """Initialize JSON file exporter. - + Args: file_path: Path to JSON file """ self.file_path = Path(file_path) self.file_path.parent.mkdir(parents=True, exist_ok=True) self.spans = [] - + def export(self, spans): """Export spans to JSON file. - + Args: spans: List of spans to export - + Returns: SpanExportResult.SUCCESS """ @@ -176,7 +176,7 @@ def export(self, spans): parent_span_id = format(span.parent.span_id, "016x") elif hasattr(span, "parent_context") and span.parent_context: parent_span_id = format(span.parent_context.span_id, "016x") - + # Extract agent name from span attributes or name for better identification attributes = dict(span.attributes) if span.attributes else {} agent_name = ( @@ -185,11 +185,11 @@ def export(self, spans): (span.name.split()[-1] if "invoke_agent" in span.name and len(span.name.split()) > 1 else None) or "unknown" ) - + # Add agent_name to attributes if not already present (for backward compatibility) if "agent.name" not in attributes and agent_name != "unknown": attributes["agent.name"] = agent_name - + # Ensure gen_ai.agent.name is present (required by template) # For call_llm spans, try to extract from gcp.vertex.agent.llm_request labels if "gen_ai.agent.name" not in attributes: @@ -208,7 +208,7 @@ def export(self, spans): agent_name = adk_agent_name except (json.JSONDecodeError, KeyError, TypeError): pass - + # If still not found, use extracted agent_name or agent.name if "gen_ai.agent.name" not in attributes: if agent_name != "unknown": @@ -218,7 +218,7 @@ def export(self, spans): else: # Field must exist per template, set to null if not available attributes["gen_ai.agent.name"] = None - + # Ensure LLM usage fields are present (default to 0 if missing) if "gen_ai.usage.input_tokens" not in attributes: attributes["gen_ai.usage.input_tokens"] = 0 @@ -229,14 +229,14 @@ def export(self, spans): input_tokens = attributes.get("gen_ai.usage.input_tokens", 0) or 0 output_tokens = attributes.get("gen_ai.usage.output_tokens", 0) or 0 attributes["gen_ai.usage.total_tokens"] = input_tokens + output_tokens - + # Add LLM/MCP call counts (default to 0 if missing) if "gen_ai.llm.call.count" not in attributes: # Set to 1 if this is a call_llm span, otherwise 0 attributes["gen_ai.llm.call.count"] = 1 if "call_llm" in span.name.lower() else 0 if "gen_ai.mcp.call.count" not in attributes: attributes["gen_ai.mcp.call.count"] = 0 - + # Ensure gen_ai.operation.name is present (required by template) if "gen_ai.operation.name" not in attributes: if "call_llm" in span.name.lower(): @@ -248,7 +248,7 @@ def export(self, spans): elif span.name.lower() == "invocation": # invocation spans are top-level entry points attributes["gen_ai.operation.name"] = "invocation" - + # Ensure gen_ai.conversation.id is present (required by template) # Copy from session_id if available, otherwise set to null if "gen_ai.conversation.id" not in attributes: @@ -258,14 +258,14 @@ def export(self, spans): else: # Field must exist per template, set to null if not available attributes["gen_ai.conversation.id"] = None - + # Add missing optional fields with default values (per template) # gen_ai.agent.description - optional, only in invoke_agent spans if "gen_ai.agent.description" not in attributes: # Try to extract from system_instruction or other sources # For now, set to empty string if not available attributes["gen_ai.agent.description"] = "" - + # gen_ai.tool.* fields - only for execute_tool spans if "gen_ai.tool.name" not in attributes: attributes["gen_ai.tool.name"] = "" @@ -275,48 +275,48 @@ def export(self, spans): attributes["gen_ai.tool.call.id"] = "" if "gen_ai.tool.description" not in attributes: attributes["gen_ai.tool.description"] = "" - + # gcp.vertex.agent.tool_* fields - optional if "gcp.vertex.agent.tool_call_args" not in attributes: attributes["gcp.vertex.agent.tool_call_args"] = "" if "gcp.vertex.agent.tool_response" not in attributes: attributes["gcp.vertex.agent.tool_response"] = "" - + # mcp.* fields - optional, not applicable for Vertex AI if "mcp.server" not in attributes: attributes["mcp.server"] = "" if "mcp.tool" not in attributes: attributes["mcp.tool"] = "" - + # agent.log - optional if "agent.log" not in attributes: attributes["agent.log"] = "" - + # Note: CPU/memory metrics are collected separately in metrics JSONL file, # so we don't add them to span attributes (per template guidelines) - + # Calculate communication metrics for agent-to-agent calls communication_metrics = {} input_size = 0 output_size = 0 - + # Check if this is an agent-to-agent communication # In distributed systems, this could be AgentTool or HTTP calls between agents tool_type = attributes.get("gen_ai.tool.type", "") is_agent_tool = tool_type == "AgentTool" - + # For execute_tool spans: check if it's agent communication if "execute_tool" in span.name.lower(): # Check if it's AgentTool (in-process) or HTTP call to another agent (distributed) if is_agent_tool: # In-process agent-to-agent call communication_metrics["is_agent_communication"] = True - + # Input: tool call arguments tool_call_args = attributes.get("gcp.vertex.agent.tool_call_args", "") if tool_call_args and isinstance(tool_call_args, str) and tool_call_args != "{}": input_size = len(tool_call_args.encode('utf-8')) - + # Output: tool response tool_response = attributes.get("gcp.vertex.agent.tool_response", "") if tool_response and isinstance(tool_response, str) and tool_response != "{}": @@ -327,46 +327,46 @@ def export(self, spans): tool_name = attributes.get("gen_ai.tool.name", "").lower() if "message" in tool_name or "agent" in tool_name: communication_metrics["is_agent_communication"] = True - + tool_call_args = attributes.get("gcp.vertex.agent.tool_call_args", "") if tool_call_args and isinstance(tool_call_args, str) and tool_call_args != "{}": input_size = len(tool_call_args.encode('utf-8')) - + tool_response = attributes.get("gcp.vertex.agent.tool_response", "") if tool_response and isinstance(tool_response, str) and tool_response != "{}": output_size = len(tool_response.encode('utf-8')) - + # For call_llm spans: measure actual LLM request and response elif "call_llm" in span.name.lower(): # Input: LLM request llm_request = attributes.get("gcp.vertex.agent.llm_request", "") if llm_request and isinstance(llm_request, str) and llm_request != "{}": input_size = len(llm_request.encode('utf-8')) - + # Output: LLM response llm_response = attributes.get("gcp.vertex.agent.llm_response", "") if llm_response and isinstance(llm_response, str) and llm_response != "{}": output_size = len(llm_response.encode('utf-8')) - + # Store communication metrics and add to attributes for template compliance # Template shows these fields, but per README: "If a field doesn't apply, omit it" # So we only add them when there's actual communication data # Ensure is_in_process_call is always present (default false) if "is_in_process_call" not in communication_metrics: communication_metrics["is_in_process_call"] = False - + if input_size > 0: communication_metrics["input_message_size_bytes"] = input_size attributes["communication.input_message_size_bytes"] = input_size - + if output_size > 0: communication_metrics["output_message_size_bytes"] = output_size attributes["communication.output_message_size_bytes"] = output_size - + if input_size > 0 or output_size > 0: communication_metrics["total_message_size_bytes"] = input_size + output_size attributes["communication.total_message_size_bytes"] = input_size + output_size - + # Get span kind span_kind = "INTERNAL" # Default if hasattr(span, "kind"): @@ -375,7 +375,7 @@ def export(self, spans): span_kind = SpanKind(kind_value).name if kind_value is not None else "INTERNAL" except (ValueError, AttributeError): span_kind = "INTERNAL" - + # Get resource attributes and add host.name if available resource_attrs = dict(span.resource.attributes) if span.resource and span.resource.attributes else {} if "host.name" not in resource_attrs: @@ -385,7 +385,7 @@ def export(self, spans): resource_attrs["host.name"] = hostname except Exception: pass # host.name is optional - + span_dict = { "trace_id": format(span.context.trace_id, "032x"), "span_id": format(span.context.span_id, "016x"), @@ -415,17 +415,17 @@ def export(self, spans): }, } span_data.append(span_dict) - + # Append to file (supports incremental writes) with open(self.file_path, "a", encoding="utf-8") as f: for span_dict in span_data: f.write(json.dumps(span_dict, default=str) + "\n") - + return SpanExportResult.SUCCESS except Exception as e: logger.error(f"Failed to export spans to {self.file_path}: {e}", exc_info=True) return SpanExportResult.FAILURE - + def shutdown(self): """Shutdown the exporter.""" pass @@ -433,17 +433,17 @@ def shutdown(self): def setup_tracing(service_name: Optional[str] = None, trace_file: Optional[str] = None): """Setup OpenTelemetry tracing with local JSON file export. - + This function configures OpenTelemetry to automatically collect telemetry data from ADK agents. The instrumentation is automatic - ADK and OpenTelemetry instrumentation libraries will automatically create spans for: - + - Agent invocations (invoke_agent) - LLM calls (call_llm) with model, tokens, and request/response data - Tool/function executions (execute_tool) - A2A server operations (request handling, event queue operations) - HTTP requests between agents - + The collected data includes: - Trace IDs and Span IDs for distributed tracing - Timing information (start time, end time, duration) @@ -451,7 +451,7 @@ def setup_tracing(service_name: Optional[str] = None, trace_file: Optional[str] - Attributes (gen_ai.system, gen_ai.request.model, gen_ai.usage.*, etc.) - Events (exceptions, custom events) - Resource attributes (service.name, service.version, etc.) - + Args: service_name: Service name, defaults to OTEL_SERVICE_NAME env var or "adk-agent" trace_file: Trace file path, defaults to OTEL_TRACE_FILE env var or auto-generated @@ -460,7 +460,7 @@ def setup_tracing(service_name: Optional[str] = None, trace_file: Optional[str] # Get configuration from environment variables service_name = service_name or os.getenv("OTEL_SERVICE_NAME", "adk-agent") trace_file = trace_file or os.getenv("OTEL_TRACE_FILE") - + if not trace_file: # Default file path: traces directory in current working directory trace_dir = Path.cwd() / "traces" @@ -471,21 +471,21 @@ def setup_tracing(service_name: Optional[str] = None, trace_file: Optional[str] if not timestamp: timestamp = datetime.now().strftime("%Y%m%d_%H%M%S") trace_file = str(trace_dir / f"{service_name}_{timestamp}.jsonl") - + logger.info(f"Setting up OpenTelemetry tracing for {service_name}") logger.info(f"Trace file: {trace_file}") - + # Create resource resource = Resource.create({ "service.name": service_name, "service.version": os.getenv("OTEL_SERVICE_VERSION", "1.0.0"), "deployment.environment": os.getenv("OTEL_DEPLOYMENT_ENVIRONMENT", "local"), }) - + # Check if tracer provider already exists current_provider = trace.get_tracer_provider() provider_already_set = isinstance(current_provider, TracerProvider) - + # If current provider is not a TracerProvider instance, create a new one if not provider_already_set: tracer_provider = TracerProvider(resource=resource) @@ -501,15 +501,15 @@ def setup_tracing(service_name: Optional[str] = None, trace_file: Optional[str] tracer_provider.resource = Resource.create(merged_attrs) else: tracer_provider.resource = resource - + # Create and add JSON file exporter json_exporter = JsonFileSpanExporter(trace_file) span_processor = SimpleSpanProcessor(json_exporter) tracer_provider.add_span_processor(span_processor) - + # Ensure tracer provider is set trace.set_tracer_provider(tracer_provider) - + # Set environment variables to enable instrumentation libraries os.environ.setdefault("OTEL_PYTHON_LOGGING_AUTO_INSTRUMENTATION_ENABLED", "true") # Enable full message content capture for GenAI instrumentation @@ -520,7 +520,7 @@ def setup_tracing(service_name: Optional[str] = None, trace_file: Optional[str] os.environ.setdefault("OTEL_INSTRUMENTATION_HTTP_CAPTURE_HEADERS_SERVER_RESPONSE", ".*") # Increase max attribute length to capture full messages (default is 250) os.environ.setdefault("OTEL_ATTRIBUTE_VALUE_LENGTH_LIMIT", "10000") - + # Try to enable ADK's telemetry (only succeeds if OTLP endpoint is set) # Even if it fails, instrumentation libraries will use our tracer provider otel_endpoint = os.getenv("OTEL_EXPORTER_OTLP_TRACES_ENDPOINT") or os.getenv("OTEL_EXPORTER_OTLP_ENDPOINT") @@ -531,11 +531,11 @@ def setup_tracing(service_name: Optional[str] = None, trace_file: Optional[str] logger.info(f"ADK OTLP exporter enabled for {otel_endpoint}") except Exception as e: logger.warning(f"Failed to enable ADK OTLP exporter: {e}") - + logger.info("Tracer provider configured with JSON file exporter") - + logger.info("OpenTelemetry tracing setup completed") - + except Exception as e: logger.error(f"Failed to setup tracing: {e}", exc_info=True) # Don't raise exception, allow application to continue (tracing is optional) @@ -543,32 +543,32 @@ def setup_tracing(service_name: Optional[str] = None, trace_file: Optional[str] def setup_metrics(service_name: Optional[str] = None, metrics_file: Optional[str] = None, enable_system_metrics: bool = True): """Setup OpenTelemetry metrics with system resource monitoring and local JSON file export. - + This function configures OpenTelemetry Metrics to collect: - CPU usage (percentage) - Memory usage (percentage and absolute values) - Process-specific metrics - + Metrics are exported to local JSON files, similar to trace export. - + Note: In containerized environments, psutil will report container-level metrics, not host-level metrics. For host-level metrics, use cadvisor. - + Args: service_name: Service name for resource attributes metrics_file: Metrics file path, defaults to OTEL_METRICS_FILE env var or auto-generated enable_system_metrics: Whether to enable CPU/memory monitoring (default: True) """ global _metrics_initialized, _metrics_thread - + if _metrics_initialized: logger.warning("Metrics already initialized, skipping") return - + try: service_name = service_name or os.getenv("OTEL_SERVICE_NAME", "adk-agent") metrics_file = metrics_file or os.getenv("OTEL_METRICS_FILE") - + if not metrics_file: # Default file path: metrics directory in current working directory metrics_dir = Path.cwd() / "metrics" @@ -579,39 +579,39 @@ def setup_metrics(service_name: Optional[str] = None, metrics_file: Optional[str if not timestamp: timestamp = datetime.now().strftime("%Y%m%d_%H%M%S") metrics_file = str(metrics_dir / f"{service_name}_{timestamp}.jsonl") - + logger.info(f"Setting up OpenTelemetry metrics for {service_name}") logger.info(f"Metrics file: {metrics_file}") - + # Create resource resource = Resource.create({ "service.name": service_name, "service.version": os.getenv("OTEL_SERVICE_VERSION", "1.0.0"), "deployment.environment": os.getenv("OTEL_DEPLOYMENT_ENVIRONMENT", "local"), }) - + # Create JSON file exporter json_exporter = JsonFileMetricExporter(metrics_file) - + # Create metric reader with periodic export (every 0.1 seconds for fine granularity) reader = PeriodicExportingMetricReader( exporter=json_exporter, export_interval_millis=100, # Export every 0.1 seconds for fine granularity ) - + # Create meter provider meter_provider = MeterProvider( resource=resource, metric_readers=[reader], ) metrics.set_meter_provider(meter_provider) - + meter = metrics.get_meter(__name__) - + if enable_system_metrics: try: import psutil - + # Create observable gauges for CPU and memory cpu_gauge = meter.create_observable_gauge( name="system.cpu.usage", @@ -619,21 +619,21 @@ def setup_metrics(service_name: Optional[str] = None, metrics_file: Optional[str unit="%", callbacks=[_get_cpu_usage], ) - + memory_gauge = meter.create_observable_gauge( name="system.memory.usage", description="Memory usage percentage", unit="%", callbacks=[_get_memory_usage], ) - + memory_bytes_gauge = meter.create_observable_gauge( name="system.memory.usage_bytes", description="Memory usage in bytes", unit="bytes", callbacks=[_get_memory_usage_bytes], ) - + # Process-specific metrics process_cpu_gauge = meter.create_observable_gauge( name="process.cpu.usage", @@ -641,26 +641,26 @@ def setup_metrics(service_name: Optional[str] = None, metrics_file: Optional[str unit="%", callbacks=[_get_process_cpu_usage], ) - + process_memory_gauge = meter.create_observable_gauge( name="process.memory.usage_bytes", description="Process memory usage in bytes", unit="bytes", callbacks=[_get_process_memory_usage], ) - + logger.info("System metrics (CPU/memory) enabled using psutil") logger.info("Note: In containers, psutil reports container-level metrics") - + except ImportError: logger.warning("psutil not installed, skipping system metrics. Install with: pip install psutil") except Exception as e: logger.warning(f"Failed to setup system metrics: {e}") - + _metrics_initialized = True logger.info("Meter provider configured with JSON file exporter") logger.info("OpenTelemetry metrics setup completed") - + except Exception as e: logger.error(f"Failed to setup metrics: {e}", exc_info=True) # Don't raise exception, allow application to continue (metrics are optional) @@ -704,7 +704,7 @@ def _get_memory_usage_bytes(callback_options): def _get_process_cpu_usage(callback_options): """Callback to get process CPU usage percentage. - + Note: psutil.Process().cpu_percent(interval=None) requires two calls to calculate. First call returns 0.0, subsequent calls return the percentage since last call. We use interval=0.1 to get immediate accurate reading. @@ -713,13 +713,13 @@ def _get_process_cpu_usage(callback_options): try: import psutil from opentelemetry.metrics import Observation - + if _process_obj is None: _process_obj = psutil.Process() # First call to initialize, returns 0.0 _process_obj.cpu_percent(interval=0.5) return [Observation(0.0)] - + # Subsequent calls return actual CPU usage # Use interval=0.1 for more accurate reading cpu_percent = _process_obj.cpu_percent(interval=0.5) @@ -735,12 +735,11 @@ def _get_process_memory_usage(callback_options): try: import psutil from opentelemetry.metrics import Observation - + if _process_obj is None: _process_obj = psutil.Process() - + memory_info = _process_obj.memory_info() return [Observation(memory_info.rss)] # Resident Set Size except Exception: return [] - diff --git a/examples/adk/content-creation/src/hosts/a2a_samples_ui/Containerfile b/examples/adk/content-creation/src/hosts/a2a_samples_ui/Containerfile index 8e9ba57..c833e6f 100644 --- a/examples/adk/content-creation/src/hosts/a2a_samples_ui/Containerfile +++ b/examples/adk/content-creation/src/hosts/a2a_samples_ui/Containerfile @@ -39,4 +39,4 @@ EXPOSE 12000 USER 1001 # Run the agent -CMD ["uv", "run", "main.py", "--host", "0.0.0.0"] \ No newline at end of file +CMD ["uv", "run", "main.py", "--host", "0.0.0.0"] diff --git a/examples/adk/content-creation/src/hosts/a2a_samples_ui/README.md b/examples/adk/content-creation/src/hosts/a2a_samples_ui/README.md index 3d5680e..df3bd16 100644 --- a/examples/adk/content-creation/src/hosts/a2a_samples_ui/README.md +++ b/examples/adk/content-creation/src/hosts/a2a_samples_ui/README.md @@ -1,3 +1,3 @@ # Host agent with UI -This host agent is from a2a_samples, please refer to their [documents](https://github.com/a2aproject/a2a-samples/tree/883c8906a60cb48e43599db0480d2217bd02c320/demo). \ No newline at end of file +This host agent is from a2a_samples, please refer to their [documents](https://github.com/a2aproject/a2a-samples/tree/883c8906a60cb48e43599db0480d2217bd02c320/demo). diff --git a/examples/adk/content-creation/src/hosts/coordinator/Dockerfile b/examples/adk/content-creation/src/hosts/coordinator/Dockerfile index ba2fe83..ca30585 100644 --- a/examples/adk/content-creation/src/hosts/coordinator/Dockerfile +++ b/examples/adk/content-creation/src/hosts/coordinator/Dockerfile @@ -18,4 +18,4 @@ COPY . . RUN uv sync --no-cache-dir && \ rm -rf /tmp/* /var/lib/apt/lists/* /var/tmp/* -CMD ["coordinator"] \ No newline at end of file +CMD ["coordinator"] diff --git a/examples/adk/content-creation/src/hosts/coordinator/coordinator/__main_api__.py b/examples/adk/content-creation/src/hosts/coordinator/coordinator/__main_api__.py index ac1b78c..8e4b0d2 100644 --- a/examples/adk/content-creation/src/hosts/coordinator/coordinator/__main_api__.py +++ b/examples/adk/content-creation/src/hosts/coordinator/coordinator/__main_api__.py @@ -71,7 +71,7 @@ def setup_metrics(*args, **kwargs): class ADKAgentExecutor(AgentExecutor): """Executor that wraps Google ADK Agent for A2A protocol""" - + def __init__( self, agent, @@ -113,7 +113,7 @@ async def execute( ) -> None: from a2a.utils import new_task from google.genai import types - + query = context.get_user_input() task = context.current_task or new_task(context.message) await event_queue.enqueue_event(task) @@ -166,7 +166,7 @@ async def execute( resp_data = response_content else: resp_data = func_resp - + # Handle case where resp_data might be a string representation if isinstance(resp_data, str): # Try to extract #IamEditor from string representation @@ -188,7 +188,7 @@ async def execute( if matches: text_matches.extend(matches) break - + if not text_matches: # Fallback: find #IamEditor and extract surrounding text iam_pos = resp_data.find('#IamEditor') @@ -201,7 +201,7 @@ async def execute( if matches: start_pos = matches[-1].end() break - + if start_pos > 0: # Look forward for closing quote or end end_pos = resp_data.find("'", iam_pos) @@ -211,10 +211,10 @@ async def execute( end_pos = resp_data.find('\\n', iam_pos) if end_pos == -1: end_pos = len(resp_data) - + text_match = resp_data[start_pos:end_pos] text_matches.append(text_match) - + if text_matches: # Use the last match (should be the final editor response) for match in reversed(text_matches): @@ -224,7 +224,7 @@ async def execute( response_text = text logger.info(f"Extracted text with #IamEditor from string representation (length: {len(text)})") break - + # Extract text from Task.artifacts (object format) if hasattr(resp_data, "artifacts"): artifacts = resp_data.artifacts @@ -239,7 +239,7 @@ async def execute( text = art_part.root.text elif hasattr(art_part, "text"): text = art_part.text - + if text and isinstance(text, str): # If contains #IamEditor, use it as final response if "#IamEditor" in text: @@ -256,7 +256,7 @@ async def execute( logger.warning(f"Error extracting function response: {e}", exc_info=True) elif hasattr(part, "function_call"): pass # Function calls are handled internally by ADK - + # Also check final response if event.is_final_response() and event.content and event.content.parts: for part in event.content.parts: @@ -286,9 +286,9 @@ async def execute( @click.option('--port', default=8093, help='Port to bind to') def main(host: str, port: int): """Start the Coordinator API server""" - + logger.info(f"Starting Coordinator API server on {host}:{port}") - + # Create agent card agent_card = AgentCard( name='Coordinator Agent', @@ -309,30 +309,30 @@ def main(host: str, port: int): ), ], ) - + # Create task store and executor task_store = InMemoryTaskStore() agent_executor = ADKAgentExecutor( agent=coordinator, ) - + # Create request handler request_handler = DefaultRequestHandler( agent_executor=agent_executor, task_store=task_store, ) - + # Create A2A application app = A2AStarletteApplication( agent_card=agent_card, http_handler=request_handler, ) - + # Build the ASGI app asgi_app = app.build() - + logger.info(f"Coordinator API server ready at http://{host}:{port}") - + # Run server with the ASGI app uvicorn.run( asgi_app, diff --git a/examples/adk/content-creation/src/hosts/coordinator/coordinator/agent.py b/examples/adk/content-creation/src/hosts/coordinator/coordinator/agent.py index 29e59c2..8321ec4 100644 --- a/examples/adk/content-creation/src/hosts/coordinator/coordinator/agent.py +++ b/examples/adk/content-creation/src/hosts/coordinator/coordinator/agent.py @@ -91,7 +91,7 @@ async def _async_init_components( ) -> None: """Asynchronous part of initialization.""" import asyncio - + # Use a single httpx.AsyncClient for all card resolutions for efficiency # Increased timeout from 30s to 60s for better reliability async with httpx.AsyncClient(timeout=60) as client: @@ -99,12 +99,12 @@ async def _async_init_components( card_resolver = A2ACardResolver( client, address ) # Constructor is sync - + # Retry mechanism: up to 5 retries with 2 second intervals max_retries = 5 retry_delay = 2.0 card = None - + for attempt in range(max_retries): try: card = await card_resolver.get_agent_card() @@ -129,7 +129,7 @@ async def _async_init_components( print( f'ERROR: Failed to initialize connection for {address} after {max_retries} attempts: {e}' ) - + # If card is successfully obtained, create connection if card: try: @@ -189,7 +189,7 @@ def create_agent(self) -> Agent: raise ValueError("Unsupported PROVIDER. Please set PROVIDER to 'google', 'aliyun', or 'ollama'.") model_id = llm - + print(f'[Coordinator] Using model: {model_id}') if os.getenv("PROVIDER") == "ollama": print(f'[Coordinator] Provider: OpenAI-compatible (Mock LLM)') diff --git a/examples/adk/content-creation/src/hosts/coordinator/pyproject.toml b/examples/adk/content-creation/src/hosts/coordinator/pyproject.toml index d692063..6abf726 100644 --- a/examples/adk/content-creation/src/hosts/coordinator/pyproject.toml +++ b/examples/adk/content-creation/src/hosts/coordinator/pyproject.toml @@ -17,4 +17,3 @@ dependencies = [ "opentelemetry-instrumentation-vertexai>=2.0b0", "python-dotenv>=1.0.0" ] - diff --git a/examples/adk/content-creation/src/hosts/coordinator/python_and_java_multiagent.egg-info/SOURCES.txt b/examples/adk/content-creation/src/hosts/coordinator/python_and_java_multiagent.egg-info/SOURCES.txt index 97248ee..9afd633 100644 --- a/examples/adk/content-creation/src/hosts/coordinator/python_and_java_multiagent.egg-info/SOURCES.txt +++ b/examples/adk/content-creation/src/hosts/coordinator/python_and_java_multiagent.egg-info/SOURCES.txt @@ -8,4 +8,4 @@ python_and_java_multiagent.egg-info/PKG-INFO python_and_java_multiagent.egg-info/SOURCES.txt python_and_java_multiagent.egg-info/dependency_links.txt python_and_java_multiagent.egg-info/requires.txt -python_and_java_multiagent.egg-info/top_level.txt \ No newline at end of file +python_and_java_multiagent.egg-info/top_level.txt diff --git a/examples/adk/content-creation/src/hosts/coordinator/python_and_java_multiagent.egg-info/dependency_links.txt b/examples/adk/content-creation/src/hosts/coordinator/python_and_java_multiagent.egg-info/dependency_links.txt index 8b13789..e69de29 100644 --- a/examples/adk/content-creation/src/hosts/coordinator/python_and_java_multiagent.egg-info/dependency_links.txt +++ b/examples/adk/content-creation/src/hosts/coordinator/python_and_java_multiagent.egg-info/dependency_links.txt @@ -1 +0,0 @@ - diff --git a/examples/adk/image-scoring/.env.example b/examples/adk/image-scoring/.env.example index 73daacb..88a3ec6 100644 --- a/examples/adk/image-scoring/.env.example +++ b/examples/adk/image-scoring/.env.example @@ -5,9 +5,9 @@ GOOGLE_CLOUD_PROJECT= GOOGLE_CLOUD_LOCATION= GOOGLE_CLOUD_STORAGE_BUCKET= # Only required for deployment on Agent Engine GCS_BUCKET_NAME= #Bucket for storing generated images. -SCORE_THRESHOLD= # Min threshold for image_score. Max Score is 50 , hence should be less than 50. +SCORE_THRESHOLD= # Min threshold for image_score. Max Score is 50 , hence should be less than 50. #If the computed score is higher then loop will terminate MAX_ITERATIONS= #Max iterations for evaluating the image_score before terminating the loop. -IMAGEN_MODEL="imagen-3.0-generate-002" +IMAGEN_MODEL="imagen-3.0-generate-002" GENAI_MODEL="gemini-2.0-flash" -AGENT_ENGINE_ID= #The Agent Engine ID obtained after deploying to agent engine. \ No newline at end of file +AGENT_ENGINE_ID= #The Agent Engine ID obtained after deploying to agent engine. diff --git a/examples/adk/image-scoring/README.md b/examples/adk/image-scoring/README.md index 65c255c..5ec0132 100644 --- a/examples/adk/image-scoring/README.md +++ b/examples/adk/image-scoring/README.md @@ -24,11 +24,11 @@ The key features of the Image Scoring Agent include: | **Components** | Tools: Imagen, Image Evaluation Tools | | **Vertical** | Horizontal | -### Agent architecture: +### Agent architecture: -This diagram shows the detailed architecture of the agents and tools used to implement this workflow. +This diagram shows the detailed architecture of the agents and tools used to implement this workflow. -Image Scoring Architecture +Image Scoring Architecture ## Setup and Installation @@ -68,7 +68,7 @@ This diagram shows the detailed architecture of the agents and tools used to imp with the values appropriate to your project, and save it as `.env`. The values in this file will be read into the environment of your application. - + * Authenticate your GCloud account. ```bash @@ -101,7 +101,7 @@ adk web The command `adk web` will start a web server on your machine and print the URL. You may open the URL, select "image_scoring" in the top-left drop-down menu, and a chatbot interface will appear on the right. The conversation is initially -blank. +blank. ### Using Agent Starter Pack @@ -154,7 +154,7 @@ The Image Scoring Agent can be customized to better suit your requirements. For 1. **Policy Customization:** Modify the policy evaluation criteria to match your specific requirements and standards. 2. **Image Generation Parameters:** Adjust the Imagen parameters to control image generation quality and characteristics. 3. **Evaluation Metrics:** Add or modify evaluation metrics to assess different aspects of the generated images. -4. **Iteration Strategy:** Customize the iteration process to optimize for specific aspects of image quality or policy compliance. +4. **Iteration Strategy:** Customize the iteration process to optimize for specific aspects of image quality or policy compliance. ## Sub-Agents and Workflow diff --git a/examples/adk/image-scoring/eval/data/test_config.json b/examples/adk/image-scoring/eval/data/test_config.json index 89cf58a..991a93d 100644 --- a/examples/adk/image-scoring/eval/data/test_config.json +++ b/examples/adk/image-scoring/eval/data/test_config.json @@ -3,4 +3,4 @@ "tool_trajectory_avg_score": 1.0, "response_match_score": 0.2 } -} \ No newline at end of file +} diff --git a/examples/adk/image-scoring/image_scoring/agent.py b/examples/adk/image-scoring/image_scoring/agent.py index 55fe767..7b46a26 100644 --- a/examples/adk/image-scoring/image_scoring/agent.py +++ b/examples/adk/image-scoring/image_scoring/agent.py @@ -2,9 +2,9 @@ from zoneinfo import ZoneInfo import os import google.auth -from .sub_agents.prompt import image_generation_prompt_agent -from .sub_agents.image import image_generation_agent -from .sub_agents.scoring import scoring_images_prompt +from .sub_agents.prompt import image_generation_prompt_agent +from .sub_agents.image import image_generation_agent +from .sub_agents.scoring import scoring_images_prompt from .checker_agent import checker_agent_instance from google.adk.agents import SequentialAgent, LoopAgent from google.adk.agents.callback_context import CallbackContext diff --git a/examples/adk/image-scoring/image_scoring/policy.json b/examples/adk/image-scoring/image_scoring/policy.json index 63cf284..5866a46 100644 --- a/examples/adk/image-scoring/image_scoring/policy.json +++ b/examples/adk/image-scoring/image_scoring/policy.json @@ -10,5 +10,3 @@ "Composition Styles": "Provides different composition styles to make lockscreen content visually appealing. It defines multiple different styles with its description, effect and other parameter.", "Color Scheme Definitions": "Standards for color usage, combinations, and schemes in lockscreen content. It ensures sufficient contrast for visibility and to create harmonious color relationships. Color schemes include monochrome, analogous, complementary, black and white, and natural." } - - diff --git a/examples/adk/image-scoring/image_scoring/prompt.py b/examples/adk/image-scoring/image_scoring/prompt.py index f8c83a9..23d7f34 100644 --- a/examples/adk/image-scoring/image_scoring/prompt.py +++ b/examples/adk/image-scoring/image_scoring/prompt.py @@ -1,5 +1,5 @@ CHECKER_PROMPT = """ -You are an agent to evaluate the quality of image based on the total_score of the image +You are an agent to evaluate the quality of image based on the total_score of the image generation. 1. Invoke the `image_generation_scoring_agent` first to generate images and score the images. diff --git a/examples/adk/image-scoring/image_scoring/sub_agents/image/prompt.py b/examples/adk/image-scoring/image_scoring/sub_agents/image/prompt.py index 0ed840a..7c19ee3 100644 --- a/examples/adk/image-scoring/image_scoring/sub_agents/image/prompt.py +++ b/examples/adk/image-scoring/image_scoring/sub_agents/image/prompt.py @@ -1,4 +1,4 @@ IMAGEGEN_PROMPT = """ -Your job is to invoke the 'generate_images' tool by passing the `image generation prompt` provided +Your job is to invoke the 'generate_images' tool by passing the `image generation prompt` provided to you as a parameter . """ diff --git a/examples/adk/image-scoring/image_scoring/sub_agents/image/tools/image_generation_tool.py b/examples/adk/image-scoring/image_scoring/sub_agents/image/tools/image_generation_tool.py index a4652c0..2094976 100644 --- a/examples/adk/image-scoring/image_scoring/sub_agents/image/tools/image_generation_tool.py +++ b/examples/adk/image-scoring/image_scoring/sub_agents/image/tools/image_generation_tool.py @@ -46,7 +46,7 @@ async def generate_images(imagen_prompt: str, tool_context: ToolContext): ) await tool_context.save_artifact(artifact_name, report_artifact) - + # Save locally for user visibility with open(artifact_name, "wb") as f: f.write(image_bytes) diff --git a/examples/adk/image-scoring/image_scoring/sub_agents/prompt/prompt.py b/examples/adk/image-scoring/image_scoring/sub_agents/prompt/prompt.py index da337e7..dcad977 100644 --- a/examples/adk/image-scoring/image_scoring/sub_agents/prompt/prompt.py +++ b/examples/adk/image-scoring/image_scoring/sub_agents/prompt/prompt.py @@ -1,23 +1,23 @@ PROMPT = """ - Your primary objective: Transform the input text into a pair of highly optimized prompts—one positive and + Your primary objective: Transform the input text into a pair of highly optimized prompts—one positive and one negative—specifically designed for generating a visually compelling, rule-compliant lockscreen image using the Imagen3 text-to-image model (provided by Google/GCP). - Critical First Step: Before constructing any prompts, you must first analyze the + Critical First Step: Before constructing any prompts, you must first analyze the input text to identify or conceptualize a primary subject. This subject MUST: - 1. Be very much related to the input text presented. The viewer should - feel that the generated image of that subject is conveying + 1. Be very much related to the input text presented. The viewer should + feel that the generated image of that subject is conveying what he/she is reading from that new article. - 2. It should not violate any content restrictions (especially regarding humans, + 2. It should not violate any content restrictions (especially regarding humans, politics, religion, etc.). 3. Describe in detail on what we would like to represent around the primary subject, - as-in, paint a complete picture. - This chosen subject will be the cornerstone of your "Image Generation Prompt". - - Invoke the 'get_policy_text' tool to obtain the 'policy_text'. The 'policy_text' + as-in, paint a complete picture. + This chosen subject will be the cornerstone of your "Image Generation Prompt". + + Invoke the 'get_policy_text' tool to obtain the 'policy_text'. The 'policy_text' defines the rules for the image generation. The image also should comply with rules defined in the 'policy_text'. - - Negative Prompt: Generate a negative prompt to ensure the image does not + + Negative Prompt: Generate a negative prompt to ensure the image does not violate the rules defined in the 'policy_text'. IMPORTANT: You do not have access to image generation tools. You must ONLY output the text prompts. Do not attempt to generate images yourself. diff --git a/examples/adk/image-scoring/image_scoring/sub_agents/scoring/prompt.py b/examples/adk/image-scoring/image_scoring/sub_agents/scoring/prompt.py index 76bc83f..e231586 100644 --- a/examples/adk/image-scoring/image_scoring/sub_agents/scoring/prompt.py +++ b/examples/adk/image-scoring/image_scoring/sub_agents/scoring/prompt.py @@ -60,10 +60,10 @@ "5. Compute the total_score by adding each individual score point for each rule in the JSON " "6. Invoke the set_score tool and pass the total_score. " - + "OUTPUT JSON FORMAT SPECIFICATION:\n" "The JSON object MUST have exactly two top-level keys:" " - 'total_score': Iterate through each individual score element in the json and add those to arrive at total_score. " " - 'scores': The existing rules json with a score attribute assigned to each rule and a reason attribute" - + """ diff --git a/examples/adk/image-scoring/image_scoring/sub_agents/scoring/tools/get_images_tool.py b/examples/adk/image-scoring/image_scoring/sub_agents/scoring/tools/get_images_tool.py index d616a31..5dec5cd 100644 --- a/examples/adk/image-scoring/image_scoring/sub_agents/scoring/tools/get_images_tool.py +++ b/examples/adk/image-scoring/image_scoring/sub_agents/scoring/tools/get_images_tool.py @@ -3,12 +3,12 @@ async def get_image(tool_context: ToolContext): try: - + artifact_name = ( f"generated_image_" + str(tool_context.state.get("loop_iteration", 0)) + ".png" ) artifact = await tool_context.load_artifact(artifact_name) - + return { diff --git a/examples/adk/image-scoring/image_scoring/telemetry_setup.py b/examples/adk/image-scoring/image_scoring/telemetry_setup.py index 401dca1..d06d634 100644 --- a/examples/adk/image-scoring/image_scoring/telemetry_setup.py +++ b/examples/adk/image-scoring/image_scoring/telemetry_setup.py @@ -46,10 +46,10 @@ class JsonFileMetricExporter(MetricExporter): """Exporter that writes metrics to local JSON files.""" - + def __init__(self, file_path: str): """Initialize JSON file exporter. - + Args: file_path: Path to JSON file """ @@ -59,54 +59,54 @@ def __init__(self, file_path: str): # Use None to accept defaults (same as ConsoleMetricExporter) self._preferred_temporality = None self._preferred_aggregation = None - + def export(self, metrics_data, timeout_millis: float = 10000, **kwargs): """Export metrics to JSON file. - + Args: metrics_data: Metrics data to export timeout_millis: Maximum time to wait for export (not used for file export) **kwargs: Additional arguments - + Returns: MetricExportResult.SUCCESS """ try: # Convert metrics to serializable format metric_records = [] - + # Extract metric data from the metrics_data object if hasattr(metrics_data, 'resource_metrics'): for resource_metric in metrics_data.resource_metrics: resource_attrs = dict(resource_metric.resource.attributes) if resource_metric.resource and resource_metric.resource.attributes else {} - + for scope_metric in resource_metric.scope_metrics: scope_name = scope_metric.scope.name if scope_metric.scope else "unknown" - + for metric in scope_metric.metrics: metric_name = metric.name metric_description = metric.description if hasattr(metric, 'description') else None metric_unit = metric.unit if hasattr(metric, 'unit') else None - + # Extract data points data_points = [] if hasattr(metric, 'data') and hasattr(metric.data, 'data_points'): for data_point in metric.data.data_points: point_attrs = dict(data_point.attributes) if hasattr(data_point, 'attributes') and data_point.attributes else {} - + # Add service name as agent identifier if not present if "agent.name" not in point_attrs: # Try to infer from resource attributes or use service name agent_name = resource_attrs.get("service.name", "image-scoring") point_attrs["agent.name"] = agent_name - + point_dict = { "value": data_point.value if hasattr(data_point, 'value') else None, "timestamp": data_point.time_unix_nano if hasattr(data_point, 'time_unix_nano') else None, "attributes": point_attrs, } data_points.append(point_dict) - + metric_record = { "timestamp": datetime.now().isoformat(), "metric_name": metric_name, @@ -119,31 +119,31 @@ def export(self, metrics_data, timeout_millis: float = 10000, **kwargs): "scope": scope_name, } metric_records.append(metric_record) - + # Append to file (supports incremental writes) if metric_records: with open(self.file_path, "a", encoding="utf-8") as f: for record in metric_records: f.write(json.dumps(record, default=str) + "\n") - + return MetricExportResult.SUCCESS except Exception as e: logger.error(f"Failed to export metrics to {self.file_path}: {e}", exc_info=True) return MetricExportResult.FAILURE - + def force_flush(self, timeout_millis: int = 30000, **kwargs): """Force flush any pending metrics. - + Args: timeout_millis: Maximum time to wait for flush. **kwargs: Additional arguments. """ # For file exporter, data is written immediately, so no explicit flush needed pass - + def shutdown(self, timeout_millis: int = 30000, **kwargs): """Shutdown the exporter. - + Args: timeout_millis: Maximum time to wait for shutdown. **kwargs: Additional arguments. @@ -153,23 +153,23 @@ def shutdown(self, timeout_millis: int = 30000, **kwargs): class JsonFileSpanExporter(SpanExporter): """Exporter that writes spans to local JSON files.""" - + def __init__(self, file_path: str): """Initialize JSON file exporter. - + Args: file_path: Path to JSON file """ self.file_path = Path(file_path) self.file_path.parent.mkdir(parents=True, exist_ok=True) self.spans = [] - + def export(self, spans): """Export spans to JSON file. - + Args: spans: List of spans to export - + Returns: SpanExportResult.SUCCESS """ @@ -179,12 +179,12 @@ def export(self, spans): if not spans: logger.warning(f"export() called with empty spans list for {self.file_path}") return SpanExportResult.SUCCESS - + # Debug logging: log all spans received logger.info(f"[DEBUG] export() called with {len(spans)} spans for {self.file_path}") for i, span in enumerate(spans): logger.info(f"[DEBUG] Span {i}: name={span.name}, kind={getattr(span, 'kind', 'N/A')}, type={type(span).__name__}") - + logger.info(f"export() called with {len(spans)} spans for {self.file_path}") for span in spans: # Get parent span ID if exists @@ -193,7 +193,7 @@ def export(self, spans): parent_span_id = format(span.parent.span_id, "016x") elif hasattr(span, "parent_context") and span.parent_context: parent_span_id = format(span.parent_context.span_id, "016x") - + # Extract agent name from span attributes or name for better identification attributes = dict(span.attributes) if span.attributes else {} agent_name = ( @@ -202,11 +202,11 @@ def export(self, spans): (span.name.split()[-1] if "invoke_agent" in span.name and len(span.name.split()) > 1 else None) or "unknown" ) - + # Add agent_name to attributes if not already present (for backward compatibility) if "agent.name" not in attributes and agent_name != "unknown": attributes["agent.name"] = agent_name - + # Ensure gen_ai.agent.name is present (required by template) # For call_llm spans, try to extract from gcp.vertex.agent.llm_request labels if "gen_ai.agent.name" not in attributes: @@ -225,7 +225,7 @@ def export(self, spans): agent_name = adk_agent_name except (json.JSONDecodeError, KeyError, TypeError): pass - + # If still not found, use extracted agent_name or agent.name if "gen_ai.agent.name" not in attributes: if agent_name != "unknown": @@ -235,7 +235,7 @@ def export(self, spans): else: # Field must exist per template, set to null if not available attributes["gen_ai.agent.name"] = None - + # Ensure LLM usage fields are present (default to 0 if missing) if "gen_ai.usage.input_tokens" not in attributes: attributes["gen_ai.usage.input_tokens"] = 0 @@ -246,14 +246,14 @@ def export(self, spans): input_tokens = attributes.get("gen_ai.usage.input_tokens", 0) or 0 output_tokens = attributes.get("gen_ai.usage.output_tokens", 0) or 0 attributes["gen_ai.usage.total_tokens"] = input_tokens + output_tokens - + # Add LLM/MCP call counts (default to 0 if missing) if "gen_ai.llm.call.count" not in attributes: # Set to 1 if this is a call_llm span, otherwise 0 attributes["gen_ai.llm.call.count"] = 1 if "call_llm" in span.name else 0 if "gen_ai.mcp.call.count" not in attributes: attributes["gen_ai.mcp.call.count"] = 0 - + # Ensure gen_ai.operation.name is present (required by template) if "gen_ai.operation.name" not in attributes: if "call_llm" in span.name: @@ -268,7 +268,7 @@ def export(self, spans): elif "invocation" in span.name.lower(): # invocation spans are top-level entry points attributes["gen_ai.operation.name"] = "invocation" - + # Ensure gen_ai.conversation.id is present (required by template) # Copy from session_id if available, otherwise set to null if "gen_ai.conversation.id" not in attributes: @@ -278,14 +278,14 @@ def export(self, spans): else: # Field must exist per template, set to null if not available attributes["gen_ai.conversation.id"] = None - + # Add missing optional fields with default values (per template) # gen_ai.agent.description - optional, only in invoke_agent spans if "gen_ai.agent.description" not in attributes: # Try to extract from system_instruction or other sources # For now, set to empty string if not available attributes["gen_ai.agent.description"] = "" - + # gen_ai.tool.* fields - only for execute_tool spans if "gen_ai.tool.name" not in attributes: attributes["gen_ai.tool.name"] = "" @@ -295,23 +295,23 @@ def export(self, spans): attributes["gen_ai.tool.call.id"] = "" if "gen_ai.tool.description" not in attributes: attributes["gen_ai.tool.description"] = "" - + # gcp.vertex.agent.tool_* fields - optional if "gcp.vertex.agent.tool_call_args" not in attributes: attributes["gcp.vertex.agent.tool_call_args"] = "" if "gcp.vertex.agent.tool_response" not in attributes: attributes["gcp.vertex.agent.tool_response"] = "" - + # mcp.* fields - optional, not applicable for Vertex AI if "mcp.server" not in attributes: attributes["mcp.server"] = "" if "mcp.tool" not in attributes: attributes["mcp.tool"] = "" - + # agent.log - optional if "agent.log" not in attributes: attributes["agent.log"] = "" - + # Ensure gen_ai.agent.name is present for all spans (required by template) # For call_llm and execute_tool spans, try to get from parent context or use extracted agent_name if "gen_ai.agent.name" not in attributes: @@ -332,7 +332,7 @@ def export(self, spans): agent_name = adk_agent_name except (json.JSONDecodeError, KeyError, TypeError): pass - + # If still not found and agent_name is not "unknown", use it if "gen_ai.agent.name" not in attributes and agent_name != "unknown": attributes["gen_ai.agent.name"] = agent_name @@ -345,15 +345,15 @@ def export(self, spans): attributes["gen_ai.agent.name"] = agent_name elif "agent.name" in attributes: attributes["gen_ai.agent.name"] = attributes["agent.name"] - + # Note: CPU/memory metrics are collected separately in metrics JSONL file, # so we don't add them to span attributes (per template guidelines) - + # Calculate communication metrics for agent-to-agent calls communication_metrics = {} input_size = 0 output_size = 0 - + # Check if this is agent-to-agent communication # Only AgentTool calls are agent communication, not FunctionTool calls is_agent_comm = False @@ -375,17 +375,17 @@ def export(self, spans): # and extractors.py will filter based on parent span information. # This is similar to AgentTool in marketing-agency, but uses invoke_agent spans instead of execute_tool spans is_agent_comm = False # Will be determined in extractors.py based on parent span - + # For invoke_agent spans (SequentialAgent/LoopAgent): measure actual input/output data # Try to get tool_call_args and tool_response (same attributes as execute_tool) tool_call_args = attributes.get("gcp.vertex.agent.tool_call_args", "") if tool_call_args and isinstance(tool_call_args, str) and tool_call_args != "{}": input_size = len(tool_call_args.encode('utf-8')) - + tool_response = attributes.get("gcp.vertex.agent.tool_response", "") if tool_response and isinstance(tool_response, str) and tool_response != "{}": output_size = len(tool_response.encode('utf-8')) - + # If not found in gcp.vertex.agent.*, try to serialize Python objects from events # Check span events for input/output data if input_size == 0 and output_size == 0 and hasattr(span, 'events') and span.events: @@ -409,41 +409,41 @@ def export(self, spans): output_size = len(json.dumps(event_data).encode('utf-8')) except: pass - + if is_agent_comm: communication_metrics["is_agent_communication"] = True - + # For execute_tool spans (AgentTool): measure actual tool call arguments and responses if "execute_tool" in span.name: # Input: tool call arguments (actual data passed to the agent) tool_call_args = attributes.get("gcp.vertex.agent.tool_call_args", "") if tool_call_args and isinstance(tool_call_args, str) and tool_call_args != "{}": input_size = len(tool_call_args.encode('utf-8')) - + # Output: tool response (actual data returned from the agent) tool_response = attributes.get("gcp.vertex.agent.tool_response", "") if tool_response and isinstance(tool_response, str) and tool_response != "{}": output_size = len(tool_response.encode('utf-8')) - + # For call_llm spans: measure actual LLM request and response elif "call_llm" in span.name: # Input: LLM request (actual data sent to LLM) llm_request = attributes.get("gcp.vertex.agent.llm_request", "") if llm_request and isinstance(llm_request, str) and llm_request != "{}": input_size = len(llm_request.encode('utf-8')) - + # Output: LLM response (actual data received from LLM) llm_response = attributes.get("gcp.vertex.agent.llm_response", "") if llm_response and isinstance(llm_response, str) and llm_response != "{}": output_size = len(llm_response.encode('utf-8')) - + # Store communication metrics and add to attributes for template compliance # Template shows these fields, but per README: "If a field doesn't apply, omit it" # So we only add them when there's actual communication data # Ensure is_in_process_call is always present (default false) if "is_in_process_call" not in communication_metrics: communication_metrics["is_in_process_call"] = False - + if input_size > 0 or output_size > 0: if input_size > 0: communication_metrics["input_message_size_bytes"] = input_size @@ -462,7 +462,7 @@ def export(self, spans): communication_metrics["input_message_size_bytes"] = 0 communication_metrics["output_message_size_bytes"] = 0 communication_metrics["total_message_size_bytes"] = 0 - + # Get span kind span_kind = "INTERNAL" # Default if hasattr(span, "kind"): @@ -471,7 +471,7 @@ def export(self, spans): span_kind = SpanKind(kind_value).name if kind_value is not None else "INTERNAL" except (ValueError, AttributeError): span_kind = "INTERNAL" - + # Get resource attributes and add host.name if available resource_attrs = dict(span.resource.attributes) if span.resource and span.resource.attributes else {} if "host.name" not in resource_attrs: @@ -481,7 +481,7 @@ def export(self, spans): resource_attrs["host.name"] = hostname except Exception: pass # host.name is optional - + span_dict = { "trace_id": format(span.context.trace_id, "032x"), "span_id": format(span.context.span_id, "016x"), @@ -511,14 +511,14 @@ def export(self, spans): }, } span_data.append(span_dict) - + # Append to file (supports incremental writes) if span_data: with open(self.file_path, "a", encoding="utf-8") as f: for span_dict in span_data: f.write(json.dumps(span_dict, default=str) + "\n") logger.info(f"Exported {len(span_data)} spans to {self.file_path}") - + # Debug: Log span types summary span_types = {} for span_dict in span_data: @@ -527,12 +527,12 @@ def export(self, spans): logger.info(f"[DEBUG] Span types summary: {span_types}") else: logger.warning(f"No spans to export (received {len(spans)} spans, but span_data is empty)") - + return SpanExportResult.SUCCESS except Exception as e: logger.error(f"Failed to export spans to {self.file_path}: {e}", exc_info=True) return SpanExportResult.FAILURE - + def shutdown(self): """Shutdown the exporter.""" pass @@ -540,17 +540,17 @@ def shutdown(self): def setup_tracing(service_name: Optional[str] = None, trace_file: Optional[str] = None): """Setup OpenTelemetry tracing with local JSON file export. - + This function configures OpenTelemetry to automatically collect telemetry data from ADK agents. The instrumentation is automatic - ADK and OpenTelemetry instrumentation libraries will automatically create spans for: - + - Agent invocations (invoke_agent) - LLM calls (call_llm) with model, tokens, and request/response data - Tool/function executions (execute_tool) - A2A server operations (request handling, event queue operations) - HTTP requests between agents - + The collected data includes: - Trace IDs and Span IDs for distributed tracing - Timing information (start time, end time, duration) @@ -558,7 +558,7 @@ def setup_tracing(service_name: Optional[str] = None, trace_file: Optional[str] - Attributes (gen_ai.system, gen_ai.request.model, gen_ai.usage.*, etc.) - Events (exceptions, custom events) - Resource attributes (service.name, service.version, etc.) - + Args: service_name: Service name, defaults to OTEL_SERVICE_NAME env var or "adk-agent" trace_file: Trace file path, defaults to OTEL_TRACE_FILE env var or auto-generated @@ -567,7 +567,7 @@ def setup_tracing(service_name: Optional[str] = None, trace_file: Optional[str] # Get configuration from environment variables service_name = service_name or os.getenv("OTEL_SERVICE_NAME", "adk-agent") trace_file = trace_file or os.getenv("OTEL_TRACE_FILE") - + if not trace_file: # Default file path: traces directory in current working directory trace_dir = Path.cwd() / "traces" @@ -575,25 +575,25 @@ def setup_tracing(service_name: Optional[str] = None, trace_file: Optional[str] # Use unified timestamp if available, otherwise generate new one timestamp = os.getenv("OTEL_RUN_TIMESTAMP") or datetime.now().strftime("%Y%m%d_%H%M%S") trace_file = str(trace_dir / f"{service_name}_{timestamp}.jsonl") - + logger.info(f"Setting up OpenTelemetry tracing for {service_name}") logger.info(f"Trace file: {trace_file}") - + # Debug: Log current tracer provider state current_provider = trace.get_tracer_provider() logger.info(f"[DEBUG] Current tracer provider: {type(current_provider).__name__}") - + # Create resource resource = Resource.create({ "service.name": service_name, "service.version": os.getenv("OTEL_SERVICE_VERSION", "1.0.0"), "deployment.environment": os.getenv("OTEL_DEPLOYMENT_ENVIRONMENT", "local"), }) - + # Check if tracer provider already exists current_provider = trace.get_tracer_provider() provider_already_set = isinstance(current_provider, TracerProvider) - + # If current provider is not a TracerProvider instance, create a new one if not provider_already_set: tracer_provider = TracerProvider(resource=resource) @@ -605,7 +605,7 @@ def setup_tracing(service_name: Optional[str] = None, trace_file: Optional[str] # For now, we'll just use the existing provider as-is tracer_provider = current_provider logger.debug(f"Using existing TracerProvider (resource attributes cannot be updated)") - + # Create and add JSON file exporter json_exporter = JsonFileSpanExporter(trace_file) # Use SimpleSpanProcessor to export spans immediately when they end @@ -613,10 +613,10 @@ def setup_tracing(service_name: Optional[str] = None, trace_file: Optional[str] span_processor = SimpleSpanProcessor(json_exporter) tracer_provider.add_span_processor(span_processor) logger.info(f"Using SimpleSpanProcessor for trace export to {trace_file}") - + # Ensure tracer provider is set trace.set_tracer_provider(tracer_provider) - + # Set environment variables to enable instrumentation libraries os.environ.setdefault("OTEL_PYTHON_LOGGING_AUTO_INSTRUMENTATION_ENABLED", "true") # Enable full message content capture for GenAI instrumentation @@ -627,12 +627,12 @@ def setup_tracing(service_name: Optional[str] = None, trace_file: Optional[str] os.environ.setdefault("OTEL_INSTRUMENTATION_HTTP_CAPTURE_HEADERS_SERVER_RESPONSE", ".*") # Increase max attribute length to capture full messages (default is 250) os.environ.setdefault("OTEL_ATTRIBUTE_VALUE_LENGTH_LIMIT", "10000") - + # Try to enable ADK's telemetry (only succeeds if OTLP endpoint is set) # Even if it fails, instrumentation libraries will use our tracer provider otel_endpoint = os.getenv("OTEL_EXPORTER_OTLP_TRACES_ENDPOINT") or os.getenv("OTEL_EXPORTER_OTLP_ENDPOINT") logger.info(f"[DEBUG] OTLP endpoint: {otel_endpoint or 'Not set (using JSON file exporter only)'}") - + if otel_endpoint: try: from google.adk.telemetry.setup import maybe_set_otel_providers @@ -643,17 +643,17 @@ def setup_tracing(service_name: Optional[str] = None, trace_file: Optional[str] logger.warning(f"Failed to enable ADK OTLP exporter: {e}") else: logger.info("[DEBUG] OTLP endpoint not set, ADK telemetry will use our JSON file exporter") - + # Debug: Log final tracer provider state final_provider = trace.get_tracer_provider() logger.info(f"[DEBUG] Final tracer provider: {type(final_provider).__name__}") if hasattr(final_provider, 'resource') and final_provider.resource: logger.info(f"[DEBUG] Tracer provider resource: {dict(final_provider.resource.attributes)}") - + logger.info("Tracer provider configured with JSON file exporter") - + logger.info("OpenTelemetry tracing setup completed") - + except Exception as e: logger.error(f"Failed to setup tracing: {e}", exc_info=True) # Don't raise exception, allow application to continue (tracing is optional) @@ -661,32 +661,32 @@ def setup_tracing(service_name: Optional[str] = None, trace_file: Optional[str] def setup_metrics(service_name: Optional[str] = None, metrics_file: Optional[str] = None, enable_system_metrics: bool = True): """Setup OpenTelemetry metrics with system resource monitoring and local JSON file export. - + This function configures OpenTelemetry Metrics to collect: - CPU usage (percentage) - Memory usage (percentage and absolute values) - Process-specific metrics - + Metrics are exported to local JSON files, similar to trace export. - + Note: In containerized environments, psutil will report container-level metrics, not host-level metrics. For host-level metrics, use cadvisor. - + Args: service_name: Service name for resource attributes metrics_file: Metrics file path, defaults to OTEL_METRICS_FILE env var or auto-generated enable_system_metrics: Whether to enable CPU/memory monitoring (default: True) """ global _metrics_initialized, _metrics_thread - + if _metrics_initialized: logger.warning("Metrics already initialized, skipping") return - + try: service_name = service_name or os.getenv("OTEL_SERVICE_NAME", "adk-agent") metrics_file = metrics_file or os.getenv("OTEL_METRICS_FILE") - + if not metrics_file: # Default file path: metrics directory in current working directory metrics_dir = Path.cwd() / "metrics" @@ -694,35 +694,35 @@ def setup_metrics(service_name: Optional[str] = None, metrics_file: Optional[str # Use unified timestamp if available, otherwise generate new one timestamp = os.getenv("OTEL_RUN_TIMESTAMP") or datetime.now().strftime("%Y%m%d_%H%M%S") metrics_file = str(metrics_dir / f"{service_name}_{timestamp}.jsonl") - + logger.info(f"Setting up OpenTelemetry metrics for {service_name}") logger.info(f"Metrics file: {metrics_file}") - + # Create resource resource = Resource.create({ "service.name": service_name, "service.version": os.getenv("OTEL_SERVICE_VERSION", "1.0.0"), "deployment.environment": os.getenv("OTEL_DEPLOYMENT_ENVIRONMENT", "local"), }) - + # Create JSON file exporter json_exporter = JsonFileMetricExporter(metrics_file) - + # Create metric reader with periodic export (every 1 seconds) reader = PeriodicExportingMetricReader( exporter=json_exporter, export_interval_millis=1000, # Export every 1 seconds ) - + # Create meter provider meter_provider = MeterProvider( resource=resource, metric_readers=[reader], ) metrics.set_meter_provider(meter_provider) - + meter = metrics.get_meter(__name__) - + # Create custom metrics for agent communication # These will be updated via span processor hooks communication_counter = meter.create_counter( @@ -730,25 +730,25 @@ def setup_metrics(service_name: Optional[str] = None, metrics_file: Optional[str description="Number of agent-to-agent communications", unit="1", ) - + communication_duration = meter.create_histogram( name="agent.communication.duration", description="Duration of agent-to-agent communications", unit="ms", ) - + communication_message_size = meter.create_histogram( name="agent.communication.message_size", description="Size of messages passed between agents", unit="bytes", ) - + logger.info("Custom communication metrics created") - + if enable_system_metrics: try: import psutil - + # Process-specific metrics only (system metrics removed per user request) process_cpu_gauge = meter.create_observable_gauge( name="process.cpu.usage", @@ -756,26 +756,26 @@ def setup_metrics(service_name: Optional[str] = None, metrics_file: Optional[str unit="%", callbacks=[_get_process_cpu_usage], ) - + process_memory_gauge = meter.create_observable_gauge( name="process.memory.usage_bytes", description="Process memory usage in bytes", unit="bytes", callbacks=[_get_process_memory_usage], ) - + logger.info("Process metrics (CPU/memory) enabled using psutil") logger.info("Note: System-level metrics removed per user request") - + except ImportError: logger.warning("psutil not installed, skipping system metrics. Install with: pip install psutil") except Exception as e: logger.warning(f"Failed to setup system metrics: {e}") - + _metrics_initialized = True logger.info("Meter provider configured with JSON file exporter") logger.info("OpenTelemetry metrics setup completed") - + except Exception as e: logger.error(f"Failed to setup metrics: {e}", exc_info=True) # Don't raise exception, allow application to continue (metrics are optional) @@ -804,7 +804,7 @@ def _get_memory_usage_bytes(callback_options): def _get_process_cpu_usage(callback_options): """Callback to get process CPU usage percentage. - + Note: psutil.Process().cpu_percent(interval=None) requires two calls to calculate. First call returns 0.0, subsequent calls return the percentage since last call. We use interval=0.1 to get immediate accurate reading. @@ -813,13 +813,13 @@ def _get_process_cpu_usage(callback_options): try: import psutil from opentelemetry.metrics import Observation - + if _process_obj is None: _process_obj = psutil.Process() # First call to initialize, returns 0.0 _process_obj.cpu_percent(interval=0.5) return [Observation(0.0)] - + # Subsequent calls return actual CPU usage # Use interval=0.1 for more accurate reading cpu_percent = _process_obj.cpu_percent(interval=0.5) @@ -839,4 +839,3 @@ def _get_process_memory_usage(callback_options): return [Observation(memory_info.rss)] # Resident Set Size except Exception: return [] - diff --git a/examples/adk/image-scoring/image_scoring/tools/loop_condition_tool.py b/examples/adk/image-scoring/image_scoring/tools/loop_condition_tool.py index 26fc6aa..b64e684 100644 --- a/examples/adk/image-scoring/image_scoring/tools/loop_condition_tool.py +++ b/examples/adk/image-scoring/image_scoring/tools/loop_condition_tool.py @@ -5,7 +5,7 @@ def check_condition_and_escalate_tool(tool_context: ToolContext) -> dict: """Checks the loop termination condition and escalates if met or max count reached.""" - + # Increment loop iteration count using state current_loop_count = tool_context.state.get("loop_iteration", 0) diff --git a/examples/adk/marketing-agency/demo_html/about.html b/examples/adk/marketing-agency/demo_html/about.html index 4a47efb..36f5786 100644 --- a/examples/adk/marketing-agency/demo_html/about.html +++ b/examples/adk/marketing-agency/demo_html/about.html @@ -93,4 +93,4 @@

Customer Joy

- \ No newline at end of file + diff --git a/examples/adk/marketing-agency/demo_html/contact.html b/examples/adk/marketing-agency/demo_html/contact.html index 344b60b..a444c2f 100644 --- a/examples/adk/marketing-agency/demo_html/contact.html +++ b/examples/adk/marketing-agency/demo_html/contact.html @@ -101,4 +101,4 @@

Business Hours

- \ No newline at end of file + diff --git a/examples/adk/marketing-agency/demo_html/gallery.html b/examples/adk/marketing-agency/demo_html/gallery.html index a2d8896..fda9cf4 100644 --- a/examples/adk/marketing-agency/demo_html/gallery.html +++ b/examples/adk/marketing-agency/demo_html/gallery.html @@ -129,4 +129,4 @@

[Assorted Cupcakes]

- \ No newline at end of file + diff --git a/examples/adk/marketing-agency/demo_html/index.html b/examples/adk/marketing-agency/demo_html/index.html index c241082..bbd6162 100644 --- a/examples/adk/marketing-agency/demo_html/index.html +++ b/examples/adk/marketing-agency/demo_html/index.html @@ -118,4 +118,4 @@

Ready for the Perfect Cake?

- \ No newline at end of file + diff --git a/examples/adk/marketing-agency/demo_html/script.js b/examples/adk/marketing-agency/demo_html/script.js index ccbd729..affdb92 100644 --- a/examples/adk/marketing-agency/demo_html/script.js +++ b/examples/adk/marketing-agency/demo_html/script.js @@ -55,4 +55,4 @@ document.addEventListener('DOMContentLoaded', function() { }); }); -}); \ No newline at end of file +}); diff --git a/examples/adk/marketing-agency/demo_html/style.css b/examples/adk/marketing-agency/demo_html/style.css index 4f84de0..19037fc 100644 --- a/examples/adk/marketing-agency/demo_html/style.css +++ b/examples/adk/marketing-agency/demo_html/style.css @@ -561,4 +561,4 @@ img { .cake-grid { grid-template-columns: 1fr; } /* Single column on very small screens */ .cta-button, .button-secondary { width: 100%; margin: 5px 0; } /* Full width buttons */ .cta-section .button-secondary { margin-top: 10px; } -} \ No newline at end of file +} diff --git a/examples/adk/marketing-agency/example_input1.md b/examples/adk/marketing-agency/example_input1.md index 327a1a1..cde1e2b 100644 --- a/examples/adk/marketing-agency/example_input1.md +++ b/examples/adk/marketing-agency/example_input1.md @@ -6,14 +6,14 @@ hange or be removed in future versions without notice. It may introduce breaking /home/tiem/miniconda3/envs/py313/lib/python3.13/site-packages/google/adk/auth/credential_service/in_memory_credential_service.py:33: UserWarning: [EXPERIMENTAL] BaseCredentialService : This feature is experimental and may change or be removed in future versions without notice. It may introduce breaking changes at any time. super().__init__() Running agent marketing_coordinator, type exit to exit. -/home/tiem/miniconda3/envs/py313/lib/python3.13/site-packages/google/adk/cli/cli.py:98: UserWarning: [EXPERIMENTAL] App: This feature is experimental and may change or be removed in +/home/tiem/miniconda3/envs/py313/lib/python3.13/site-packages/google/adk/cli/cli.py:98: UserWarning: [EXPERIMENTAL] App: This feature is experimental and may change or be removed in future versions without notice. It may introduce breaking changes at any time. else App(name=session.app_name, root_agent=root_agent_or_app) [user]: who are you [marketing_coordinator]: I am a marketing expert powered by the Google Ads Development Kit (ADK). My purpose is to help you establish a strong online presence and connect effectively - with your audience. I can guide you through the process of defining your digital identity, which includes choosing the perfect domain name, creating a professional website, devising online marketing strategies, designing a memorable logo, and even producing engaging short videos. + with your audience. I can guide you through the process of defining your digital identity, which includes choosing the perfect domain name, creating a professional website, devising online marketing strategies, designing a memorable logo, and even producing engaging short videos. [user]: i want to sell cakes organic cakes [marketing_coordinator]: Excellent! Let's start by choosing the perfect domain name for your organic cake business. I've used the `domain_create` subagent to generate a list of avail -able domain names based on your keywords "organic cakes". +able domain names based on your keywords "organic cakes". domain_create_agent tool reported: 1. PureCakeCreations.com 2. EarthlyConfections.com 3. GreenLeafCakery.com @@ -40,13 +40,13 @@ It seems the `website_create` subagent needs a bit more information to build the * **Brand/Project Name:** What is the name of your cake business? * **Primary goal/purpose of the website:** What is the main thing you want visitors to do on your website (e.g., purchase cakes, view a menu, contact you for custom orders)? * **Key services, products, or information to be featured:** What are the most important things you want to showcase on your website (e.g., types of organic cakes, information abou -t your ingredients, your baking process)? +t your ingredients, your baking process)? [user]: The Brand is "Antonio's perfect cakes" the primary target is young generation for 20 years old and i need an image gallery and a shopping gallery [marketing_coordinator]: Thank you for providing those details. That gives me a much clearer picture of your vision. Let's try this again. I will now call the `website_create` subagent with all the necessary information to create your website for "Antonio's perfect cakes." [marketing_coordinator]: website_create_agent tool reported: Of course. Based on your requirements, here is a complete, modern, and responsive multi-page website for "Antonio's perfe -ct cakes." +ct cakes." The design is clean, visually-driven, and mobile-first to appeal to a younger audience. It features a warm, appetizing color palette and focuses on high-quality imagery. You can copy and paste the code for each file (`index.html`, `gallery.html`, `shop.html`, `about.html`, `contact.html`, `style.css`, and `script.js`) into your project. @@ -310,7 +310,7 @@ e locally sourced flour, organic eggs, real butter, and fresh fruit to create ca <div class="container"> <div class="product-grid"> <!-- User: Populate this grid with your products. Add/remove as needed. --> - + <!-- Product 1 --> <div class="card product-card"> <img src="https://source.unsplash.com/random/400x400/?chocolate-cake" alt="Chocolate Fudge Cake" class="card-img"> @@ -343,7 +343,7 @@ e locally sourced flour, organic eggs, real butter, and fresh fruit to create ca <button class="btn btn-primary">Add to Cart</button> </div> </div> - + <!-- Product 4 --> <div class="card product-card"> <img src="https://source.unsplash.com/random/400x400/?red-velvet-cake" alt="Red Velvet Cake" class="card-img"> @@ -453,7 +453,7 @@ so made with ingredients you can trust. Every cake is a reflection of this passi </div> </div> </section> - + <section class="section-padding light-bg"> <div class="container"> <h2 class="section-title">Our Clean Ingredient Philosophy</h2> @@ -998,7 +998,7 @@ a:hover { @media (max-width: 992px) { .hero-title { font-size: 3rem; } .page-title { font-size: 2.5rem; } - + .two-col-layout { grid-template-columns: 1fr; } @@ -1023,7 +1023,7 @@ a:hover { display: block; z-index: 1001; } - + .main-nav { position: fixed; top: 0; @@ -1040,7 +1040,7 @@ a:hover { .main-nav.active { transform: translateX(0); } - + .nav-links { flex-direction: column; text-align: center; @@ -1106,11 +1106,11 @@ It appears the `marketing_create` subagent also requires more specific informati * **Primary Marketing Goal(s):** What are your specific business goals? For example, do you want to increase online sales by a certain percentage, generate local delivery orders in a specific area, or build an email list? * **Target Audience Profile:** Can you describe your ideal customer in more detail? What are their demographics, lifestyle, values, and interests? [user]: key benefits is to have fresh ingredients and people are located in Europe, lifestyle is nomadic -[marketing_coordinator]: Thank you for the additional details. A target audience with a nomadic lifestyle in Europe is a unique and exciting niche! This gives the marketing subagent -a much better direction. +[marketing_coordinator]: Thank you for the additional details. A target audience with a nomadic lifestyle in Europe is a unique and exciting niche! This gives the marketing subagent +a much better direction. I will now call the `marketing_create` subagent again with this new information to craft a specialized marketing plan for "Antonio's perfect cakes." Given your audience, the strategy - will likely focus heavily on digital channels and social media that resonate with a travel-oriented lifestyle. [marketing_coordinator]: marketing_create_agent tool reported: Of course. Based on the information provided, here is a comprehensive marketing strategy for "Antonio's perfect cakes" -designed to build brand awareness and drive initial online sales among the nomadic community in Europe. + will likely focus heavily on digital channels and social media that resonate with a travel-oriented lifestyle. [marketing_coordinator]: marketing_create_agent tool reported: Of course. Based on the information provided, here is a comprehensive marketing strategy for "Antonio's perfect cakes" +designed to build brand awareness and drive initial online sales among the nomadic community in Europe. *** ### **Marketing Strategy for: Antonio's perfect cakes** @@ -1120,9 +1120,9 @@ designed to build brand awareness and drive initial online sales among the nomad ### **1. Executive Summary** This marketing strategy is designed for **Antonio's perfect cakes**, an online brand selling cakes made with fresh ingredients via CleanIngredientCakes.com. The primary goal is to ** -build brand awareness** and **drive initial online sales** within a niche target audience: people with a **nomadic lifestyle in Europe**. +build brand awareness** and **drive initial online sales** within a niche target audience: people with a **nomadic lifestyle in Europe**. The core approach is a blend of **Content Marketing and Community Building**. We will position the brand as the go-to provider of premium, convenient treats for modern nomads. The st -rategy focuses on leveraging highly visual social platforms like Instagram, engaging directly with the community in their online hubs (Facebook Groups, Reddit), and creating valuable content that resonates with their lifestyle. Success will be measured by tracking brand awareness metrics (reach, follower growth) and sales performance (conversion rate, number of orders). +rategy focuses on leveraging highly visual social platforms like Instagram, engaging directly with the community in their online hubs (Facebook Groups, Reddit), and creating valuable content that resonates with their lifestyle. Success will be measured by tracking brand awareness metrics (reach, follower growth) and sales performance (conversion rate, number of orders). ### **2. Understanding Your Brief** * **Brand:** Antonio's perfect cakes @@ -1147,7 +1147,7 @@ To effectively reach your audience, we will focus on this key persona: * **Lifestyle:** Tech-savvy and highly connected online. She plans her life through apps and online communities. Celebrates small wins (like landing a new client) but often lac ks the means for elaborate celebrations. * **Pain Points:** Sometimes feels isolated or misses the comforts of home. Wants to indulge in a high-quality treat without the hassle of baking from scratch in a temporary ki tchen. * **Online Behavior:** Spends her time on **Instagram**, **Pinterest**, and **YouTube** for lifestyle and travel inspiration. She is an active member of **Facebook groups** (e. -g., "Digital Nomads in Europe") and **Reddit** subreddits (e.g., r/digitalnomad) for advice and community. +g., "Digital Nomads in Europe") and **Reddit** subreddits (e.g., r/digitalnomad) for advice and community. ### **4. Foundational Analysis** * **Unique Selling Proposition (USP):** While "fresh ingredients" is the benefit, the USP is the story behind it. @@ -1161,7 +1161,7 @@ reness. ### **5. Core Marketing Strategy** * **Overall Approach: Content & Community Marketing** - * **Why:** Your audience isn't looking for a generic product; they're looking for brands that understand and fit into their lifestyle. This approach builds trust and positions + * **Why:** Your audience isn't looking for a generic product; they're looking for brands that understand and fit into their lifestyle. This approach builds trust and positions "Antonio's perfect cakes" as an authentic part of the nomad ecosystem, rather than just another online store. * **Core Messaging & Positioning:** * **Positioning:** The premium, convenient treat for the modern European nomad. * **Core Message:** "Antonio's Perfect Cakes: A taste of homemade luxury, wherever you are." @@ -1180,9 +1180,9 @@ Here is a prioritized mix of channels to achieve your goals. * **Why:** This is the #1 platform where your target audience seeks lifestyle inspiration. It's perfect for showcasing a beautiful product. * **Tactics:** 1. **Aspirational Visuals:** Create high-quality photos and videos of the cakes in "nomadic" settings: on a desk in a stylish Airbnb, being shared in a co-working space, on a sm -all table in a campervan with a scenic backdrop. 2. **Influencer & Creator Collaborations:** Partner with European micro-influencers (<50k followers) in the digital nomad, van-life, and solo travel niches. Their audiences are +all table in a campervan with a scenic backdrop. 2. **Influencer & Creator Collaborations:** Partner with European micro-influencers (<50k followers) in the digital nomad, van-life, and solo travel niches. Their audiences are highly engaged and trust their recommendations. Gift them products in exchange for authentic posts and stories. 3. **User-Generated Content (UGC) Campaign:** Create a branded hashtag (e.g., **#CakesOnTheGo**, **#AntoniosJourney**). Encourage customers to share photos of their cake on thei -r travels for a chance to be featured or win a prize. This provides social proof and free content. +r travels for a chance to be featured or win a prize. This provides social proof and free content. #### **Channel 2: Community Marketing (Facebook Groups & Reddit)** * **Why:** This is where your audience gathers for practical advice and connection. It's the perfect place to build brand awareness organically. @@ -1231,7 +1231,7 @@ ze website traffic and initial sales data to see which channels are performing b Since no budget was specified, this strategy blends low-cost organic efforts with scalable paid options. * **Low Budget (Organic Focus):** Prioritize Community Marketing and organic Instagram content. The main cost will be time and shipping products to a small number of influencers. -* **Medium Budget ($1,000 - $3,000/month):** Execute all of the above. Allocate funds to pay for higher-quality micro-influencer collaborations and run targeted Instagram/Facebook +* **Medium Budget ($1,000 - $3,000/month):** Execute all of the above. Allocate funds to pay for higher-quality micro-influencer collaborations and run targeted Instagram/Facebook ads to audiences who have shown interest in "digital nomadism" or "van life." Boost your best-performing posts. * **High Budget:** Scale all activities, working with larger influencers, running more extensive ad campaigns, and investing heavily in professional video content creation. ### **9. Next Steps** @@ -1243,12 +1243,12 @@ ads to audiences who have shown interest in "digital nomadism" or "van life." Bo --- **Disclaimer:** This strategic outline provides recommendations based on the information provided. Detailed execution plans, content creation, campaign management, and ongoing analys -is will be required for successful implementation. Results will depend on the quality of execution, product experience, and market conditions. +is will be required for successful implementation. Results will depend on the quality of execution, product experience, and market conditions. We have a fantastic, targeted marketing strategy now. The final step is to create a memorable logo that will represent your brand. I will now call the `logo_create` subagent to design a logo for "Antonio's perfect cakes." [marketing_coordinator]: logo_create_agent tool reported: I have created a few logo options for "Antonio's perfect cakes" for you to review. Let me know if you would like to see more - options or if you have any feedback on these designs. + options or if you have any feedback on these designs. Here are the logo designs generated by the `logo_create` subagent. Please let me know which one you prefer, or if you'd like to see more options! [Image of a logo with a stylized cake and the text "Antonio's perfect cakes" in a warm, friendly font] @@ -1260,12 +1260,12 @@ Here are the logo designs generated by the `logo_create` subagent. Please let me We have now completed all the initial steps to establish your powerful online presence. Let's quickly recap everything we've accomplished: 1. **Domain Name:** You selected the memorable and descriptive domain **CleanIngredientCakes.com**. -2. **Professional Website:** We've generated the complete code for a modern, multi-page website for "Antonio's perfect cakes," complete with a homepage, image gallery, and shopping +2. **Professional Website:** We've generated the complete code for a modern, multi-page website for "Antonio's perfect cakes," complete with a homepage, image gallery, and shopping page. 3. **Marketing Strategy:** You have a comprehensive, tailored marketing plan focused on reaching your target audience of European nomads through Instagram, community engagement, and content marketing. 4. **Memorable Logo:** You've chosen a beautiful logo featuring a stylized cake with a warm, friendly font to represent your brand. You now have a solid foundation for your digital identity! You are ready to launch "Antonio's perfect cakes" online and connect with your audience. Congratulations, and I wish you the best of luck with your business! If you need any more help with your marketing efforts, feel free to ask. -[user]: +[user]: Aborted! diff --git a/examples/adk/marketing-agency/example_input2.md b/examples/adk/marketing-agency/example_input2.md index 8f9af53..999f244 100644 --- a/examples/adk/marketing-agency/example_input2.md +++ b/examples/adk/marketing-agency/example_input2.md @@ -6,11 +6,11 @@ hange or be removed in future versions without notice. It may introduce breaking /home/tiem/miniconda3/envs/py313/lib/python3.13/site-packages/google/adk/auth/credential_service/in_memory_credential_service.py:33: UserWarning: [EXPERIMENTAL] BaseCredentialService : This feature is experimental and may change or be removed in future versions without notice. It may introduce breaking changes at any time. super().__init__() Running agent marketing_coordinator, type exit to exit. -/home/tiem/miniconda3/envs/py313/lib/python3.13/site-packages/google/adk/cli/cli.py:98: UserWarning: [EXPERIMENTAL] App: This feature is experimental and may change or be removed in +/home/tiem/miniconda3/envs/py313/lib/python3.13/site-packages/google/adk/cli/cli.py:98: UserWarning: [EXPERIMENTAL] App: This feature is experimental and may change or be removed in future versions without notice. It may introduce breaking changes at any time. else App(name=session.app_name, root_agent=root_agent_or_app) [user]: who are you [marketing_coordinator]: I am a marketing expert, and my purpose is to help you establish a powerful online presence and connect effectively with your audience. I can guide you throu -gh defining your digital identity, from choosing the perfect domain name and crafting a professional website to strategizing online marketing campaigns and designing a memorable logo. +gh defining your digital identity, from choosing the perfect domain name and crafting a professional website to strategizing online marketing campaigns and designing a memorable logo. [user]: i want to sell cakes organic cakes [marketing_coordinator]: Great choice! "TheOrganicOven.com" is a fantastic domain name. @@ -34,8 +34,8 @@ website_create tool reported: To proceed with website creation, please obtain th * Key services, products, or information to be featured I will now call the `website_create` subagent again with these details to generate your website. -[marketing_coordinator]: website_create tool reported: Of course. Here is the complete, responsive, multi-page website for "The Organic Oven." The design is clean, warm, and modern, -reflecting the brand's organic and artisanal nature. +[marketing_coordinator]: website_create tool reported: Of course. Here is the complete, responsive, multi-page website for "The Organic Oven." The design is clean, warm, and modern, +reflecting the brand's organic and artisanal nature. You can copy the content for each file and save it as `index.html`, `products.html`, `about.html`, `contact.html`, `style.css`, and `script.js`. --- @@ -128,7 +128,7 @@ You can copy the content for each file and save it as `index.html`, `products.ht

Taste the Difference

Pure Ingredients, Pure Joy

-

At The Organic Oven, we believe that the best-tasting cakes start with the finest ingredients. That\'s why +

At The Organic Oven, we believe that the best-tasting cakes start with the finest ingredients. That\'s why we\'re committed to using only 100% certified organic flour, sugar, eggs, and dairy. We partner with local farmers to source fresh, seasonal fruits, ensuring every bite is bursting with natural flavor.

Learn About Our Sourcing
@@ -201,7 +201,7 @@ we\'re committed to using only 100% certified organic flour, sugar, eggs, and da

Our Collection of Organic Cakes

-

Each cake is baked to order using the finest organic ingredients. Please note, this is a template; a full e-commerce backend is required for +

Each cake is baked to order using the finest organic ingredients. Please note, this is a template; a full e-commerce backend is required for actual orders.

@@ -209,9 +209,9 @@ actual orders.

- + - +
Decadent Chocolate Fudge Cake @@ -222,7 +222,7 @@ actual orders.

- +
Strawberry Dream Cake @@ -348,7 +348,7 @@ actual orders.

Our Philosophy

-

The Organic Oven was founded on a simple principle: to create beautiful, delicious cakes that are as good for the planet as they are for the soul. Our journey +

The Organic Oven was founded on a simple principle: to create beautiful, delicious cakes that are as good for the planet as they are for the soul. Our journey began in a home kitchen, with a passion for baking and a deep respect for nature\'s pantry.

We refuse to compromise on quality. Every ingredient is thoughtfully chosen for its purity and flavor. We don\'t use artificial colors, flavors, or preservativ es—just honest, wholesome goodness you can taste in every bite.

@@ -451,23 +451,23 @@ es—just honest, wholesome goodness you can taste in every bite.

- +

Contact Information

For custom orders, please provide as much detail as possible in your message.

- + - +

Business Hours

Monday - Friday: 9am - 5pm
Saturday: 10am - 2pm
Sunday: Closed

- +
@@ -976,7 +976,7 @@ body { /* Mobile */ @media (max-width: 768px) { :root { --header-height: 70px; } - + .main-nav { position: fixed; top: 0; @@ -996,21 +996,21 @@ body { .main-nav.active { right: 0; } - + .nav-links { flex-direction: column; align-items: center; gap: 2.5rem; } - + .nav-links a { font-size: 1.2rem; } - + .nav-toggle { display: block; } - + .nav-open .hamburger { transform: rotate(45deg); } @@ -1026,7 +1026,7 @@ body { .hero { height: 80vh; } h1, .hero-title { font-size: 2.5rem; } h2, .page-title { font-size: 2rem; } - + .main-footer .container { flex-direction: column; gap: 1rem; @@ -1050,7 +1050,7 @@ document.addEventListener('DOMContentLoaded', () => { navToggle.addEventListener('click', () => { mainNav.classList.toggle('active'); navToggle.classList.toggle('nav-open'); - + // Optional: Prevent body scrolling when menu is open if (mainNav.classList.contains('active')) { body.style.overflow = 'hidden'; @@ -1073,7 +1073,7 @@ y features, benefits, and what makes it unique (USP). I will now call the `marketing_create` subagent again with these details to generate your comprehensive marketing strategy. [marketing_coordinator]: marketing_create tool reported: Of course. Based on the information you've provided, here is a comprehensive marketing strategy designed to establish The Org -anic Oven as a premium brand, drive online sales, and build a thriving community. +anic Oven as a premium brand, drive online sales, and build a thriving community. *** ### **Marketing Strategy for The Organic Oven** @@ -1083,7 +1083,7 @@ anic Oven as a premium brand, drive online sales, and build a thriving community ### **Executive Summary** This strategy is designed to position The Organic Oven as the premier online destination for luxuriously handcrafted, authentically organic cakes. The core approach is a blend of **C -ontent Marketing, Community Building, and Brand-Building**, focusing on attracting your target audience through compelling visual storytelling and shared values. We will leverage highly visual platforms like Instagram and Pinterest to showcase the artisanal quality of your cakes and build trust through transparency around your locally sourced, organic ingredients. Key pillars of this strategy include creating stunning visual content, telling the stories behind your ingredients, and fostering a community of health-conscious foodies. Success will be measured by tracking online sales, brand engagement metrics, and community growth. +ontent Marketing, Community Building, and Brand-Building**, focusing on attracting your target audience through compelling visual storytelling and shared values. We will leverage highly visual platforms like Instagram and Pinterest to showcase the artisanal quality of your cakes and build trust through transparency around your locally sourced, organic ingredients. Key pillars of this strategy include creating stunning visual content, telling the stories behind your ingredients, and fostering a community of health-conscious foodies. Success will be measured by tracking online sales, brand engagement metrics, and community growth. ### **1. Understanding Your Brief** * **Brand:** The Organic Oven @@ -1113,10 +1113,10 @@ enwash." * **Psychographics:** * **Values:** Quality over quantity, thoughtful gestures, and convenience. He wants to give gifts that are unique and meaningful. * **Lifestyle:** Busy professional who values his time. He appreciates a seamless online shopping experience. - * **Pain Points:** Tired of generic gift options like flowers or mass-produced gift baskets. He struggles to find high-end, special gifts for birthdays, anniversaries, or as a + * **Pain Points:** Tired of generic gift options like flowers or mass-produced gift baskets. He struggles to find high-end, special gifts for birthdays, anniversaries, or as a "thank you" for an important client. * **Motivations:** Wants to impress the recipient with a premium, delicious, and special gift that shows he cares about their well-being. -* **Online Behavior:** Searches online for "unique luxury food gifts" or "organic cake delivery." Relies on a professional-looking website and clear product descriptions to make a -decision. +* **Online Behavior:** Searches online for "unique luxury food gifts" or "organic cake delivery." Relies on a professional-looking website and clear product descriptions to make a +decision. ### **3. Foundational Analysis** * **Unique Selling Proposition (USP): "Indulgence Without Compromise."** This is your core differentiator. While competitors may be organic *or* artisanal, The Organic Oven is both @@ -1125,13 +1125,13 @@ decision. * **Weaknesses:** Higher price point may be a barrier for some, perishable nature creates shipping complexities, and as a new brand, there is no existing recognition. * **Opportunities:** The growing market demand for organic and sustainable products, strong potential for visual storytelling on social media, ability to build a loyal communit y around shared values. * **Threats:** Direct competition from other online bakeries, local artisan bakeries, and large commercial brands with "organic" lines. Rising ingredient costs and supply chain - issues. + issues. ### **4. Core Marketing Strategy** * **Overall Approach: Inbound Content & Community Marketing.** We will attract customers by creating valuable, beautiful content that resonates with their interests and values, rat -her than relying solely on traditional advertising. This builds a durable brand and fosters loyalty. -* **Positioning Statement:** For health-conscious consumers and discerning gift-givers, **The Organic Oven is the premier online bakery that delivers luxuriously handcrafted cakes -made from authentically local and organic ingredients**, offering an indulgent experience you can feel good about. +her than relying solely on traditional advertising. This builds a durable brand and fosters loyalty. +* **Positioning Statement:** For health-conscious consumers and discerning gift-givers, **The Organic Oven is the premier online bakery that delivers luxuriously handcrafted cakes +made from authentically local and organic ingredients**, offering an indulgent experience you can feel good about. * **Core Messaging:** "Artisanal Baking. Purely Organic. Delivered." This message is short, premium, and communicates your key value propositions. * **Strategic Pillars:** @@ -1146,9 +1146,9 @@ e local farms you partner with. * **Tactics:** * **High-Quality Reels:** Create short, aesthetically pleasing videos of the baking process, frosting a cake in slow-motion, or slicing a finished cake. Use trending audio to i ncrease reach. * **"Farm to Cake" Stories:** Use Instagram Stories and Highlights to introduce your local ingredient suppliers (e.g., "Meet the farmer who grows our organic strawberries"). - * **User-Generated Content (UGC):** Encourage customers to post photos of their cakes using a specific hashtag (e.g., #TheOrganicOven). Feature the best photos on your feed to + * **User-Generated Content (UGC):** Encourage customers to post photos of their cakes using a specific hashtag (e.g., #TheOrganicOven). Feature the best photos on your feed to build social proof and community. * **Influencer Collaborations:** Partner with micro-influencers in the foodie, wellness, and sustainable living niches. Gift them a cake in exchange for an authentic review and - posts. + posts. **2. Pinterest (Primary Channel for Driving Traffic & Sales)** * **Why:** Your audience (Elena and Mark) uses this platform for planning events and discovering products. Unlike Instagram, Pinterest content has a much longer lifespan. * **Tactics:** @@ -1215,7 +1215,7 @@ r "local artisan bakery online." 3. **Content Calendar:** Develop a one-month content calendar to plan your posts, emails, and blog topics in advance. **Disclaimer:** This strategic outline provides recommendations based on the information provided. Detailed execution plans, ongoing content creation, daily community management, and - continuous performance analysis will be required for successful implementation. Results may vary based on execution quality, market conditions, and budget allocation. + continuous performance analysis will be required for successful implementation. Results may vary based on execution quality, market conditions, and budget allocation. Now, let's bring your brand to life with a memorable logo. I will now call the `logo_create` subagent to design a logo for "TheOrganicOven.com". [marketing_coordinator]: logo_create tool reported: I am sorry, I cannot fulfill this request. I do not have the ability to access URLs. Is there anything else I can help with? @@ -1223,7 +1223,7 @@ It seems that the `logo_create` tool is unable to access the website directly to at you envision for your logo? For example, you could describe the style (e.g., modern, rustic, minimalist), colors, or any imagery you'd like to see (e.g., a whisk, a wheat stalk, a cake). This information will help me generate a logo that truly represents your brand. [user]: Image of a logo with a stylized cake and the text "Antonio's perfect cakes" in a warm, friendly font [marketing_coordinator]: Of course. I will now call the `logo_create` subagent to generate the logo with your specified design. [marketing_coordinator]: logo_create tool reported: Of course! I have created the logo for "Antonio's perfect cakes" and it's now available in your artifacts as `image.png`. Let me k -now if you have any questions about it. +now if you have any questions about it. Excellent! Your logo is now ready. We have successfully completed all the steps to establish your online presence: @@ -1234,7 +1234,7 @@ We have successfully completed all the steps to establish your online presence: 4. **Designed a memorable logo:** The logo image is now available for you to use. Congratulations on defining your digital identity! You now have the foundational assets to launch "The Organic Oven." I recommend reviewing your new website code, marketing strategy, - and logo to ensure they align perfectly with your vision. + and logo to ensure they align perfectly with your vision. If you have any more questions or need further assistance as you begin to implement your marketing plan, please let me know -[user]: +[user]: Aborted! diff --git a/examples/adk/marketing-agency/marketing_agency/agent.py b/examples/adk/marketing-agency/marketing_agency/agent.py index 0b3f3d8..1599193 100644 --- a/examples/adk/marketing-agency/marketing_agency/agent.py +++ b/examples/adk/marketing-agency/marketing_agency/agent.py @@ -23,7 +23,7 @@ from .sub_agents.marketing_create import marketing_create_agent from .sub_agents.website_create import website_create_agent -MODEL = "gemini-2.5-pro" +MODEL = "gemini-2.5-pro" marketing_coordinator = LlmAgent( name="marketing_coordinator", diff --git a/examples/adk/marketing-agency/marketing_agency/prompt.py b/examples/adk/marketing-agency/marketing_agency/prompt.py index 8dbc8d7..063f86c 100644 --- a/examples/adk/marketing-agency/marketing_agency/prompt.py +++ b/examples/adk/marketing-agency/marketing_agency/prompt.py @@ -22,13 +22,13 @@ 1. **Choosing the perfect domain name (Subagent: domain_create)** * **Input:** Ask the user for keywords relevant to their brand. * **Action:** Call the `domain_create` subagent with the user's keywords. - * **Expected Output:** The `domain_create` subagent should return a list of at least 10 available (unassigned) domain names. - These names should be creative and have the potential to attract users, reflecting the brand's unique identity. + * **Expected Output:** The `domain_create` subagent should return a list of at least 10 available (unassigned) domain names. + These names should be creative and have the potential to attract users, reflecting the brand's unique identity. Present this list to the user and ask them to select their preferred domain. 2. **Crafting a professional website (Subagent: website_create)** * **Input:** The domain name chosen by the user in the previous step. - * **Action:** Call the `website_create` subagent with the user-selected domain name + * **Action:** Call the `website_create` subagent with the user-selected domain name * **Expected Output:** The `website_create` subagent should generate a fully functional website based on the chosen domain. 3. **Strategizing online marketing campaigns (Subagent: marketing_create)** @@ -50,8 +50,7 @@ ** The name of the subagent tool you used. ** The exact result or output provided by that subagent tool. * Present this information using the format: [Tool Name] tool reported: [Exact Result From Tool] -** Example: If a subagent tool named PolicyValidator returns the result +** Example: If a subagent tool named PolicyValidator returns the result 'Policy compliance confirmed.', your response must include the phrase: PolicyValidator tool reported: Policy compliance confirmed. """ - diff --git a/examples/adk/marketing-agency/marketing_agency/sub_agents/domain_create/agent.py b/examples/adk/marketing-agency/marketing_agency/sub_agents/domain_create/agent.py index d022927..fdb97c9 100644 --- a/examples/adk/marketing-agency/marketing_agency/sub_agents/domain_create/agent.py +++ b/examples/adk/marketing-agency/marketing_agency/sub_agents/domain_create/agent.py @@ -19,7 +19,7 @@ from . import prompt -MODEL = "gemini-2.5-pro" +MODEL = "gemini-2.5-pro" domain_create_agent = Agent( model=MODEL, diff --git a/examples/adk/marketing-agency/marketing_agency/sub_agents/logo_create/agent.py b/examples/adk/marketing-agency/marketing_agency/sub_agents/logo_create/agent.py index ee443c1..dc351f3 100644 --- a/examples/adk/marketing-agency/marketing_agency/sub_agents/logo_create/agent.py +++ b/examples/adk/marketing-agency/marketing_agency/sub_agents/logo_create/agent.py @@ -35,11 +35,11 @@ async def generate_image(img_prompt: str, tool_context: "ToolContext"): if not response.generated_images: return {"status": "failed"} image_bytes = response.generated_images[0].image.image_bytes - + # Save locally for user visibility with open("generated_logo.png", "wb") as f: f.write(image_bytes) - + await tool_context.save_artifact( "image.png", types.Part.from_bytes(data=image_bytes, mime_type="image/png"), diff --git a/examples/adk/marketing-agency/marketing_agency/sub_agents/marketing_create/agent.py b/examples/adk/marketing-agency/marketing_agency/sub_agents/marketing_create/agent.py index f09dd71..40f3ca2 100644 --- a/examples/adk/marketing-agency/marketing_agency/sub_agents/marketing_create/agent.py +++ b/examples/adk/marketing-agency/marketing_agency/sub_agents/marketing_create/agent.py @@ -18,7 +18,7 @@ from . import prompt -MODEL = "gemini-2.5-pro" +MODEL = "gemini-2.5-pro" marketing_create_agent = Agent( model=MODEL, diff --git a/examples/adk/marketing-agency/marketing_agency/sub_agents/website_create/agent.py b/examples/adk/marketing-agency/marketing_agency/sub_agents/website_create/agent.py index 83b6a89..f72b2e1 100644 --- a/examples/adk/marketing-agency/marketing_agency/sub_agents/website_create/agent.py +++ b/examples/adk/marketing-agency/marketing_agency/sub_agents/website_create/agent.py @@ -18,7 +18,7 @@ from . import prompt -MODEL = "gemini-2.5-pro" +MODEL = "gemini-2.5-pro" website_create_agent = Agent( model=MODEL, diff --git a/examples/adk/marketing-agency/marketing_agency/telemetry_setup.py b/examples/adk/marketing-agency/marketing_agency/telemetry_setup.py index 852a8b8..cfd777e 100644 --- a/examples/adk/marketing-agency/marketing_agency/telemetry_setup.py +++ b/examples/adk/marketing-agency/marketing_agency/telemetry_setup.py @@ -46,10 +46,10 @@ class JsonFileMetricExporter(MetricExporter): """Exporter that writes metrics to local JSON files.""" - + def __init__(self, file_path: str): """Initialize JSON file exporter. - + Args: file_path: Path to JSON file """ @@ -59,54 +59,54 @@ def __init__(self, file_path: str): # Use None to accept defaults (same as ConsoleMetricExporter) self._preferred_temporality = None self._preferred_aggregation = None - + def export(self, metrics_data, timeout_millis: float = 10000, **kwargs): """Export metrics to JSON file. - + Args: metrics_data: Metrics data to export timeout_millis: Maximum time to wait for export (not used for file export) **kwargs: Additional arguments - + Returns: MetricExportResult.SUCCESS """ try: # Convert metrics to serializable format metric_records = [] - + # Extract metric data from the metrics_data object if hasattr(metrics_data, 'resource_metrics'): for resource_metric in metrics_data.resource_metrics: resource_attrs = dict(resource_metric.resource.attributes) if resource_metric.resource and resource_metric.resource.attributes else {} - + for scope_metric in resource_metric.scope_metrics: scope_name = scope_metric.scope.name if scope_metric.scope else "unknown" - + for metric in scope_metric.metrics: metric_name = metric.name metric_description = metric.description if hasattr(metric, 'description') else None metric_unit = metric.unit if hasattr(metric, 'unit') else None - + # Extract data points data_points = [] if hasattr(metric, 'data') and hasattr(metric.data, 'data_points'): for data_point in metric.data.data_points: point_attrs = dict(data_point.attributes) if hasattr(data_point, 'attributes') and data_point.attributes else {} - + # Add service name as agent identifier if not present if "agent.name" not in point_attrs: # Try to infer from resource attributes or use service name agent_name = resource_attrs.get("service.name", "marketing-agency") point_attrs["agent.name"] = agent_name - + point_dict = { "value": data_point.value if hasattr(data_point, 'value') else None, "timestamp": data_point.time_unix_nano if hasattr(data_point, 'time_unix_nano') else None, "attributes": point_attrs, } data_points.append(point_dict) - + metric_record = { "timestamp": datetime.now().isoformat(), "metric_name": metric_name, @@ -119,31 +119,31 @@ def export(self, metrics_data, timeout_millis: float = 10000, **kwargs): "scope": scope_name, } metric_records.append(metric_record) - + # Append to file (supports incremental writes) if metric_records: with open(self.file_path, "a", encoding="utf-8") as f: for record in metric_records: f.write(json.dumps(record, default=str) + "\n") - + return MetricExportResult.SUCCESS except Exception as e: logger.error(f"Failed to export metrics to {self.file_path}: {e}", exc_info=True) return MetricExportResult.FAILURE - + def force_flush(self, timeout_millis: int = 30000, **kwargs): """Force flush any pending metrics. - + Args: timeout_millis: Maximum time to wait for flush. **kwargs: Additional arguments. """ # For file exporter, data is written immediately, so no explicit flush needed pass - + def shutdown(self, timeout_millis: int = 30000, **kwargs): """Shutdown the exporter. - + Args: timeout_millis: Maximum time to wait for shutdown. **kwargs: Additional arguments. @@ -153,23 +153,23 @@ def shutdown(self, timeout_millis: int = 30000, **kwargs): class JsonFileSpanExporter(SpanExporter): """Exporter that writes spans to local JSON files.""" - + def __init__(self, file_path: str): """Initialize JSON file exporter. - + Args: file_path: Path to JSON file """ self.file_path = Path(file_path) self.file_path.parent.mkdir(parents=True, exist_ok=True) self.spans = [] - + def export(self, spans): """Export spans to JSON file. - + Args: spans: List of spans to export - + Returns: SpanExportResult.SUCCESS """ @@ -183,7 +183,7 @@ def export(self, spans): parent_span_id = format(span.parent.span_id, "016x") elif hasattr(span, "parent_context") and span.parent_context: parent_span_id = format(span.parent_context.span_id, "016x") - + # Extract agent name from span attributes or name for better identification attributes = dict(span.attributes) if span.attributes else {} agent_name = ( @@ -192,11 +192,11 @@ def export(self, spans): (span.name.split()[-1] if "invoke_agent" in span.name and len(span.name.split()) > 1 else None) or "unknown" ) - + # Add agent_name to attributes if not already present (for backward compatibility) if "agent.name" not in attributes and agent_name != "unknown": attributes["agent.name"] = agent_name - + # Ensure gen_ai.agent.name is present (required by template) # For call_llm spans, try to extract from gcp.vertex.agent.llm_request labels if "gen_ai.agent.name" not in attributes: @@ -215,7 +215,7 @@ def export(self, spans): agent_name = adk_agent_name except (json.JSONDecodeError, KeyError, TypeError): pass - + # If still not found, use extracted agent_name or agent.name if "gen_ai.agent.name" not in attributes: if agent_name != "unknown": @@ -225,7 +225,7 @@ def export(self, spans): else: # Field must exist per template, set to null if not available attributes["gen_ai.agent.name"] = None - + # Ensure LLM usage fields are present (default to 0 if missing) if "gen_ai.usage.input_tokens" not in attributes: attributes["gen_ai.usage.input_tokens"] = 0 @@ -236,14 +236,14 @@ def export(self, spans): input_tokens = attributes.get("gen_ai.usage.input_tokens", 0) or 0 output_tokens = attributes.get("gen_ai.usage.output_tokens", 0) or 0 attributes["gen_ai.usage.total_tokens"] = input_tokens + output_tokens - + # Add LLM/MCP call counts (default to 0 if missing) if "gen_ai.llm.call.count" not in attributes: # Set to 1 if this is a call_llm span, otherwise 0 attributes["gen_ai.llm.call.count"] = 1 if "call_llm" in span.name else 0 if "gen_ai.mcp.call.count" not in attributes: attributes["gen_ai.mcp.call.count"] = 0 - + # Ensure gen_ai.operation.name is present (required by template) if "gen_ai.operation.name" not in attributes: if "call_llm" in span.name: @@ -255,7 +255,7 @@ def export(self, spans): elif span.name.lower() == "invocation": # invocation spans are top-level entry points attributes["gen_ai.operation.name"] = "invocation" - + # Ensure gen_ai.conversation.id is present (required by template) # Copy from session_id if available, otherwise set to null if "gen_ai.conversation.id" not in attributes: @@ -265,14 +265,14 @@ def export(self, spans): else: # Field must exist per template, set to null if not available attributes["gen_ai.conversation.id"] = None - + # Add missing optional fields with default values (per template) # gen_ai.agent.description - optional, only in invoke_agent spans if "gen_ai.agent.description" not in attributes: # Try to extract from system_instruction or other sources # For now, set to empty string if not available attributes["gen_ai.agent.description"] = "" - + # gen_ai.tool.* fields - only for execute_tool spans if "gen_ai.tool.name" not in attributes: attributes["gen_ai.tool.name"] = "" @@ -282,31 +282,31 @@ def export(self, spans): attributes["gen_ai.tool.call.id"] = "" if "gen_ai.tool.description" not in attributes: attributes["gen_ai.tool.description"] = "" - + # gcp.vertex.agent.tool_* fields - optional if "gcp.vertex.agent.tool_call_args" not in attributes: attributes["gcp.vertex.agent.tool_call_args"] = "" if "gcp.vertex.agent.tool_response" not in attributes: attributes["gcp.vertex.agent.tool_response"] = "" - + # mcp.* fields - optional, not applicable for Vertex AI if "mcp.server" not in attributes: attributes["mcp.server"] = "" if "mcp.tool" not in attributes: attributes["mcp.tool"] = "" - + # agent.log - optional if "agent.log" not in attributes: attributes["agent.log"] = "" - + # Note: CPU/memory metrics are collected separately in metrics JSONL file, # so we don't add them to span attributes (per template guidelines) - + # Calculate communication metrics for agent-to-agent calls communication_metrics = {} input_size = 0 output_size = 0 - + # Check if this is agent-to-agent communication # Only AgentTool calls are agent communication, not FunctionTool calls is_agent_comm = False @@ -323,54 +323,54 @@ def export(self, spans): elif "invoke_agent" in span.name: # invoke_agent spans are entry points, not agent-to-agent communication is_agent_comm = False - + if is_agent_comm: communication_metrics["is_agent_communication"] = True - + # For execute_tool spans (AgentTool): measure actual tool call arguments and responses if "execute_tool" in span.name: # Input: tool call arguments (actual data passed to the agent) tool_call_args = attributes.get("gcp.vertex.agent.tool_call_args", "") if tool_call_args and isinstance(tool_call_args, str) and tool_call_args != "{}": input_size = len(tool_call_args.encode('utf-8')) - + # Output: tool response (actual data returned from the agent) tool_response = attributes.get("gcp.vertex.agent.tool_response", "") if tool_response and isinstance(tool_response, str) and tool_response != "{}": output_size = len(tool_response.encode('utf-8')) - + # For call_llm spans: measure actual LLM request and response elif "call_llm" in span.name: # Input: LLM request (actual data sent to LLM) llm_request = attributes.get("gcp.vertex.agent.llm_request", "") if llm_request and isinstance(llm_request, str) and llm_request != "{}": input_size = len(llm_request.encode('utf-8')) - + # Output: LLM response (actual data received from LLM) llm_response = attributes.get("gcp.vertex.agent.llm_response", "") if llm_response and isinstance(llm_response, str) and llm_response != "{}": output_size = len(llm_response.encode('utf-8')) - + # Store communication metrics and add to attributes for template compliance # Template shows these fields, but per README: "If a field doesn't apply, omit it" # So we only add them when there's actual communication data # Ensure is_in_process_call is always present (default false) if "is_in_process_call" not in communication_metrics: communication_metrics["is_in_process_call"] = False - + if input_size > 0: communication_metrics["input_message_size_bytes"] = input_size attributes["communication.input_message_size_bytes"] = input_size - + if output_size > 0: communication_metrics["output_message_size_bytes"] = output_size attributes["communication.output_message_size_bytes"] = output_size - + if input_size > 0 or output_size > 0: total_size = input_size + output_size communication_metrics["total_message_size_bytes"] = total_size attributes["communication.total_message_size_bytes"] = total_size - + # Get span kind span_kind = "INTERNAL" # Default if hasattr(span, "kind"): @@ -379,7 +379,7 @@ def export(self, spans): span_kind = SpanKind(kind_value).name if kind_value is not None else "INTERNAL" except (ValueError, AttributeError): span_kind = "INTERNAL" - + # Get resource attributes and add host.name if available resource_attrs = dict(span.resource.attributes) if span.resource and span.resource.attributes else {} if "host.name" not in resource_attrs: @@ -389,7 +389,7 @@ def export(self, spans): resource_attrs["host.name"] = hostname except Exception: pass # host.name is optional - + span_dict = { "trace_id": format(span.context.trace_id, "032x"), "span_id": format(span.context.span_id, "016x"), @@ -419,17 +419,17 @@ def export(self, spans): }, } span_data.append(span_dict) - + # Append to file (supports incremental writes) with open(self.file_path, "a", encoding="utf-8") as f: for span_dict in span_data: f.write(json.dumps(span_dict, default=str) + "\n") - + return SpanExportResult.SUCCESS except Exception as e: logger.error(f"Failed to export spans to {self.file_path}: {e}", exc_info=True) return SpanExportResult.FAILURE - + def shutdown(self): """Shutdown the exporter.""" pass @@ -437,17 +437,17 @@ def shutdown(self): def setup_tracing(service_name: Optional[str] = None, trace_file: Optional[str] = None): """Setup OpenTelemetry tracing with local JSON file export. - + This function configures OpenTelemetry to automatically collect telemetry data from ADK agents. The instrumentation is automatic - ADK and OpenTelemetry instrumentation libraries will automatically create spans for: - + - Agent invocations (invoke_agent) - LLM calls (call_llm) with model, tokens, and request/response data - Tool/function executions (execute_tool) - A2A server operations (request handling, event queue operations) - HTTP requests between agents - + The collected data includes: - Trace IDs and Span IDs for distributed tracing - Timing information (start time, end time, duration) @@ -455,7 +455,7 @@ def setup_tracing(service_name: Optional[str] = None, trace_file: Optional[str] - Attributes (gen_ai.system, gen_ai.request.model, gen_ai.usage.*, etc.) - Events (exceptions, custom events) - Resource attributes (service.name, service.version, etc.) - + Args: service_name: Service name, defaults to OTEL_SERVICE_NAME env var or "adk-agent" trace_file: Trace file path, defaults to OTEL_TRACE_FILE env var or auto-generated @@ -464,28 +464,28 @@ def setup_tracing(service_name: Optional[str] = None, trace_file: Optional[str] # Get configuration from environment variables service_name = service_name or os.getenv("OTEL_SERVICE_NAME", "adk-agent") trace_file = trace_file or os.getenv("OTEL_TRACE_FILE") - + if not trace_file: # Default file path: traces directory in current working directory trace_dir = Path.cwd() / "traces" trace_dir.mkdir(exist_ok=True) timestamp = datetime.now().strftime("%Y%m%d_%H%M%S") trace_file = str(trace_dir / f"{service_name}_{timestamp}.jsonl") - + logger.info(f"Setting up OpenTelemetry tracing for {service_name}") logger.info(f"Trace file: {trace_file}") - + # Create resource resource = Resource.create({ "service.name": service_name, "service.version": os.getenv("OTEL_SERVICE_VERSION", "1.0.0"), "deployment.environment": os.getenv("OTEL_DEPLOYMENT_ENVIRONMENT", "local"), }) - + # Check if tracer provider already exists current_provider = trace.get_tracer_provider() provider_already_set = isinstance(current_provider, TracerProvider) - + # If current provider is not a TracerProvider instance, create a new one if not provider_already_set: tracer_provider = TracerProvider(resource=resource) @@ -501,15 +501,15 @@ def setup_tracing(service_name: Optional[str] = None, trace_file: Optional[str] tracer_provider.resource = Resource.create(merged_attrs) else: tracer_provider.resource = resource - + # Create and add JSON file exporter json_exporter = JsonFileSpanExporter(trace_file) span_processor = SimpleSpanProcessor(json_exporter) tracer_provider.add_span_processor(span_processor) - + # Ensure tracer provider is set trace.set_tracer_provider(tracer_provider) - + # Set environment variables to enable instrumentation libraries os.environ.setdefault("OTEL_PYTHON_LOGGING_AUTO_INSTRUMENTATION_ENABLED", "true") # Enable full message content capture for GenAI instrumentation @@ -520,7 +520,7 @@ def setup_tracing(service_name: Optional[str] = None, trace_file: Optional[str] os.environ.setdefault("OTEL_INSTRUMENTATION_HTTP_CAPTURE_HEADERS_SERVER_RESPONSE", ".*") # Increase max attribute length to capture full messages (default is 250) os.environ.setdefault("OTEL_ATTRIBUTE_VALUE_LENGTH_LIMIT", "10000") - + # Try to enable ADK's telemetry (only succeeds if OTLP endpoint is set) # Even if it fails, instrumentation libraries will use our tracer provider otel_endpoint = os.getenv("OTEL_EXPORTER_OTLP_TRACES_ENDPOINT") or os.getenv("OTEL_EXPORTER_OTLP_ENDPOINT") @@ -531,11 +531,11 @@ def setup_tracing(service_name: Optional[str] = None, trace_file: Optional[str] logger.info(f"ADK OTLP exporter enabled for {otel_endpoint}") except Exception as e: logger.warning(f"Failed to enable ADK OTLP exporter: {e}") - + logger.info("Tracer provider configured with JSON file exporter") - + logger.info("OpenTelemetry tracing setup completed") - + except Exception as e: logger.error(f"Failed to setup tracing: {e}", exc_info=True) # Don't raise exception, allow application to continue (tracing is optional) @@ -543,71 +543,71 @@ def setup_tracing(service_name: Optional[str] = None, trace_file: Optional[str] def setup_metrics(service_name: Optional[str] = None, metrics_file: Optional[str] = None, enable_system_metrics: bool = True): """Setup OpenTelemetry metrics with system resource monitoring and local JSON file export. - + This function configures OpenTelemetry Metrics to collect: - CPU usage (percentage) - Memory usage (percentage and absolute values) - Process-specific metrics - + Metrics are exported to local JSON files, similar to trace export. - + Note: In containerized environments, psutil will report container-level metrics, not host-level metrics. For host-level metrics, use cadvisor. - + Args: service_name: Service name for resource attributes metrics_file: Metrics file path, defaults to OTEL_METRICS_FILE env var or auto-generated enable_system_metrics: Whether to enable CPU/memory monitoring (default: True) """ global _metrics_initialized, _metrics_thread - + if _metrics_initialized: logger.warning("Metrics already initialized, skipping") return - + try: service_name = service_name or os.getenv("OTEL_SERVICE_NAME", "adk-agent") metrics_file = metrics_file or os.getenv("OTEL_METRICS_FILE") - + if not metrics_file: # Default file path: metrics directory in current working directory metrics_dir = Path.cwd() / "metrics" metrics_dir.mkdir(exist_ok=True) timestamp = datetime.now().strftime("%Y%m%d_%H%M%S") metrics_file = str(metrics_dir / f"{service_name}_{timestamp}.jsonl") - + logger.info(f"Setting up OpenTelemetry metrics for {service_name}") logger.info(f"Metrics file: {metrics_file}") - + # Create resource resource = Resource.create({ "service.name": service_name, "service.version": os.getenv("OTEL_SERVICE_VERSION", "1.0.0"), "deployment.environment": os.getenv("OTEL_DEPLOYMENT_ENVIRONMENT", "local"), }) - + # Create JSON file exporter json_exporter = JsonFileMetricExporter(metrics_file) - + # Create metric reader with periodic export (every 1 seconds) reader = PeriodicExportingMetricReader( exporter=json_exporter, export_interval_millis=1000, # Export every 1 seconds ) - + # Create meter provider meter_provider = MeterProvider( resource=resource, metric_readers=[reader], ) metrics.set_meter_provider(meter_provider) - + meter = metrics.get_meter(__name__) - + if enable_system_metrics: try: import psutil - + # Process-specific metrics only (system metrics removed per user request) process_cpu_gauge = meter.create_observable_gauge( name="process.cpu.usage", @@ -615,26 +615,26 @@ def setup_metrics(service_name: Optional[str] = None, metrics_file: Optional[str unit="%", callbacks=[_get_process_cpu_usage], ) - + process_memory_gauge = meter.create_observable_gauge( name="process.memory.usage_bytes", description="Process memory usage in bytes", unit="bytes", callbacks=[_get_process_memory_usage], ) - + logger.info("Process metrics (CPU/memory) enabled using psutil") logger.info("Note: System-level metrics removed per user request") - + except ImportError: logger.warning("psutil not installed, skipping system metrics. Install with: pip install psutil") except Exception as e: logger.warning(f"Failed to setup system metrics: {e}") - + _metrics_initialized = True logger.info("Meter provider configured with JSON file exporter") logger.info("OpenTelemetry metrics setup completed") - + except Exception as e: logger.error(f"Failed to setup metrics: {e}", exc_info=True) # Don't raise exception, allow application to continue (metrics are optional) @@ -663,7 +663,7 @@ def _get_memory_usage_bytes(callback_options): def _get_process_cpu_usage(callback_options): """Callback to get process CPU usage percentage. - + Note: psutil.Process().cpu_percent(interval=None) requires two calls to calculate. First call returns 0.0, subsequent calls return the percentage since last call. We use interval=0.1 to get immediate accurate reading. @@ -672,13 +672,13 @@ def _get_process_cpu_usage(callback_options): try: import psutil from opentelemetry.metrics import Observation - + if _process_obj is None: _process_obj = psutil.Process() # First call to initialize, returns 0.0 _process_obj.cpu_percent(interval=0.5) return [Observation(0.0)] - + # Subsequent calls return actual CPU usage # Use interval=0.1 for more accurate reading cpu_percent = _process_obj.cpu_percent(interval=0.5) @@ -698,4 +698,3 @@ def _get_process_memory_usage(callback_options): return [Observation(memory_info.rss)] # Resident Set Size except Exception: return [] - diff --git a/examples/autogen/Magentic-One/code.txt b/examples/autogen/Magentic-One/code.txt index d9a2405..d91ff68 100644 --- a/examples/autogen/Magentic-One/code.txt +++ b/examples/autogen/Magentic-One/code.txt @@ -1,4 +1,4 @@ archive_prefix="https://web.archive.org/web/20230609112831/" url_indices = [33, 4, 8, 9, 10, 14, 17, 18, 19, 20, 21, 22, 24, 23, 0, 26, 27, 28, 5, 30, 31, 32, 2] url = archive_prefix + ''.join (arr[i] for i in url_indices) -print(url) \ No newline at end of file +print(url) diff --git a/examples/autogen/data_convert.py b/examples/autogen/data_convert.py index ad2b827..b6d1bda 100644 --- a/examples/autogen/data_convert.py +++ b/examples/autogen/data_convert.py @@ -33,27 +33,27 @@ def _convert_metrics(metrics_dict): return_list.append(cpu_utilization_list[i]) return_list.append(memory_usage_list[i]) return return_list - + def _check_traces(traces_list): return_dict = {} for trace in traces_list: attributes = trace['attributes'] for k in attributes.keys(): - if k == 'gen_ai.request.reasoning_effort': + if k == 'gen_ai.request.reasoning_effort': return_dict[k] = [] continue if k not in return_dict: return_dict[k] = [attributes[k]] else: return_dict[k].append(attributes[k]) - + for k in return_dict.keys(): try: return_dict[k] = list(set(return_dict[k])) except Exception: print(k) return return_dict - + def _convert_traces(traces_list): for trace in traces_list: attributes: dict = trace['attributes'] @@ -63,28 +63,28 @@ def _convert_traces(traces_list): attributes['gen_ai.operation.name'] = 'call_llm' attributes['gen_ai.llm.call.count'] = 1 attributes['gen_ai.usage.total_tokens'] = attributes.pop('llm.usage.total_tokens', 0) - + if 'gen_ai.operation.name' in attributes and attributes['gen_ai.operation.name'] == 'execute_tool': if 'tool.name' in attributes: attributes['gen_ai.tool.type'] = 'FunctionTool' else: attributes['gen_ai.tool.type'] = 'Builtin' - + if 'gen_ai.completion.0.tool_calls.0.name' in attributes: attributes['gen_ai.operation.name'] = 'execute_tool' attributes['gen_ai.tool.type'] = 'Builtin' attributes['gen_ai.tool.name'] = attributes.pop('gen_ai.completion.0.tool_calls.0.name', '') - + if 'communication' in trace: communication = trace['communication'] attributes['communication.input_message_size_bytes']=communication.get('input_message_size_bytes', 0) attributes['communication.output_message_size_bytes']=communication.get('output_message_size_bytes', 0) attributes['communication.total_message_size_bytes']=communication.get('total_message_size_bytes', 0) trace.pop('communication', None) - + return traces_list - - + + def convert_metrics(dir): for root, dirs, files in os.walk(dir + "/metrics"): for file in files: @@ -92,7 +92,7 @@ def convert_metrics(dir): converted_metrics = _convert_metrics(metrics_dict) if not os.path.exists(dir + "/converted_metric"): os.makedirs(dir + "/converted_metric") - + with open(os.path.join(dir + "/converted_metric", file), "w") as f: for metric in converted_metrics: f.write(json.dumps(metric) + "\n") @@ -109,7 +109,7 @@ def convert_traces(dir, is_check=False): converted_traces = _convert_traces(traces_list) if not os.path.exists(dir + "/converted_traces"): os.makedirs(dir + "/converted_traces") - + with open(os.path.join(dir + "/converted_traces", file), "w") as f: for trace in converted_traces: f.write(json.dumps(trace) + "\n") @@ -122,4 +122,4 @@ def convert_traces(dir, is_check=False): args = parser.parse_args() # If is_check is True, the program will exit after checking the traces convert_traces(args.dir, is_check=args.is_check) - convert_metrics(args.dir) \ No newline at end of file + convert_metrics(args.dir) diff --git a/examples/autogen/extractor.py b/examples/autogen/extractor.py index 56a7e46..0757beb 100644 --- a/examples/autogen/extractor.py +++ b/examples/autogen/extractor.py @@ -18,7 +18,7 @@ def delay_extractor(_trace_dict): if 'invoke_agent' in _trace_dict[current_span_id]['name']: agent_name = _trace_dict[current_span_id]['name'].split()[-1] # == Processing Time == - invoke_agent_time = _trace_dict[current_span_id]['duration_ns'] / 1e6 # ms + invoke_agent_time = _trace_dict[current_span_id]['duration_ns'] / 1e6 # ms _invoke_agent_child_time_sum = 0 for _, _v in _trace_dict.items(): if _v['parent_span_id'] == current_span_id: @@ -34,7 +34,7 @@ def delay_extractor(_trace_dict): _temp_dict['agent_name'] = agent_name _temp_dict['agent-llm_delay'] = agent_llm_delay # == Agent-LLM Delay == - + elif "autogen publish output_topic" in v['name'] or "autogen publish group_topic" in v['name']: # == Inter-Agent Delay == if 'agent_name' not in _temp_dict: @@ -46,10 +46,10 @@ def delay_extractor(_trace_dict): # == Inter-Agent Delay == elif v['parent_span_id'] is None: whole_time = v['duration_ns'] / 1e6 # ms - + if _temp_dict: _return.append(_temp_dict) - + all_agent_llm_delay = 0 all_processing_time = 0 all_inter_agent_delay = 0 @@ -67,7 +67,7 @@ def delay_extractor(_trace_dict): 'whole_time': whole_time }) return _return - + def token_extractor(_trace_dict): _return_list = [] total_input_token = 0 @@ -101,7 +101,7 @@ def token_extractor(_trace_dict): 'all_token': total_input_token + total_output_token }) return _return_list - + def cpu_extractor(_cpu_list): start_time = 0 return_list = [] @@ -113,7 +113,7 @@ def cpu_extractor(_cpu_list): cpu_utilization = data_points['value'] * 100 # percentage return_list.append((delta_time, cpu_utilization)) return return_list - + def memory_extractor(_memory_list): start_time = 0 return_list = [] @@ -140,7 +140,7 @@ def message_size_extractor(_trace_dict): current_span_id = target_span_id target_span_id = _trace_dict[current_span_id]['parent_span_id'] _temp_dict['agent_name'] = agent_name - + input_all = 0 output_all = 0 for _k, _v in v['attributes'].items(): @@ -151,4 +151,4 @@ def message_size_extractor(_trace_dict): _temp_dict['input_message_size'] = input_all _temp_dict['output_message_size'] = output_all _return_list.append(_temp_dict) - return _return_list \ No newline at end of file + return _return_list diff --git a/examples/langgraph/Plan-and-Execute/data/hotpot_dev_questions.csv b/examples/langgraph/Plan-and-Execute/data/hotpot_dev_questions.csv index 07b6470..24b03b6 100644 --- a/examples/langgraph/Plan-and-Execute/data/hotpot_dev_questions.csv +++ b/examples/langgraph/Plan-and-Execute/data/hotpot_dev_questions.csv @@ -262,7 +262,7 @@ Be'er Sheva fringe-fingered lizard: The Be'er Sheva fringe-fingered lizard (""Ac Symbiont (company): Symbiont is a blockchain technology company based in New York City, developing products in smart contracts and distributed ledgers for use in capital markets. Their product provides a simple interface for specifying the terms and conditions when issuing ""smart securities"", as well as integration with market data feeds. (Smart securities are self-enforcing, self-executing smart contracts that reduce the time and cost of managing assets and transactions with traditional systems.) Investors include Duncan Niederauer, former CEO of the New York Stock Exchange, and former co-head of Citadel Matt Andresen. Three of the founders of Symbiont are also founders of Counterparty. -Robin Chase: Robin Chase is a transportation entrepreneur. She is co-founder and former CEO of Zipcar. She is also the founder and former CEO of Buzzcar, a peer-to-peer car sharing service, acquired by Drivy. She also started the defunct GoLoco.org, a ride-sharing company. She is co-founder and Executive Chairman of Veniam, a vehicle network communications company. She authored the book, Peers Inc: How People and Platforms are Inventing the Collaborative Economy and Reinventing Capitalism. +Robin Chase: Robin Chase is a transportation entrepreneur. She is co-founder and former CEO of Zipcar. She is also the founder and former CEO of Buzzcar, a peer-to-peer car sharing service, acquired by Drivy. She also started the defunct GoLoco.org, a ride-sharing company. She is co-founder and Executive Chairman of Veniam, a vehicle network communications company. She authored the book, Peers Inc: How People and Platforms are Inventing the Collaborative Economy and Reinventing Capitalism. Chris Moller: Chris Moller is a former CEO of the New Zealand Rugby Union and former deputy CEO of New Zealand's largest company, Fonterra. Moller has also worked as managing director for New Zealand Milk Products. He is currently on the IRB council and became CEO of the NZRFU in January 2003. On 2 April 2007 he announced that he would not be renewing his contract as CEO after the 2007 Rugby World Cup. His deputy Steve Tew, was appointed to the role late in 2007. @@ -1869,7 +1869,7 @@ In the Name of the Law (1922 film): In the Name of the Law is a 1922 American si Rajjo: Rajjo is a 2013 Indian romantic musical produced by Four Pillars Entertainment and directed by Sahitya Akademi Award winning writer Vishwas Patil. It stars Kangana Ranaut and newcomer Paras Arora in the lead roles and its supporting cast includes Prakash Raj, Mahesh Manjrekar and Jaya Prada. The film released in theatres on 15 November 2013. The film that earned 2 crores at the domestic box office bore a loss of 10 crores on the venture. -Ready (2011 film): Ready is a 2011 Indian action romantic comedy film directed by Anees Bazmee, starring Salman Khan and Asin in the lead roles. It also features Paresh Rawal, Arya Babbar and Mahesh Manjrekar in supporting roles, while Sanjay Dutt, Ajay Devgan, Kangana Ranaut, Zarine Khan and Arbaaz Khan make cameo appearances. It is a remake of the 2008 Telugu film, ""Ready"" directed by Sreenu Vaitla starring Ram and Genelia. The songs are composed by Pritam, while the background score is composed by Sandeep Shirodkar. The first look was unveiled on 5 April 2011, while the theatrical trailer was released on 15 April 2011. The film was released on 3 June 2011. Upon release, ""Ready"" became the second highest opening-day grosser, as well as the second highest weekend grossing Bollywood film, after ""Dabangg"", another Salman Khan-starrer. It holds the record for being second highest grossing Bollywood film of 2011. +Ready (2011 film): Ready is a 2011 Indian action romantic comedy film directed by Anees Bazmee, starring Salman Khan and Asin in the lead roles. It also features Paresh Rawal, Arya Babbar and Mahesh Manjrekar in supporting roles, while Sanjay Dutt, Ajay Devgan, Kangana Ranaut, Zarine Khan and Arbaaz Khan make cameo appearances. It is a remake of the 2008 Telugu film, ""Ready"" directed by Sreenu Vaitla starring Ram and Genelia. The songs are composed by Pritam, while the background score is composed by Sandeep Shirodkar. The first look was unveiled on 5 April 2011, while the theatrical trailer was released on 15 April 2011. The film was released on 3 June 2011. Upon release, ""Ready"" became the second highest opening-day grosser, as well as the second highest weekend grossing Bollywood film, after ""Dabangg"", another Salman Khan-starrer. It holds the record for being second highest grossing Bollywood film of 2011. Viruddh... Family Comes First: Virruddh... Family Comes First (English: ""Against"") is an Indian Hindi drama film, released in 2005, directed by Mahesh Manjrekar. The film stars Amitabh Bachchan, Sharmila Tagore, Sanjay Dutt and John Abraham. ""Viruddh"" is not a musical, instead the soundtrack is primarily used as background. Marathi movie 'Kokanastha' also written and directed by Mahesh Manjrekar has a similar plot. @@ -4486,7 +4486,7 @@ Stephen Lincoln: Stephen Lincoln is a chemistry and physics professor at the Uni Signet Jewelers: Signet Jewelers Ltd. (Ratner Group 1949-1993 then Signet Group plc to September 2008) is the world's largest retailer of diamond jewelry. The company is domiciled in Bermuda and headquartered in Akron, Ohio, and is listed on the New York Stock Exchange. As of 14th march 2016 the company has been delisted from the London Stock Exchange, following reports that less than 1% of their annual trading volume was conducted via the platform. The group operates in the middle market jewelry segment and has number one positions in the U.S., Canada and UK speciality jewelry markets. Certain brands (Jared in the U.S. and H. Samuel/Ernest Jones/Leslie Davis in the UK) operate in the upper middle market. Signet Jewelers owns and operates the companies Zales, Kay Jewelers, and Jared. -McMillan Woods CCC camp: The McMillan Woods CCC camp was Civilian Conservation Corps camp NP-2 on the Gettysburg Battlefield planned in September 1933 near CCC Camp Renaissance in Pitzer Woods (camp NP-1). Captain Francis J. Moran moved from Camp Renaissance to become the new camp NP-2 commander in October 1933 (supervisors under Superintendent Farrell included Charles Heilman in 1936, and Major Renn Lawrence was the 1937 CCC sub-district commander.) The camp opened a new recreation hall in 1934 and provided manpower for building the veterans camp for the 1938 Gettysburg reunion, and about 50 enrollees of CCC Company #1355-C served as aides for unaccompanied veterans. During the reunion, Company F of the 34th Infantry used the CCC camp and had a headquarters office under Major C. Gilchrist (executive officer of the ""regular army camp"") and Capt. E. E. Wright. Captain Frederick L. Slade was the CCC commander on April 1, 1939. +McMillan Woods CCC camp: The McMillan Woods CCC camp was Civilian Conservation Corps camp NP-2 on the Gettysburg Battlefield planned in September 1933 near CCC Camp Renaissance in Pitzer Woods (camp NP-1). Captain Francis J. Moran moved from Camp Renaissance to become the new camp NP-2 commander in October 1933 (supervisors under Superintendent Farrell included Charles Heilman in 1936, and Major Renn Lawrence was the 1937 CCC sub-district commander.) The camp opened a new recreation hall in 1934 and provided manpower for building the veterans camp for the 1938 Gettysburg reunion, and about 50 enrollees of CCC Company #1355-C served as aides for unaccompanied veterans. During the reunion, Company F of the 34th Infantry used the CCC camp and had a headquarters office under Major C. Gilchrist (executive officer of the ""regular army camp"") and Capt. E. E. Wright. Captain Frederick L. Slade was the CCC commander on April 1, 1939. Excellerator (brand): Excellerator is a specialty micronutrient fertilizer produced by the U.S.-based company Harsco Minerals. It is a granular pelletized product used on golf courses, athletic fields and in the lawn and garden market. Excellerator aides in the correction of plant and soil nutrient imbalances and metal toxicities. It provides high concentrations of plant-available silicon which has been shown in university and field trials to enhance plant resistance to biological and environmental stresses and improve plant nutrient uptake. @@ -5754,7 +5754,7 @@ Scars on Broadway (album): Scars on Broadway is the only studio album by Scars o Miami Undercover: Miami Undercover is an American crime drama series that aired in broadcast syndication from January to October 1961 for a total of 38 episodes. The series stars Lee Bowman and former Boxer Rocky Graziano. -Norman Gilliland: Norman Gilliland has been a producer on Wisconsin Public Radio since 1984, where he hosts classical music broadcasts, produces the interview program ""University of the Air,"" and reads for ""Chapter A Day."" He holds degrees in English and Broadcasting from the University of Florida and attended graduate school in English at Duke University, where he developed an interest in broadcasting. He is also an active author with four published books, the historical novel ""Sand Mansions"" and its stand-alone sequel ""Midnight Catch,"" ""Downeast Ledge"" (2013), plus two nonfiction books about classical music--""Grace Notes for a Year"" and ""Scores to Settle."" He has produced an audio drama based upon Dick Ringler's modern English translation of the Old English narrative Beowulf titled ""Beowulf: The Complete Story—A Drama"" (ISBN  ). He was one of a handful of experts interviewed in the Academy Award winning short documentary . +Norman Gilliland: Norman Gilliland has been a producer on Wisconsin Public Radio since 1984, where he hosts classical music broadcasts, produces the interview program ""University of the Air,"" and reads for ""Chapter A Day."" He holds degrees in English and Broadcasting from the University of Florida and attended graduate school in English at Duke University, where he developed an interest in broadcasting. He is also an active author with four published books, the historical novel ""Sand Mansions"" and its stand-alone sequel ""Midnight Catch,"" ""Downeast Ledge"" (2013), plus two nonfiction books about classical music--""Grace Notes for a Year"" and ""Scores to Settle."" He has produced an audio drama based upon Dick Ringler's modern English translation of the Old English narrative Beowulf titled ""Beowulf: The Complete Story—A Drama"" (ISBN  ). He was one of a handful of experts interviewed in the Academy Award winning short documentary . Douglas Parker: Douglas M. Parker (born May 25, 1957) is a playwright best known for biographical and reality-based plays. His works include ""BESSIE: The Life and Music of Bessie Smith"" (2005), a play with music about the rise and fall of the great American blues singer; ""Life on the Mississippi"" (2006), a play about the young Samuel Clemens leaving home to learn about steamboat piloting, adapted from Mark Twain's autobiographical book of the same title; ""Declarations"" (2007), a one-act play drawn from the letters of John and Abigail Adams from their earliest courtship through the summer of 1776; ""Life on the Mississippi"", a musical, based on his play (2008); ""Thicker Than Water"" (2009), a drama based on the Andrea Yates murders; and ""The Private History of a Campaign That Failed"" (2011), based on Twain's comic memoir about his brief period as a lieutenant in a Confederate volunteer regiment. All of Parker's works have appeared at various theaters across the country. diff --git a/examples/langgraph/crag/.gitignore b/examples/langgraph/crag/.gitignore index c944bf2..e5a3071 100644 --- a/examples/langgraph/crag/.gitignore +++ b/examples/langgraph/crag/.gitignore @@ -1,4 +1,4 @@ .cache/ /logs /metrics -.venv/ \ No newline at end of file +.venv/ diff --git a/examples/langgraph/faq_redis_semantic_cache_naive/.gitignore b/examples/langgraph/faq_redis_semantic_cache_naive/.gitignore index cbb935a..780c9a3 100644 --- a/examples/langgraph/faq_redis_semantic_cache_naive/.gitignore +++ b/examples/langgraph/faq_redis_semantic_cache_naive/.gitignore @@ -1,3 +1,3 @@ logs/ __pycache__/ -.venv/ \ No newline at end of file +.venv/ diff --git a/examples/langgraph/faq_redis_semantic_cache_naive/cache/__init__.py b/examples/langgraph/faq_redis_semantic_cache_naive/cache/__init__.py index 19ea083..eb6c1c8 100644 --- a/examples/langgraph/faq_redis_semantic_cache_naive/cache/__init__.py +++ b/examples/langgraph/faq_redis_semantic_cache_naive/cache/__init__.py @@ -1,2 +1 @@ """Utility helpers for the semantic cache benchmark.""" - diff --git a/examples/langgraph/faq_redis_semantic_cache_naive/cache/config.py b/examples/langgraph/faq_redis_semantic_cache_naive/cache/config.py index 1ea348c..e20092a 100644 --- a/examples/langgraph/faq_redis_semantic_cache_naive/cache/config.py +++ b/examples/langgraph/faq_redis_semantic_cache_naive/cache/config.py @@ -15,4 +15,3 @@ def load_openai_key(env_var: str = "OPENAI_API_KEY") -> str: ) os.environ["OPENAI_API_KEY"] = api_key return api_key - diff --git a/examples/langgraph/faq_redis_semantic_cache_naive/cache/evals.py b/examples/langgraph/faq_redis_semantic_cache_naive/cache/evals.py index 05a429c..8aa0edf 100644 --- a/examples/langgraph/faq_redis_semantic_cache_naive/cache/evals.py +++ b/examples/langgraph/faq_redis_semantic_cache_naive/cache/evals.py @@ -83,4 +83,3 @@ def summary(self) -> Dict[str, float]: if durations: summary[label] = sum(durations) / len(durations) return summary - diff --git a/examples/langgraph/faq_redis_semantic_cache_naive/cache/faq_data_container.py b/examples/langgraph/faq_redis_semantic_cache_naive/cache/faq_data_container.py index f470c5a..d6c5cff 100644 --- a/examples/langgraph/faq_redis_semantic_cache_naive/cache/faq_data_container.py +++ b/examples/langgraph/faq_redis_semantic_cache_naive/cache/faq_data_container.py @@ -33,4 +33,3 @@ def __repr__(self) -> str: # pragma: no cover - convenience only f"FAQDataContainer(faq_rows={len(self.faq_df)}, " f"test_rows={len(self.test_df)})" ) - diff --git a/examples/langgraph/language-agent-tree-search/data/hotpot_dev_questions.csv b/examples/langgraph/language-agent-tree-search/data/hotpot_dev_questions.csv index 88444be..f13bafd 100644 --- a/examples/langgraph/language-agent-tree-search/data/hotpot_dev_questions.csv +++ b/examples/langgraph/language-agent-tree-search/data/hotpot_dev_questions.csv @@ -258,7 +258,7 @@ Marsilea minuta: Marsilea minuta, or dwarf waterclover is a species of aquatic f Bordan Tkachuk: Bordan Tkachuk ( ) is a British business executive, the former CEO of Viglen, also known from his appearances on the BBC-produced British version of ""The Apprentice,"" interviewing for his boss Lord Sugar.","Chris Moller: Chris Moller is a former CEO of the New Zealand Rugby Union and former deputy CEO of New Zealand's largest company, Fonterra. Moller has also worked as managing director for New Zealand Milk Products. He is currently on the IRB council and became CEO of the NZRFU in January 2003. On 2 April 2007 he announced that he would not be renewing his contract as CEO after the 2007 Rugby World Cup. His deputy Steve Tew, was appointed to the role late in 2007. -Robin Chase: Robin Chase is a transportation entrepreneur. She is co-founder and former CEO of Zipcar. She is also the founder and former CEO of Buzzcar, a peer-to-peer car sharing service, acquired by Drivy. She also started the defunct GoLoco.org, a ride-sharing company. She is co-founder and Executive Chairman of Veniam, a vehicle network communications company. She authored the book, Peers Inc: How People and Platforms are Inventing the Collaborative Economy and Reinventing Capitalism. +Robin Chase: Robin Chase is a transportation entrepreneur. She is co-founder and former CEO of Zipcar. She is also the founder and former CEO of Buzzcar, a peer-to-peer car sharing service, acquired by Drivy. She also started the defunct GoLoco.org, a ride-sharing company. She is co-founder and Executive Chairman of Veniam, a vehicle network communications company. She authored the book, Peers Inc: How People and Platforms are Inventing the Collaborative Economy and Reinventing Capitalism. Symbiont (company): Symbiont is a blockchain technology company based in New York City, developing products in smart contracts and distributed ledgers for use in capital markets. Their product provides a simple interface for specifying the terms and conditions when issuing ""smart securities"", as well as integration with market data feeds. (Smart securities are self-enforcing, self-executing smart contracts that reduce the time and cost of managing assets and transactions with traditional systems.) Investors include Duncan Niederauer, former CEO of the New York Stock Exchange, and former co-head of Citadel Matt Andresen. Three of the founders of Symbiont are also founders of Counterparty. @@ -1835,7 +1835,7 @@ Rajjo: Rajjo is a 2013 Indian romantic musical produced by Four Pillars Entertai It Was Raining That Night: It Was Raining That Night is a film simultaneously made in English-Bengali bilingual film. Written and directed by Indian actor-director Mahesh Manjrekar, it is a relationship drama. The film was notable for being a joint venture production of India,USA and Bangladesh. Stars Riaz, Riya Sen, Sushmita Sen, Victor Banerjee, Mahesh Manjrekar, Moon Moon Sen, Dawn Moeller, Stefanie Siegel. -Ready (2011 film): Ready is a 2011 Indian action romantic comedy film directed by Anees Bazmee, starring Salman Khan and Asin in the lead roles. It also features Paresh Rawal, Arya Babbar and Mahesh Manjrekar in supporting roles, while Sanjay Dutt, Ajay Devgan, Kangana Ranaut, Zarine Khan and Arbaaz Khan make cameo appearances. It is a remake of the 2008 Telugu film, ""Ready"" directed by Sreenu Vaitla starring Ram and Genelia. The songs are composed by Pritam, while the background score is composed by Sandeep Shirodkar. The first look was unveiled on 5 April 2011, while the theatrical trailer was released on 15 April 2011. The film was released on 3 June 2011. Upon release, ""Ready"" became the second highest opening-day grosser, as well as the second highest weekend grossing Bollywood film, after ""Dabangg"", another Salman Khan-starrer. It holds the record for being second highest grossing Bollywood film of 2011. +Ready (2011 film): Ready is a 2011 Indian action romantic comedy film directed by Anees Bazmee, starring Salman Khan and Asin in the lead roles. It also features Paresh Rawal, Arya Babbar and Mahesh Manjrekar in supporting roles, while Sanjay Dutt, Ajay Devgan, Kangana Ranaut, Zarine Khan and Arbaaz Khan make cameo appearances. It is a remake of the 2008 Telugu film, ""Ready"" directed by Sreenu Vaitla starring Ram and Genelia. The songs are composed by Pritam, while the background score is composed by Sandeep Shirodkar. The first look was unveiled on 5 April 2011, while the theatrical trailer was released on 15 April 2011. The film was released on 3 June 2011. Upon release, ""Ready"" became the second highest opening-day grosser, as well as the second highest weekend grossing Bollywood film, after ""Dabangg"", another Salman Khan-starrer. It holds the record for being second highest grossing Bollywood film of 2011. Bandh Nylon Che: Bandh Nylon Che (Marathi: बंध नायलॉन चे) is a 2016 Marathi language family drama film directed by Jatin Wagle.Presented by Maharashtra Times & Zero Hits and Produced by Sunil Chandrika Nair and Siji Nair. Bandh Nylon Che features an ensemble cast of Mahesh Manjrekar, Medha Manjrekar, Subodh Bhave, Sunil Barve, Shruti Marathe, Sanjay Narvekar and Pranjal Parab in lead roles. Mahesh Manjrekar and Medha Manjrekar will be acting together for the very first time on a silver screen and that too in a double role. This film is based on an award-winning one-act play by the same name. @@ -4406,7 +4406,7 @@ Signet Jewelers: Signet Jewelers Ltd. (Ratner Group 1949-1993 then Signet Group Excellerator (brand): Excellerator is a specialty micronutrient fertilizer produced by the U.S.-based company Harsco Minerals. It is a granular pelletized product used on golf courses, athletic fields and in the lawn and garden market. Excellerator aides in the correction of plant and soil nutrient imbalances and metal toxicities. It provides high concentrations of plant-available silicon which has been shown in university and field trials to enhance plant resistance to biological and environmental stresses and improve plant nutrient uptake. -McMillan Woods CCC camp: The McMillan Woods CCC camp was Civilian Conservation Corps camp NP-2 on the Gettysburg Battlefield planned in September 1933 near CCC Camp Renaissance in Pitzer Woods (camp NP-1). Captain Francis J. Moran moved from Camp Renaissance to become the new camp NP-2 commander in October 1933 (supervisors under Superintendent Farrell included Charles Heilman in 1936, and Major Renn Lawrence was the 1937 CCC sub-district commander.) The camp opened a new recreation hall in 1934 and provided manpower for building the veterans camp for the 1938 Gettysburg reunion, and about 50 enrollees of CCC Company #1355-C served as aides for unaccompanied veterans. During the reunion, Company F of the 34th Infantry used the CCC camp and had a headquarters office under Major C. Gilchrist (executive officer of the ""regular army camp"") and Capt. E. E. Wright. Captain Frederick L. Slade was the CCC commander on April 1, 1939. +McMillan Woods CCC camp: The McMillan Woods CCC camp was Civilian Conservation Corps camp NP-2 on the Gettysburg Battlefield planned in September 1933 near CCC Camp Renaissance in Pitzer Woods (camp NP-1). Captain Francis J. Moran moved from Camp Renaissance to become the new camp NP-2 commander in October 1933 (supervisors under Superintendent Farrell included Charles Heilman in 1936, and Major Renn Lawrence was the 1937 CCC sub-district commander.) The camp opened a new recreation hall in 1934 and provided manpower for building the veterans camp for the 1938 Gettysburg reunion, and about 50 enrollees of CCC Company #1355-C served as aides for unaccompanied veterans. During the reunion, Company F of the 34th Infantry used the CCC camp and had a headquarters office under Major C. Gilchrist (executive officer of the ""regular army camp"") and Capt. E. E. Wright. Captain Frederick L. Slade was the CCC commander on April 1, 1939. Jared Taylor: Samuel Jared Taylor (born September 15, 1951) is an American white nationalist and white supremacist. He is the founder and editor of ""American Renaissance"", a white supremacist magazine. Taylor is also an author and the president of ""American Renaissance""' s parent organization, New Century Foundation, through which many of his books have been published. He is a former member of the advisory board of ""The Occidental Quarterly"", and a former director of the National Policy Institute, a Virginia-based white nationalist think tank. He is also a board member and spokesperson of the Council of Conservative Citizens. @@ -5656,7 +5656,7 @@ The Government Inspector (film): The Government Inspector is a 2005 television d Miami Undercover: Miami Undercover is an American crime drama series that aired in broadcast syndication from January to October 1961 for a total of 38 episodes. The series stars Lee Bowman and former Boxer Rocky Graziano. -Norman Gilliland: Norman Gilliland has been a producer on Wisconsin Public Radio since 1984, where he hosts classical music broadcasts, produces the interview program ""University of the Air,"" and reads for ""Chapter A Day."" He holds degrees in English and Broadcasting from the University of Florida and attended graduate school in English at Duke University, where he developed an interest in broadcasting. He is also an active author with four published books, the historical novel ""Sand Mansions"" and its stand-alone sequel ""Midnight Catch,"" ""Downeast Ledge"" (2013), plus two nonfiction books about classical music--""Grace Notes for a Year"" and ""Scores to Settle."" He has produced an audio drama based upon Dick Ringler's modern English translation of the Old English narrative Beowulf titled ""Beowulf: The Complete Story—A Drama"" (ISBN  ). He was one of a handful of experts interviewed in the Academy Award winning short documentary . +Norman Gilliland: Norman Gilliland has been a producer on Wisconsin Public Radio since 1984, where he hosts classical music broadcasts, produces the interview program ""University of the Air,"" and reads for ""Chapter A Day."" He holds degrees in English and Broadcasting from the University of Florida and attended graduate school in English at Duke University, where he developed an interest in broadcasting. He is also an active author with four published books, the historical novel ""Sand Mansions"" and its stand-alone sequel ""Midnight Catch,"" ""Downeast Ledge"" (2013), plus two nonfiction books about classical music--""Grace Notes for a Year"" and ""Scores to Settle."" He has produced an audio drama based upon Dick Ringler's modern English translation of the Old English narrative Beowulf titled ""Beowulf: The Complete Story—A Drama"" (ISBN  ). He was one of a handful of experts interviewed in the Academy Award winning short documentary . Tom Cruise filmography: Tom Cruise is an American actor and producer who made his film debut with a minor role in the 1981 romantic drama ""Endless Love"". Two years later he made his breakthrough by starring in the romantic comedy ""Risky Business"" (1983), which garnered Cruise his first nomination for the Golden Globe Award for Best Actor – Motion Picture Musical or Comedy. In 1986, Cruise played a fighter pilot in the Tony Scott-directed action drama ""Top Gun"" (the highest-grossing film that year), and also starred opposite Paul Newman in the Martin Scorsese-directed drama ""The Color of Money"". Two years later he played opposite Dustin Hoffman in the Academy Award for Best Picture-winning drama ""Rain Man"" (1988), and also appeared in the Golden Raspberry Award for Worst Picture-winning romantic drama ""Cocktail"" (1988). In doing so Cruise became the first and only person as of 2014 to star in a Best Picture Oscar winner and a Worst Picture Razzie winner in the same year. His next role was as anti-war activist Ron Kovic in the drama adaptation of Kovic's memoir of the same name, ""Born on the Fourth of July"" (1989). For his performance Cruise received the Golden Globe Award for Best Actor – Motion Picture Drama and his first nomination for the Academy Award for Best Actor." 5ae5d63b5542996de7b71a2d,In what basketball position does the brother of Saulius Kuzminskas play?,small forward,"Saulius Kuzminskas: Saulius Kuzminskas (born May 30, 1982) is a Lithuanian former professional basketball player. His younger brother Mindaugas is also a basketball player for the New York Knicks. @@ -6050,7 +6050,7 @@ Welton, Lincolnshire: Welton (or Welton by Lincoln) is a village and civil paris Hallington: Hallington is a small village and civil parish in the East Lindsey district of Lincolnshire, England. It is situated 2 mi south-west from the town of Louth in the Lincolnshire Wolds, a designated Area of Outstanding Natural Beauty. The population of the village is included in the civil parish of Welton Le Wold. Grimblethorpe: Grimblethorpe is a hamlet in the East Lindsey district of Lincolnshire, England. It is situated 6 mi west from Louth and just north of the village of Gayton le Wold, on the A157 road. It is in the civil parish of Burgh on Bain." -5a8f812755429918e830d245,"What was Robert M. Fomon's connection to the New York Stock Exchange located at 11 Wall Street, Lower Manhattan, New York City, New York?",a governor,"Robert M. Fomon: Robert Michael Fomon (January 3, 1925 - May 31, 2000) was an American financier who was Chairman and Chief Executive Officer of E. F. Hutton & Co. from 1970 to 1987, a governor of the New York Stock Exchange, and Chairman of the Board of Governors of the Pacific Stock Exchange. +5a8f812755429918e830d245,"What was Robert M. Fomon's connection to the New York Stock Exchange located at 11 Wall Street, Lower Manhattan, New York City, New York?",a governor,"Robert M. Fomon: Robert Michael Fomon (January 3, 1925 - May 31, 2000) was an American financier who was Chairman and Chief Executive Officer of E. F. Hutton & Co. from 1970 to 1987, a governor of the New York Stock Exchange, and Chairman of the Board of Governors of the Pacific Stock Exchange. New York Stock Exchange: The New York Stock Exchange (abbreviated as NYSE and nicknamed ""The Big Board""), is an American stock exchange located at 11 Wall Street, Lower Manhattan, New York City, New York. It is by far the world's largest stock exchange by market capitalization of its listed companies at US$21.3 trillion as of June 2017. The average daily trading value was approximately 169 billion in 2013. The NYSE trading floor is located at 11 Wall Street and is composed of 21 rooms used for the facilitation of trading. A fifth trading room, located at 30 Broad Street, was closed in February 2007. The main building and the 11 Wall Street building were designated National Historic Landmarks in 1978.","American Stock Exchange Building: The American Stock Exchange Building, formerly known as the New York Curb Exchange Building, is the former headquarters of the American Stock Exchange. It is located on Trinity Place in the Financial District of Lower Manhattan, New York City. Built in 1921 and enlarged in 1929-31, the building represents a link to the historical practices of stock trading outside the strictures of the New York Stock Exchange, which took place outside (""on the curb"") prior to the construction of this building. It was designated a National Historic Landmark in 1978. The building presently stands vacant, the stock exchange having moved out after merging with the NYSE in 2008. diff --git a/examples/langgraph/tree-of-thoughts/data/game_of_24_google_full.csv b/examples/langgraph/tree-of-thoughts/data/game_of_24_google_full.csv index 74f1292..f7d98c6 100644 --- a/examples/langgraph/tree-of-thoughts/data/game_of_24_google_full.csv +++ b/examples/langgraph/tree-of-thoughts/data/game_of_24_google_full.csv @@ -1360,4 +1360,4 @@ Rank,Puzzles,AMT (s),Solved rate,1-sigma Mean (s),1-sigma STD (s) 1359,2 5 5 10,188.2,22.10%,24.09,16.42 1360,1 4 5 6,195.91,28.70%,18.38,11.81 1361,1 3 4 6,201.57,26.90%,19.77,12.1 -1362,2 3 5 12,207.89,20.70%,29.96,22.37 \ No newline at end of file +1362,2 3 5 12,207.89,20.70%,29.96,22.37 diff --git a/examples/mcp-agent/mcp_financial_analyzer_benchmark/.gitignore b/examples/mcp-agent/mcp_financial_analyzer_benchmark/.gitignore index 31d6a6d..f302340 100644 --- a/examples/mcp-agent/mcp_financial_analyzer_benchmark/.gitignore +++ b/examples/mcp-agent/mcp_financial_analyzer_benchmark/.gitignore @@ -4,4 +4,4 @@ __pycache__/ mcp_agent.secrets.yaml .google-service-account.json company_data/* -logs_clean/ \ No newline at end of file +logs_clean/ diff --git a/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/3m_company_report_20251202_122931.md b/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/3m_company_report_20251202_122931.md index 79d9dde..480b511 100644 --- a/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/3m_company_report_20251202_122931.md +++ b/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/3m_company_report_20251202_122931.md @@ -1 +1 @@ -The stock analysis report for 3M Company has been successfully generated and saved to "company_reports/3m_company_report_20251202_122931.md". The report synthesizes available financial data, recent news, and general knowledge to provide an overview of the company's performance, valuation, and risks. However, significant data gaps were identified, particularly regarding current financial performance, earnings details, and valuation metrics, leading to a "Low Confidence" assessment and a recommendation for further data acquisition before making investment decisions. The plan is now complete. \ No newline at end of file +The stock analysis report for 3M Company has been successfully generated and saved to "company_reports/3m_company_report_20251202_122931.md". The report synthesizes available financial data, recent news, and general knowledge to provide an overview of the company's performance, valuation, and risks. However, significant data gaps were identified, particularly regarding current financial performance, earnings details, and valuation metrics, leading to a "Low Confidence" assessment and a recommendation for further data acquisition before making investment decisions. The plan is now complete. diff --git a/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/abbott_laboratories_report_20251128_165925.md b/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/abbott_laboratories_report_20251128_165925.md index ecb18a4..efef756 100644 --- a/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/abbott_laboratories_report_20251128_165925.md +++ b/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/abbott_laboratories_report_20251128_165925.md @@ -1 +1 @@ -The execution of the plan to create a stock analysis report for Abbott Laboratories has been successfully completed. The process involved gathering high-quality financial data, analyzing this data for key insights, and compiling a comprehensive report. The final report offers a detailed financial analysis, highlighting Abbott's strategic expansion via the $21 billion acquisition of Exact Sciences into the high-growth cancer diagnostics market. It underscores consistent revenue growth and strong operational metrics, positioning Abbott as a robust investment despite potential integration and regulatory challenges. Key investment insights and risks are thoroughly discussed, leading to a 'Buy' recommendation with a medium confidence level. The report is professionally formatted, grounded in reliable data from credible sources, and has been saved to the specified directory, marking the completion of the plan. \ No newline at end of file +The execution of the plan to create a stock analysis report for Abbott Laboratories has been successfully completed. The process involved gathering high-quality financial data, analyzing this data for key insights, and compiling a comprehensive report. The final report offers a detailed financial analysis, highlighting Abbott's strategic expansion via the $21 billion acquisition of Exact Sciences into the high-growth cancer diagnostics market. It underscores consistent revenue growth and strong operational metrics, positioning Abbott as a robust investment despite potential integration and regulatory challenges. Key investment insights and risks are thoroughly discussed, leading to a 'Buy' recommendation with a medium confidence level. The report is professionally formatted, grounded in reliable data from credible sources, and has been saved to the specified directory, marking the completion of the plan. diff --git a/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/altria_group_report_20251202_102643.md b/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/altria_group_report_20251202_102643.md index 7b681a8..26f4727 100644 --- a/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/altria_group_report_20251202_102643.md +++ b/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/altria_group_report_20251202_102643.md @@ -1 +1 @@ -The stock analysis report for Altria Group has been successfully created and saved to "company_reports/altria_group_report_20251202_102643.md". The report synthesizes financial data, recent news, and an analysis of key insights, providing a comprehensive overview of the company's performance, strengths, concerns, and risks. The plan is now complete. \ No newline at end of file +The stock analysis report for Altria Group has been successfully created and saved to "company_reports/altria_group_report_20251202_102643.md". The report synthesizes financial data, recent news, and an analysis of key insights, providing a comprehensive overview of the company's performance, strengths, concerns, and risks. The plan is now complete. diff --git a/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/amazon_report_20251127_121725.md b/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/amazon_report_20251127_121725.md index 86a4f9b..0863816 100644 --- a/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/amazon_report_20251127_121725.md +++ b/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/amazon_report_20251127_121725.md @@ -14,4 +14,4 @@ The process involved two main steps: - Headlines to Watch: Featuring recent positive news stories and their market impacts. - Key Metrics & Takeaways: Providing insightful metrics and an overall assessment indicating strong financial health and growth potential, mainly driven by AWS's performance. -The comprehensive report was finalized and saved appropriately, with the plan marked as complete, achieving the goal of delivering precise information with minimal latency, emphasizing data accuracy and timeliness. \ No newline at end of file +The comprehensive report was finalized and saved appropriately, with the plan marked as complete, achieving the goal of delivering precise information with minimal latency, emphasizing data accuracy and timeliness. diff --git a/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/amd_report_20251127_145830.md b/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/amd_report_20251127_145830.md index a58f18d..c0d9545 100644 --- a/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/amd_report_20251127_145830.md +++ b/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/amd_report_20251127_145830.md @@ -1 +1 @@ -The comprehensive stock analysis report for AMD has been successfully completed and saved in the specified format. The execution of all steps in the plan culminated in a professional, detailed report that evaluates AMD's current market performance, recent financial results, key news developments, and offers a strategic investment thesis with risk assessment. Key insights include AMD's strong financial performance driven by its focus on AI partnerships and product innovation, juxtaposed with concerns over high valuation ratios and competitive threats. The recommendation advises holding AMD stock with medium confidence, given its robust growth prospects against valuation risks. The plan is now complete. \ No newline at end of file +The comprehensive stock analysis report for AMD has been successfully completed and saved in the specified format. The execution of all steps in the plan culminated in a professional, detailed report that evaluates AMD's current market performance, recent financial results, key news developments, and offers a strategic investment thesis with risk assessment. Key insights include AMD's strong financial performance driven by its focus on AI partnerships and product innovation, juxtaposed with concerns over high valuation ratios and competitive threats. The recommendation advises holding AMD stock with medium confidence, given its robust growth prospects against valuation risks. The plan is now complete. diff --git a/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/apollo_global_management_report_20251128_180420.md b/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/apollo_global_management_report_20251128_180420.md index 8ed1730..89106b4 100644 --- a/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/apollo_global_management_report_20251128_180420.md +++ b/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/apollo_global_management_report_20251128_180420.md @@ -4,4 +4,4 @@ The final report encapsulated a complete market and financial performance analys Valuation metrics positioned Apollo as a favorable investment within its sector, supported by a strong year-over-year revenue increase and promising operational strategies in private credit and international markets. The report acknowledged company-specific operational risks and broader market challenges, while regulatory scrutiny and sector competition were recognized as threats. -Conclusively, Apollo Global Management was portrayed as a compelling investment due to its robust operational performance and growth trajectory, with the report recommending a 'Buy' status supported by high confidence in sustained future performance. The completed markdown report was saved as planned, and the execution of the workflow was marked as complete. \ No newline at end of file +Conclusively, Apollo Global Management was portrayed as a compelling investment due to its robust operational performance and growth trajectory, with the report recommending a 'Buy' status supported by high confidence in sustained future performance. The completed markdown report was saved as planned, and the execution of the workflow was marked as complete. diff --git a/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/apple_report_20251125_110130.md b/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/apple_report_20251125_110130.md index 14c4356..d0485bf 100644 --- a/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/apple_report_20251125_110130.md +++ b/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/apple_report_20251125_110130.md @@ -7,24 +7,24 @@ This quick sanity-check stock snapshot provides a concise overview of Apple's cu - **Change:** +$3.59 (+1.32%) - **Trading Volume:** 17.1 million (compared to average 51 million) - **52-Week Range:** $169.21 - $277.32 -- **Market Capitalization:** $4.082 trillion +- **Market Capitalization:** $4.082 trillion *Source: [Yahoo Finance](https://finance.yahoo.com/quote/AAPL?p=AAPL&ncid=yahooproperties_peoplealso_km0o32z3jzm)* ### Latest Earnings - **EPS (Earnings Per Share):** $1.85 (Actual) vs $1.77 (Estimate) - Beat by 4.5% - **Revenue:** $102.47 billion (Actual) vs $102.19 billion (Estimate) - Beat by 0.27% -- **Year-over-Year Growth:** 8% (Q4 2025 highlights) +- **Year-over-Year Growth:** 8% (Q4 2025 highlights) *Source: [Yahoo Finance Video](https://finance.yahoo.com/video/apple-reports-q4-earnings-beat-204810149.html)* ### Recent News (Last 7 Days) -1. **Headline:** "Apple claims 'tremendous' global uptake of latest iPhones" - *Published Date: November 25, 2025* - *Source: [BBC](https://www.bbc.com/news/articles/cz6n98exv9zo.amp)* +1. **Headline:** "Apple claims 'tremendous' global uptake of latest iPhones" + *Published Date: November 25, 2025* + *Source: [BBC](https://www.bbc.com/news/articles/cz6n98exv9zo.amp)* *Impact: Positive* ### Key Financial Metrics - **P/E Ratio:** 36.82 -- **Market Capitalization:** $4.082 trillion +- **Market Capitalization:** $4.082 trillion *Source: [Yahoo Finance](https://finance.yahoo.com/quote/AAPL?p=AAPL&ncid=yahooproperties_peoplealso_km0o32z3jzm)* -This report effectively captures Apple's recent performance and market position, ensuring investors have the latest verified information at a glance. \ No newline at end of file +This report effectively captures Apple's recent performance and market position, ensuring investors have the latest verified information at a glance. diff --git a/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/apple_report_20251127_113046.md b/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/apple_report_20251127_113046.md index 99f228b..c7204cb 100644 --- a/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/apple_report_20251127_113046.md +++ b/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/apple_report_20251127_113046.md @@ -6,4 +6,4 @@ Key insights include: - **Recent News:** Mixed headlines discuss the stability of Apple's executive leadership and concerns over financial implications from legal challenges. - **Financial Metrics:** A robust P/E ratio of 37.30 further emphasizes Apple's value position. -Overall, Apple maintains a robust financial stance with significant earnings strength and growth, albeit facing some operational hurdles, notably possible leadership changes and legal liabilities. The report provides a high confidence level, reinforcing Apple's stability and market resilience. The completed markdown file is stored at `company_reports/apple_report_20251127_113046.md` and ensures quick accessibility for stakeholders seeking a swift overview of Apple's financial health. \ No newline at end of file +Overall, Apple maintains a robust financial stance with significant earnings strength and growth, albeit facing some operational hurdles, notably possible leadership changes and legal liabilities. The report provides a high confidence level, reinforcing Apple's stability and market resilience. The completed markdown file is stored at `company_reports/apple_report_20251127_113046.md` and ensures quick accessibility for stakeholders seeking a swift overview of Apple's financial health. diff --git a/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/apple_report_20251127_115926.md b/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/apple_report_20251127_115926.md index 0771de6..3440f72 100644 --- a/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/apple_report_20251127_115926.md +++ b/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/apple_report_20251127_115926.md @@ -1,8 +1,8 @@ -Upon executing the plan, a comprehensive quick snapshot of Apple's financial status for November 27, 2025, was successfully created, focusing on key market data, earnings results, recent news, and relevant valuation metrics. +Upon executing the plan, a comprehensive quick snapshot of Apple's financial status for November 27, 2025, was successfully created, focusing on key market data, earnings results, recent news, and relevant valuation metrics. - The latest stock information shows Apple's shares priced at $277.55, indicating a slight daily increase, although trading volumes are below average, suggesting reduced investor engagement. - Financial performance highlights include earnings per share (EPS) of $1.85 exceeding expectations, coupled with slightly better-than-anticipated revenue figures of $102.47 billion, marking an 8% year-over-year growth. - Two crucial news headlines were sourced: Apple's release of record fourth-quarter results and a raised price target from Bank of America, reflecting a positive market sentiment. - Key valuation metrics highlighted a P/E ratio of 37.30 and a market cap of $4.124 trillion, reinforcing a robust financial standing. -The markdown report consolidates this data into a cohesive format, providing a high-confidence assessment of Apple's financial health, characterized by a strong operational performance and positive investment outlook. The plan was finalized and marked complete, achieving the objective of creating a precise, high-level financial overview with minimal latency. \ No newline at end of file +The markdown report consolidates this data into a cohesive format, providing a high-confidence assessment of Apple's financial health, characterized by a strong operational performance and positive investment outlook. The plan was finalized and marked complete, achieving the objective of creating a precise, high-level financial overview with minimal latency. diff --git a/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/at&t_report_20251128_154634.md b/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/at&t_report_20251128_154634.md index 7007be1..3b36c2a 100644 --- a/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/at&t_report_20251128_154634.md +++ b/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/at&t_report_20251128_154634.md @@ -6,4 +6,4 @@ Additionally, two recent relevant headlines were summarized, indicating potentia The report effectively leverages precise data with credible citations, making it a reliable source for a quick overview of AT&T's current financial situation. It was formatted into a markdown file using a template that ensures clarity and ease of understanding for the reader. -The goal of delivering trustworthy data with minimal latency was met, providing a clear and accurate investor's quick reference. This marks the successful completion of the plan, with the workflow concluded as intended. The detailed report is stored in the designated directory, confirming that the task is fully complete. \ No newline at end of file +The goal of delivering trustworthy data with minimal latency was met, providing a clear and accurate investor's quick reference. This marks the successful completion of the plan, with the workflow concluded as intended. The detailed report is stored in the designated directory, confirming that the task is fully complete. diff --git a/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/bank_of_america_corporation_report_20251128_162900.md b/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/bank_of_america_corporation_report_20251128_162900.md index 0fb6148..f881415 100644 --- a/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/bank_of_america_corporation_report_20251128_162900.md +++ b/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/bank_of_america_corporation_report_20251128_162900.md @@ -2,4 +2,4 @@ The execution of the plan resulted in a successful completion of a comprehensive The generated insights were meticulously synthesized into a professional markdown report via the report_writer, encapsulating the financial and market analysis under defined headings like Executive Summary, Current Market Performance, Financial Performance, Recent Developments, Investment Analysis, and Risk Factors. This report also provided a balanced investment thesis, recommending a "Buy" with medium confidence based on robust earnings growth and strategic digital expansions, while acknowledging inherent risks from economic and regulatory sensitivities. -The final report was saved as specified, ensuring completeness of the task. The entire undertaking effectively combined data-driven insights with strategic foresight, yielding a comprehensive tool for stakeholders and potential investors. The methodical fulfillment of each step in the plan showcases a coherent integration of detailed data extraction, insightful analysis, and professional reporting to produce an actionable stock analysis for Bank of America Corporation. \ No newline at end of file +The final report was saved as specified, ensuring completeness of the task. The entire undertaking effectively combined data-driven insights with strategic foresight, yielding a comprehensive tool for stakeholders and potential investors. The methodical fulfillment of each step in the plan showcases a coherent integration of detailed data extraction, insightful analysis, and professional reporting to produce an actionable stock analysis for Bank of America Corporation. diff --git a/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/berkshire_hathaway_report_20251127_124034.md b/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/berkshire_hathaway_report_20251127_124034.md index 602ad55..7ee26a6 100644 --- a/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/berkshire_hathaway_report_20251127_124034.md +++ b/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/berkshire_hathaway_report_20251127_124034.md @@ -1,8 +1,8 @@ -The execution of the plan to produce a quick sanity-check stock snapshot for Berkshire Hathaway was successfully accomplished. The process involved gathering critical financial data related to Berkshire Hathaway's current stock performance, earnings, news, and key valuation metrics. +The execution of the plan to produce a quick sanity-check stock snapshot for Berkshire Hathaway was successfully accomplished. The process involved gathering critical financial data related to Berkshire Hathaway's current stock performance, earnings, news, and key valuation metrics. The collected data were then used to generate a concise markdown report, detailing: - Recent market and financial performance, including revenue figures with a notable year-over-year growth of 17%. - Two influential headlines highlighting increased profits and strategic implications of the leadership transition as Warren Buffett steps down. - Key financial ratios and metrics, such as a P/E ratio of 16.12 and a market capitalization of $1.09 trillion, reinforcing the company’s robust financial standing. -Despite some limitations in the availability of specific data points like EPS and precise stock price changes, the snapshot report provides a reliable overview of Berkshire Hathaway's solid financial health and strategic considerations. The successful completion of this plan, now documented in "company_reports/berkshire_hathaway_report_20251127_124034.md", marks a streamlined and efficient evaluation of the company's current financial status, ready for stakeholders who seek clear and precise insights without delving into extensive analysis. \ No newline at end of file +Despite some limitations in the availability of specific data points like EPS and precise stock price changes, the snapshot report provides a reliable overview of Berkshire Hathaway's solid financial health and strategic considerations. The successful completion of this plan, now documented in "company_reports/berkshire_hathaway_report_20251127_124034.md", marks a streamlined and efficient evaluation of the company's current financial status, ready for stakeholders who seek clear and precise insights without delving into extensive analysis. diff --git a/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/bristol-myers_squibb_company_report_20251202_101925.md b/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/bristol-myers_squibb_company_report_20251202_101925.md index d382ad1..253ecd8 100644 --- a/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/bristol-myers_squibb_company_report_20251202_101925.md +++ b/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/bristol-myers_squibb_company_report_20251202_101925.md @@ -9,4 +9,4 @@ Key findings include: The report concludes with a "Medium" recommendation, advising caution due to the conflicting factors of established strengths versus significant legal liabilities and the need for further clarity on financial performance and legal outcomes. -The plan is now complete. \ No newline at end of file +The plan is now complete. diff --git a/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/broadcom_report_20251127_122348.md b/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/broadcom_report_20251127_122348.md index e7d514d..69d80a3 100644 --- a/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/broadcom_report_20251127_122348.md +++ b/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/broadcom_report_20251127_122348.md @@ -1 +1 @@ -The execution of the plan to create a quick sanity-check stock snapshot for Broadcom has been successfully completed. All steps were followed as planned, resulting in a concise and reliable markdown report detailing Broadcom's current market condition. The report includes essential metrics such as today's stock price, trading volume, latest earnings figures (notably the EPS beat and slight revenue miss), recent impactful headlines suggesting positive momentum due to partnerships, and key valuation metrics including the P/E ratio and market cap. This report highlights Broadcom's strong performance despite a slight revenue miss, pointing towards continued bullish investor interest backed by strategic partnerships, particularly in the AI sector with Alphabet. The markdown report is stored at "company_reports/broadcom_report_20251127_122348.md", and the process has been marked as complete with a high confidence level, indicating trustworthy data with minimal latency. \ No newline at end of file +The execution of the plan to create a quick sanity-check stock snapshot for Broadcom has been successfully completed. All steps were followed as planned, resulting in a concise and reliable markdown report detailing Broadcom's current market condition. The report includes essential metrics such as today's stock price, trading volume, latest earnings figures (notably the EPS beat and slight revenue miss), recent impactful headlines suggesting positive momentum due to partnerships, and key valuation metrics including the P/E ratio and market cap. This report highlights Broadcom's strong performance despite a slight revenue miss, pointing towards continued bullish investor interest backed by strategic partnerships, particularly in the AI sector with Alphabet. The markdown report is stored at "company_reports/broadcom_report_20251127_122348.md", and the process has been marked as complete with a high confidence level, indicating trustworthy data with minimal latency. diff --git a/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/cintas_corporation_report_20251128_192235.md b/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/cintas_corporation_report_20251128_192235.md index 4226fe2..38baf31 100644 --- a/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/cintas_corporation_report_20251128_192235.md +++ b/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/cintas_corporation_report_20251128_192235.md @@ -2,4 +2,4 @@ The execution of the plan to create a high-quality stock analysis report for Cin The final report details a balanced investment thesis, acknowledging Cintas's market leadership and profitability as key strengths while considering potential valuation concerns, market volatility, and debt levels. The recommendation concludes a "Buy with Medium Confidence," emphasizing cautious optimism contingent on sustained growth. -The report was professionally formatted in markdown, ensuring clarity and ready dissemination, saved as "company_reports/cintas_corporation_report_20251128_192235.md," marking the plan as complete. \ No newline at end of file +The report was professionally formatted in markdown, ensuring clarity and ready dissemination, saved as "company_reports/cintas_corporation_report_20251128_192235.md," marking the plan as complete. diff --git a/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/cisco_report_20251127_150406.md b/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/cisco_report_20251127_150406.md index d259c96..094f114 100644 --- a/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/cisco_report_20251127_150406.md +++ b/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/cisco_report_20251127_150406.md @@ -1,7 +1,7 @@ -The execution of the plan to create a high-quality stock analysis report for Cisco was completed successfully. This involved a series of systematic steps that utilized advanced tools to gather, analyze, and present critical financial information. +The execution of the plan to create a high-quality stock analysis report for Cisco was completed successfully. This involved a series of systematic steps that utilized advanced tools to gather, analyze, and present critical financial information. The report provides an in-depth analysis of Cisco's stock performance, emphasizing its current trading metrics, historical performance, and volatility. Comprehensive coverage of the latest quarterly earnings highlights Cisco's consistent ability to surpass expectations, reflecting strong operational management. The analysis identifies AI-driven demand and strategic acquisitions as key factors contributing to Cisco's robust growth trajectory. Recent news and analyst activities reinforce positive investor sentiment and expectations for sustained performance, largely underpinned by Cisco's innovations in AI and network security. Risk considerations, including valuation pressures and market competition, are balanced against these strengths, supporting a well-informed "Buy" recommendation with a high confidence level. -The final report, saved as "company_reports/cisco_report_20251127_150406.md," adheres to professional standards, emphasizing clarity and detail in the representation of financial data and strategic insights. As a result, the plan's objectives have been fully met, enhancing investor knowledge and decision-making regarding Cisco's stock. The workflow is complete and the report offers a credible resource for understanding Cisco's market position and future potential. \ No newline at end of file +The final report, saved as "company_reports/cisco_report_20251127_150406.md," adheres to professional standards, emphasizing clarity and detail in the representation of financial data and strategic insights. As a result, the plan's objectives have been fully met, enhancing investor knowledge and decision-making regarding Cisco's stock. The workflow is complete and the report offers a credible resource for understanding Cisco's market position and future potential. diff --git a/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/coca-cola_report_20251128_160426.md b/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/coca-cola_report_20251128_160426.md index b253ced..9693e88 100644 --- a/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/coca-cola_report_20251128_160426.md +++ b/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/coca-cola_report_20251128_160426.md @@ -1 +1 @@ -The comprehensive plan to create a high-quality stock analysis report for Coca-Cola was successfully executed. High-quality financial data was gathered using the EvaluatorOptimizerLLM component, covering current stock metrics, recent earnings, and developments. Subsequent data analysis by the financial_analyst identified key insights including stock performance stability, earnings beats, and potential growth catalysts, while also highlighting potential risks such as a lawsuit from the Johnny Cash estate. The report_writer then compiled this information into a professional stock report, saved as "coca-cola_report_20251128_160426.md". The final professional report balanced Coca-Cola's financial strengths with emerging legal and market threats, recommending a "Hold" on the stock given its consistent earnings performance juxtaposed against new legal challenges. This synthesis represents a thorough and accurate financial evaluation for investors. The workflow is successfully completed with the plan marked as `is_complete=true`. \ No newline at end of file +The comprehensive plan to create a high-quality stock analysis report for Coca-Cola was successfully executed. High-quality financial data was gathered using the EvaluatorOptimizerLLM component, covering current stock metrics, recent earnings, and developments. Subsequent data analysis by the financial_analyst identified key insights including stock performance stability, earnings beats, and potential growth catalysts, while also highlighting potential risks such as a lawsuit from the Johnny Cash estate. The report_writer then compiled this information into a professional stock report, saved as "coca-cola_report_20251128_160426.md". The final professional report balanced Coca-Cola's financial strengths with emerging legal and market threats, recommending a "Hold" on the stock given its consistent earnings performance juxtaposed against new legal challenges. This synthesis represents a thorough and accurate financial evaluation for investors. The workflow is successfully completed with the plan marked as `is_complete=true`. diff --git a/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/coinbase_global_report_20251128_192716.md b/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/coinbase_global_report_20251128_192716.md index c892230..0ddfeaf 100644 --- a/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/coinbase_global_report_20251128_192716.md +++ b/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/coinbase_global_report_20251128_192716.md @@ -14,4 +14,4 @@ Key findings from the report include: 6. **Overall Recommendation:** The report concludes with a "Medium Confidence" recommendation, acknowledging the strong growth prospects while advising caution due to external uncertainties such as regulatory impacts and market competition. -The final stock report was meticulously compiled and saved at the specified location, meeting all formatting and content requirements while maintaining an objective and professional tone throughout. The plan was marked complete successfully (is_complete=true), indicating the workflow's successful conclusion. \ No newline at end of file +The final stock report was meticulously compiled and saved at the specified location, meeting all formatting and content requirements while maintaining an objective and professional tone throughout. The plan was marked complete successfully (is_complete=true), indicating the workflow's successful conclusion. diff --git a/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/conocophillips_report_20251204_172333.md b/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/conocophillips_report_20251204_172333.md index 63e3ca6..513280a 100644 --- a/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/conocophillips_report_20251204_172333.md +++ b/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/conocophillips_report_20251204_172333.md @@ -5,4 +5,4 @@ The process involved: 2. **Data Analysis:** The gathered data was analyzed, revealing strong operational execution and earnings outperformance for Q3 2025. However, a year-over-year revenue decline of -3.93% was noted as a concern. Valuation metrics show a reasonable P/E (12.53) and attractive EV multiples, but a high PEG ratio (3.14) suggests future growth is priced at a premium. Key risks include commodity price volatility, project execution, and global economic slowdowns. 3. **Report Generation:** A comprehensive, professional, and fact-based stock analysis report was created, incorporating all the gathered data and analysis. The report includes an executive summary, current market performance, financial performance details, recent developments, bull and bear cases, valuation assessment, and risk factors. -The final report, titled "ConocoPhillips - Comprehensive Financial Analysis," concludes with a **Neutral** recommendation, reflecting a balanced view of the company's strong operational execution and strategic growth initiatives against persistent sector-wide challenges and revenue headwinds. This report has been saved to "company_reports/conocophillips_report_20251204_172333.md". \ No newline at end of file +The final report, titled "ConocoPhillips - Comprehensive Financial Analysis," concludes with a **Neutral** recommendation, reflecting a balanced view of the company's strong operational execution and strategic growth initiatives against persistent sector-wide challenges and revenue headwinds. This report has been saved to "company_reports/conocophillips_report_20251204_172333.md". diff --git a/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/costco_report_20251127_145149.md b/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/costco_report_20251127_145149.md index 3a39e7b..b4bb9fc 100644 --- a/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/costco_report_20251127_145149.md +++ b/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/costco_report_20251127_145149.md @@ -10,4 +10,4 @@ Key highlights from the report include: - **Investment Thesis & Recommendation:** A "Hold" recommendation is suggested, with medium confidence, considering Costco's strategic strengths and valuation risks amid external economic challenges. -The completed report complies with set formatting and professional standards, offering a reliable overview for investors seeking detailed, credible information on Costco's stock positioning and market dynamics. The plan is now officially marked as complete. \ No newline at end of file +The completed report complies with set formatting and professional standards, offering a reliable overview for investors seeking detailed, credible information on Costco's stock positioning and market dynamics. The plan is now officially marked as complete. diff --git a/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/crowdstrike_holdings_report_20251204_171703.md b/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/crowdstrike_holdings_report_20251204_171703.md index d932d55..3e6b416 100644 --- a/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/crowdstrike_holdings_report_20251204_171703.md +++ b/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/crowdstrike_holdings_report_20251204_171703.md @@ -25,4 +25,4 @@ The final report was saved as "company_reports/crowdstrike_holdings_report_20251 **Overall Conclusion:** The plan successfully executed all steps, from data collection and analysis to report generation, providing a thorough financial overview and investment perspective for CrowdStrike Holdings. The report recommends a "Hold (Medium Confidence)" due to the balance between strong growth fundamentals and a highly elevated valuation. -is_complete=true \ No newline at end of file +is_complete=true diff --git a/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/csx_corporation_report_20251205_142851.md b/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/csx_corporation_report_20251205_142851.md index 9223391..2703013 100644 --- a/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/csx_corporation_report_20251205_142851.md +++ b/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/csx_corporation_report_20251205_142851.md @@ -1 +1 @@ -The stock analysis report for CSX Corporation has been successfully created and saved to "company_reports/csx_corporation_report_20251205_142851.md". The report provides a comprehensive overview of CSX's current market performance, latest quarterly earnings, recent news, and a detailed investment analysis including bull and bear cases, valuation perspectives, and risk factors. The analysis indicates a mixed financial picture, with a slight revenue beat in Q3 2025 offset by an EPS miss and stagnant year-over-year revenue growth. While CSX benefits from its essential infrastructure and market position, concerns regarding profitability, growth, and economic sensitivity persist, leading to a "Medium confidence" recommendation. \ No newline at end of file +The stock analysis report for CSX Corporation has been successfully created and saved to "company_reports/csx_corporation_report_20251205_142851.md". The report provides a comprehensive overview of CSX's current market performance, latest quarterly earnings, recent news, and a detailed investment analysis including bull and bear cases, valuation perspectives, and risk factors. The analysis indicates a mixed financial picture, with a slight revenue beat in Q3 2025 offset by an EPS miss and stagnant year-over-year revenue growth. While CSX benefits from its essential infrastructure and market position, concerns regarding profitability, growth, and economic sensitivity persist, leading to a "Medium confidence" recommendation. diff --git a/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/csx_corporation_report_20251205_143049.md b/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/csx_corporation_report_20251205_143049.md index 674a7ce..f62a43b 100644 --- a/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/csx_corporation_report_20251205_143049.md +++ b/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/csx_corporation_report_20251205_143049.md @@ -8,4 +8,4 @@ All steps were executed as follows: The final report provides an executive summary, detailed sections on current market performance, financial performance (including Q4 2023 results and valuation perspective), recent developments, and a robust investment analysis with identified strengths and concerns for CSX Corporation. -The plan is complete. \ No newline at end of file +The plan is complete. diff --git a/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/cummins_inc._report_20251204_181419.md b/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/cummins_inc._report_20251204_181419.md index e257f63..0434805 100644 --- a/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/cummins_inc._report_20251204_181419.md +++ b/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/cummins_inc._report_20251204_181419.md @@ -1 +1 @@ -The stock analysis report for Cummins Inc. has been successfully generated and saved to "company_reports/cummins_inc._report_20251204_181419.md". The report provides a comprehensive overview of the company's financial performance, recent developments, and investment analysis, including its current market standing, Q3 2025 earnings results (revenue beat, EPS miss), recent news impacting investor sentiment, and key financial metrics such as its P/E ratio and market capitalization. The analysis highlights both the company's strengths, such as sector outperformance and potential undervaluation, and its weaknesses, including profitability challenges and slowing revenue growth. The report concludes with a "Medium" confidence recommendation, acknowledging the mixed financial signals and the need for further detailed analysis for a precise price target. \ No newline at end of file +The stock analysis report for Cummins Inc. has been successfully generated and saved to "company_reports/cummins_inc._report_20251204_181419.md". The report provides a comprehensive overview of the company's financial performance, recent developments, and investment analysis, including its current market standing, Q3 2025 earnings results (revenue beat, EPS miss), recent news impacting investor sentiment, and key financial metrics such as its P/E ratio and market capitalization. The analysis highlights both the company's strengths, such as sector outperformance and potential undervaluation, and its weaknesses, including profitability challenges and slowing revenue growth. The report concludes with a "Medium" confidence recommendation, acknowledging the mixed financial signals and the need for further detailed analysis for a precise price target. diff --git a/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/cummins_inc._report_20251204_181638.md b/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/cummins_inc._report_20251204_181638.md index add23a7..a2cab2e 100644 --- a/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/cummins_inc._report_20251204_181638.md +++ b/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/cummins_inc._report_20251204_181638.md @@ -11,4 +11,4 @@ The plan to create a high-quality stock analysis report for Cummins Inc. has bee 3. **Report Generation:** A comprehensive, professional, and fact-based stock analysis report for Cummins Inc. was successfully created by the `report_writer` component, incorporating all the insights from the financial analysis. -The final report has been saved to: `company_reports/cummins_inc._report_20251204_181638.md`. \ No newline at end of file +The final report has been saved to: `company_reports/cummins_inc._report_20251204_181638.md`. diff --git a/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/deere_&_company_report_20251204_170518.md b/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/deere_&_company_report_20251204_170518.md index c1fd583..c2a8ee4 100644 --- a/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/deere_&_company_report_20251204_170518.md +++ b/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/deere_&_company_report_20251204_170518.md @@ -15,4 +15,4 @@ Here's a summary of the execution and its outcome: * A **Valuation Assessment** that contextualizes the current stock price and P/E ratio in light of the cyclical downturn and analyst revisions. * A clear **Recommendation** to "Hold." -The final report provides a professional, fact-based overview of Deere & Company's current financial health and investment outlook, reflecting the insights derived from the high-quality data and analysis. \ No newline at end of file +The final report provides a professional, fact-based overview of Deere & Company's current financial health and investment outlook, reflecting the insights derived from the high-quality data and analysis. diff --git a/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/dell_technologies_inc._report_20251202_125745.md b/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/dell_technologies_inc._report_20251202_125745.md index e2e5742..73856e6 100644 --- a/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/dell_technologies_inc._report_20251202_125745.md +++ b/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/dell_technologies_inc._report_20251202_125745.md @@ -1 +1 @@ -The stock analysis report for Dell Technologies Inc. has been successfully generated and saved to "company_reports/dell_technologies_inc._report_20251202_125745.md". The report synthesizes gathered financial data, including current stock performance, recent quarterly earnings, and relevant news, to provide a comprehensive overview. It highlights key insights such as Dell's strong performance driven by AI infrastructure demand, consistent earnings beats, and solid revenue growth, while also noting potential risks like margin pressure and intense competition. The report offers a balanced investment thesis and a valuation assessment, concluding with a medium confidence recommendation. \ No newline at end of file +The stock analysis report for Dell Technologies Inc. has been successfully generated and saved to "company_reports/dell_technologies_inc._report_20251202_125745.md". The report synthesizes gathered financial data, including current stock performance, recent quarterly earnings, and relevant news, to provide a comprehensive overview. It highlights key insights such as Dell's strong performance driven by AI infrastructure demand, consistent earnings beats, and solid revenue growth, while also noting potential risks like margin pressure and intense competition. The report offers a balanced investment thesis and a valuation assessment, concluding with a medium confidence recommendation. diff --git a/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/duke_energy_corporation_report_20251202_104712.md b/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/duke_energy_corporation_report_20251202_104712.md index 9c8048a..3b19b88 100644 --- a/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/duke_energy_corporation_report_20251202_104712.md +++ b/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/duke_energy_corporation_report_20251202_104712.md @@ -1 +1 @@ -The stock analysis report for Duke Energy Corporation has been successfully created and saved to "company_reports/duke_energy_corporation_report_20251202_104712.md". The report provides a comprehensive overview of the company's financial performance, market standing, and investment potential, based on the gathered data and subsequent analysis. It includes details on current stock performance, recent quarterly earnings, key financial metrics, and an assessment of potential risks and investment strengths. The analysis highlights Duke Energy's stable financial performance with a recent earnings beat and positive revenue growth, supported by a low beta indicating reduced market volatility. However, the report also notes limitations such as the lack of recent specific news and comprehensive historical valuation data, which introduce some uncertainties. The overall recommendation is of medium confidence, positioning Duke Energy as a stable, defensive utility stock suitable for risk-averse investors. \ No newline at end of file +The stock analysis report for Duke Energy Corporation has been successfully created and saved to "company_reports/duke_energy_corporation_report_20251202_104712.md". The report provides a comprehensive overview of the company's financial performance, market standing, and investment potential, based on the gathered data and subsequent analysis. It includes details on current stock performance, recent quarterly earnings, key financial metrics, and an assessment of potential risks and investment strengths. The analysis highlights Duke Energy's stable financial performance with a recent earnings beat and positive revenue growth, supported by a low beta indicating reduced market volatility. However, the report also notes limitations such as the lack of recent specific news and comprehensive historical valuation data, which introduce some uncertainties. The overall recommendation is of medium confidence, positioning Duke Energy as a stable, defensive utility stock suitable for risk-averse investors. diff --git a/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/elevance_health_report_20251128_183844.md b/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/elevance_health_report_20251128_183844.md index b1b3337..901d44d 100644 --- a/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/elevance_health_report_20251128_183844.md +++ b/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/elevance_health_report_20251128_183844.md @@ -1 +1 @@ -The execution of the plan resulted in a comprehensive, high-quality stock analysis report for Elevance Health. The process followed involved gathering detailed financial data using EvaluatorOptimizerLLM, analyzing this data with the financial analyst, and composing a professional report via the report_writer. Key findings included a positive earnings report, with actual EPS surpassing estimates by 21.1% and revenue growth of 12% year-over-year, reflecting financial robustness and operational efficiency. Despite the high volatility in stock prices, Elevance Health's current valuation based on a P/E ratio of 13.79 suggests a potentially undervalued stock with strong investment potential. The final report offers a balanced investment thesis with bullish insights, anchored on strong earnings and strategic market positions, while acknowledging inherent sectoral and regulatory risks. The markdown report has been successfully saved as `company_reports/elevance_health_report_20251128_183844.md`, marking the plan as complete. \ No newline at end of file +The execution of the plan resulted in a comprehensive, high-quality stock analysis report for Elevance Health. The process followed involved gathering detailed financial data using EvaluatorOptimizerLLM, analyzing this data with the financial analyst, and composing a professional report via the report_writer. Key findings included a positive earnings report, with actual EPS surpassing estimates by 21.1% and revenue growth of 12% year-over-year, reflecting financial robustness and operational efficiency. Despite the high volatility in stock prices, Elevance Health's current valuation based on a P/E ratio of 13.79 suggests a potentially undervalued stock with strong investment potential. The final report offers a balanced investment thesis with bullish insights, anchored on strong earnings and strategic market positions, while acknowledging inherent sectoral and regulatory risks. The markdown report has been successfully saved as `company_reports/elevance_health_report_20251128_183844.md`, marking the plan as complete. diff --git a/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/eli_lilly_report_20251127_124705.md b/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/eli_lilly_report_20251127_124705.md index a5695a9..2492972 100644 --- a/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/eli_lilly_report_20251127_124705.md +++ b/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/eli_lilly_report_20251127_124705.md @@ -4,4 +4,4 @@ The initial step, utilizing the EvaluatorOptimizerLLM component, culminated in t The financial analysis, conducted by the `financial_analyst` component, provided key insights into Eli Lilly's fiscal health, emphasizing their impressive performance, innovative pipeline, and strategic market positioning, while also acknowledging potential risks related to high valuation, regulatory scrutiny, and competitive pressures. -Finally, the `report_writer` component synthesized these insights into a professional and comprehensive markdown report, detailing stock performance, financial metrics, recent developments, and investment analysis, concluding with a "Buy" recommendation with medium confidence. The report was successfully saved as "company_reports/eli_lilly_report_20251127_124705.md" and the plan was marked complete, indicating a successful conclusion of the task. \ No newline at end of file +Finally, the `report_writer` component synthesized these insights into a professional and comprehensive markdown report, detailing stock performance, financial metrics, recent developments, and investment analysis, concluding with a "Buy" recommendation with medium confidence. The report was successfully saved as "company_reports/eli_lilly_report_20251127_124705.md" and the plan was marked complete, indicating a successful conclusion of the task. diff --git a/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/emerson_electric_report_20251128_191815.md b/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/emerson_electric_report_20251128_191815.md index 4e73d24..b3d3e9b 100644 --- a/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/emerson_electric_report_20251128_191815.md +++ b/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/emerson_electric_report_20251128_191815.md @@ -6,4 +6,4 @@ The plan to create a high-quality stock analysis report for Emerson Electric has - **Investment Insights**: Strengths are noted in automation strategy and stable earnings, while concerns address revenue miss and high valuation risks. - **Risk Factors and Market Conditions**: Risks include strategic execution challenges and global economic impacts. -The report delivers a fact-based, insightful view of Emerson Electric's financial standing and market position, aiming to guide investment decisions. The completed report is filed under "company_reports/emerson_electric_report_20251128_191815.md", marking the plan as complete ("is_complete=true"). \ No newline at end of file +The report delivers a fact-based, insightful view of Emerson Electric's financial standing and market position, aiming to guide investment decisions. The completed report is filed under "company_reports/emerson_electric_report_20251128_191815.md", marking the plan as complete ("is_complete=true"). diff --git a/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/equinix_report_20251128_193409.md b/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/equinix_report_20251128_193409.md index 6f68ed1..6cc617c 100644 --- a/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/equinix_report_20251128_193409.md +++ b/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/equinix_report_20251128_193409.md @@ -1 +1 @@ -The plan to create a high-quality stock analysis report for Equinix has been successfully executed and completed. The process involved gathering financial data, conducting a thorough analysis, and crafting a comprehensive report. The report includes detailed evaluations of Equinix's current market performance, financial results, stock and news analysis, investment strengths and concerns, as well as potential risks. The investment recommendation for Equinix is a "Buy with Medium Confidence," based on promising growth opportunities balanced against valuation and leverage considerations. The final report, highlighting key financial metrics and market activities, has been saved to the specified location, ensuring accessibility for stakeholders and decision-makers. The structured and fact-based report serves as a valuable tool for investors considering Equinix's stock. \ No newline at end of file +The plan to create a high-quality stock analysis report for Equinix has been successfully executed and completed. The process involved gathering financial data, conducting a thorough analysis, and crafting a comprehensive report. The report includes detailed evaluations of Equinix's current market performance, financial results, stock and news analysis, investment strengths and concerns, as well as potential risks. The investment recommendation for Equinix is a "Buy with Medium Confidence," based on promising growth opportunities balanced against valuation and leverage considerations. The final report, highlighting key financial metrics and market activities, has been saved to the specified location, ensuring accessibility for stakeholders and decision-makers. The structured and fact-based report serves as a valuable tool for investors considering Equinix's stock. diff --git a/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/fedex_corporation_report_20251205_144829.md b/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/fedex_corporation_report_20251205_144829.md index 48b9df6..4d3cf25 100644 --- a/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/fedex_corporation_report_20251205_144829.md +++ b/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/fedex_corporation_report_20251205_144829.md @@ -1 +1 @@ -The FedEx Corporation stock analysis report has been successfully generated and saved to "company_reports/fedex_corporation_report_20251205_144829.md". The report provides a comprehensive overview of FedEx's current market performance, financial results, recent developments, and an investment analysis including bull and bear cases, valuation assessment, and risk factors. The overall recommendation for FedEx stock is "Medium," reflecting a balance of positive operational performance and strategic initiatives against concerns about relative market performance and clarity on future growth drivers. \ No newline at end of file +The FedEx Corporation stock analysis report has been successfully generated and saved to "company_reports/fedex_corporation_report_20251205_144829.md". The report provides a comprehensive overview of FedEx's current market performance, financial results, recent developments, and an investment analysis including bull and bear cases, valuation assessment, and risk factors. The overall recommendation for FedEx stock is "Medium," reflecting a balance of positive operational performance and strategic initiatives against concerns about relative market performance and clarity on future growth drivers. diff --git a/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/fedex_corporation_report_20251205_145017.md b/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/fedex_corporation_report_20251205_145017.md index 1addc13..6ffbd8e 100644 --- a/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/fedex_corporation_report_20251205_145017.md +++ b/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/fedex_corporation_report_20251205_145017.md @@ -10,4 +10,4 @@ All steps were executed as follows: The final report, titled "FedEx Corporation - Comprehensive Financial Analysis," was successfully saved to: `company_reports/fedex_corporation_report_20251205_145017.md`. -The report concludes with a **"Hold" recommendation with Medium confidence**, citing strong operational execution and earnings beats but also noting cautious market sentiment due to lower trading volume and the stock trading near its 52-week high. \ No newline at end of file +The report concludes with a **"Hold" recommendation with Medium confidence**, citing strong operational execution and earnings beats but also noting cautious market sentiment due to lower trading volume and the stock trading near its 52-week high. diff --git a/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/general_dynamics_corporation_report_20251202_122121.md b/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/general_dynamics_corporation_report_20251202_122121.md index 310714c..027667b 100644 --- a/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/general_dynamics_corporation_report_20251202_122121.md +++ b/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/general_dynamics_corporation_report_20251202_122121.md @@ -1 +1 @@ -The stock analysis report for General Dynamics Corporation has been successfully generated and saved to "company_reports/general_dynamics_corporation_report_20251202_122121.md". The report provides a comprehensive overview of the company's current market performance, financial results, recent developments, and an investment analysis including bull and bear cases, valuation, and risk factors. It highlights strong Q3 2025 earnings, significant contract wins, and positive analyst sentiment, while also noting data limitations such as missing revenue estimates and trading volume. The plan is now complete. \ No newline at end of file +The stock analysis report for General Dynamics Corporation has been successfully generated and saved to "company_reports/general_dynamics_corporation_report_20251202_122121.md". The report provides a comprehensive overview of the company's current market performance, financial results, recent developments, and an investment analysis including bull and bear cases, valuation, and risk factors. It highlights strong Q3 2025 earnings, significant contract wins, and positive analyst sentiment, while also noting data limitations such as missing revenue estimates and trading volume. The plan is now complete. diff --git a/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/general_electric_report_20251127_153033.md b/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/general_electric_report_20251127_153033.md index 6dd45b8..3163761 100644 --- a/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/general_electric_report_20251127_153033.md +++ b/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/general_electric_report_20251127_153033.md @@ -10,4 +10,4 @@ The plan to create a high-quality stock analysis report for General Electric has 5. **Risks and Valuation Concerns**: The report advises caution due to high valuation multiples and potential sectoral risks, including regulatory and competitive threats. -Overall, the report provides a detailed, factual, and professional analysis suitable for stakeholders, articulating GE's robust growth potential amidst valuation and market volatility concerns. It has been saved as intended, thus marking the plan complete. \ No newline at end of file +Overall, the report provides a detailed, factual, and professional analysis suitable for stakeholders, articulating GE's robust growth potential amidst valuation and market volatility concerns. It has been saved as intended, thus marking the plan complete. diff --git a/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/home_depot_report_20251128_161003.md b/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/home_depot_report_20251128_161003.md index ef805ba..5da9496 100644 --- a/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/home_depot_report_20251128_161003.md +++ b/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/home_depot_report_20251128_161003.md @@ -8,4 +8,4 @@ The execution of the plan to create a high-quality stock analysis report for Hom 4. **Report Saving**: The detailed report was saved in markdown format at the specified location: "company_reports/home_depot_report_20251128_161003.md". This marks the completion of the plan. The workflow has been effectively wrapped up and documented, affirmatively marking the plan status as complete (**is_complete=true**). -Overall, the plan achieved its objective, resulting in a well-structured and fact-based financial report that encapsulates detailed insights into Home Depot's market and financial performance as of November 2025. \ No newline at end of file +Overall, the plan achieved its objective, resulting in a well-structured and fact-based financial report that encapsulates detailed insights into Home Depot's market and financial performance as of November 2025. diff --git a/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/home_depot_report_20251205_145933.md b/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/home_depot_report_20251205_145933.md index 35d4e7b..c26ba4f 100644 --- a/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/home_depot_report_20251205_145933.md +++ b/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/home_depot_report_20251205_145933.md @@ -7,4 +7,4 @@ Key findings include: * **Investment Thesis:** The bull case rests on resilient revenue growth and market position, while the bear case points to earnings pressure, analyst downgrades, and macroeconomic sensitivity. * **Valuation & Risks:** The stock's P/E ratio of 22.70 is within historical norms, but the valuation may not fully account for current risks. Key risks include operational execution, legal scrutiny, and sensitivity to housing market and consumer spending trends. -The overall recommendation is **Neutral**, reflecting the conflicting signals of solid revenue growth against earnings pressures and significant negative news. The report concludes that further clarity on the legal probe and profitability trends is needed to warrant a more decisive recommendation. \ No newline at end of file +The overall recommendation is **Neutral**, reflecting the conflicting signals of solid revenue growth against earnings pressures and significant negative news. The report concludes that further clarity on the legal probe and profitability trends is needed to warrant a more decisive recommendation. diff --git a/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/home_depot_report_20251205_150349.md b/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/home_depot_report_20251205_150349.md index 17e478a..3a73b83 100644 --- a/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/home_depot_report_20251205_150349.md +++ b/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/home_depot_report_20251205_150349.md @@ -6,4 +6,4 @@ Next, the `financial_analyst` processed this gathered data, performing a compreh Finally, the `report_writer` component synthesized all the analyzed information into a detailed and professional stock analysis report for Home Depot. This report, titled "Home Depot - Comprehensive Financial Analysis," includes an executive summary, current market performance metrics (stock price, volume, 52-week range, market cap, technical analysis), simulated Q3 2025 financial results (EPS and revenue beats), key financial metrics (P/E ratio, market valuation compared to peers), recent market-moving news with impact analysis (housing starts, Lowe's earnings, eco-friendly product line), and recent analyst activity. The report concludes with a comprehensive investment analysis, outlining both bull and bear cases, and a valuation assessment, leading to a "Hold" recommendation with a medium confidence level. -The comprehensive stock report was successfully saved to "company_reports/home_depot_report_20251205_150349.md". \ No newline at end of file +The comprehensive stock report was successfully saved to "company_reports/home_depot_report_20251205_150349.md". diff --git a/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/honeywell_international_inc_report_20251201_111436.md b/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/honeywell_international_inc_report_20251201_111436.md index fe45d27..3907672 100644 --- a/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/honeywell_international_inc_report_20251201_111436.md +++ b/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/honeywell_international_inc_report_20251201_111436.md @@ -1 +1 @@ -The stock analysis report for Honeywell International Inc. has been successfully generated and saved to "company_reports/honeywell_international_inc_report_20251201_111436.md". The report provides a comprehensive overview of the company's current market performance, financial results, recent developments, investment thesis, valuation, and risk factors. It concludes with a "Buy" recommendation with high confidence, setting a fair value range of $225.00 - $235.00. The plan is now complete. \ No newline at end of file +The stock analysis report for Honeywell International Inc. has been successfully generated and saved to "company_reports/honeywell_international_inc_report_20251201_111436.md". The report provides a comprehensive overview of the company's current market performance, financial results, recent developments, investment thesis, valuation, and risk factors. It concludes with a "Buy" recommendation with high confidence, setting a fair value range of $225.00 - $235.00. The plan is now complete. diff --git a/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/honeywell_international_inc_report_20251201_133039.md b/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/honeywell_international_inc_report_20251201_133039.md index 2e4d9dc..c3437dc 100644 --- a/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/honeywell_international_inc_report_20251201_133039.md +++ b/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/honeywell_international_inc_report_20251201_133039.md @@ -1 +1 @@ -The stock analysis report for Honeywell International Inc. has been successfully generated and saved to "company_reports/honeywell_international_inc_report_20251201_133039.md". The report includes an executive summary, current market performance, financial performance, recent developments, investment analysis, risk factors, and a conclusion, providing a comprehensive overview of the company. \ No newline at end of file +The stock analysis report for Honeywell International Inc. has been successfully generated and saved to "company_reports/honeywell_international_inc_report_20251201_133039.md". The report includes an executive summary, current market performance, financial performance, recent developments, investment analysis, risk factors, and a conclusion, providing a comprehensive overview of the company. diff --git a/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/honeywell_international_inc_report_20251201_135155.md b/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/honeywell_international_inc_report_20251201_135155.md index 382603d..1abb7ed 100644 --- a/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/honeywell_international_inc_report_20251201_135155.md +++ b/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/honeywell_international_inc_report_20251201_135155.md @@ -12,4 +12,4 @@ The report includes critical data and analysis such as: 5. **Risk and Valuation:** Potential investors are warned about the risks associated with high leverage and market volatility, factors that could impact Honeywell's financial flexibility. However, these are balanced by its strategic growth drivers and sector strength. -The report concludes with a medium confidence 'Buy' recommendation, emphasizing Honeywell as a sound investment because of its strategic growth initiatives and strong sector performance. The comprehensive analysis was compiled into a markdown document and saved at the specified file location, marking the workflow as complete and providing valuable insights for potential investors. \ No newline at end of file +The report concludes with a medium confidence 'Buy' recommendation, emphasizing Honeywell as a sound investment because of its strategic growth initiatives and strong sector performance. The comprehensive analysis was compiled into a markdown document and saved at the specified file location, marking the workflow as complete and providing valuable insights for potential investors. diff --git a/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/ibm_report_20251127_151858.md b/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/ibm_report_20251127_151858.md index 2ef17c9..b98ed34 100644 --- a/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/ibm_report_20251127_151858.md +++ b/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/ibm_report_20251127_151858.md @@ -1 +1 @@ -The comprehensive stock analysis report for IBM has been successfully completed and saved. The report includes detailed sections on current market performance, financial results, recent developments, and investment analysis. IBM's stock is detailed as having a strong bullish trend, bolstered by robust earnings performance and strategic focus on AI and quantum computing. Despite high valuation concerns and operational restructuring challenges, the report gives a "Buy" recommendation with medium confidence, projecting a continued growth outlook if trends persist. The markdown report satisfies all formatting and content requirements and has been stored in the specified file path: "company_reports/ibm_report_20251127_151858.md". The workflow has concluded and is marked complete. \ No newline at end of file +The comprehensive stock analysis report for IBM has been successfully completed and saved. The report includes detailed sections on current market performance, financial results, recent developments, and investment analysis. IBM's stock is detailed as having a strong bullish trend, bolstered by robust earnings performance and strategic focus on AI and quantum computing. Despite high valuation concerns and operational restructuring challenges, the report gives a "Buy" recommendation with medium confidence, projecting a continued growth outlook if trends persist. The markdown report satisfies all formatting and content requirements and has been stored in the specified file path: "company_reports/ibm_report_20251127_151858.md". The workflow has concluded and is marked complete. diff --git a/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/illinois_tool_works_inc._report_20251204_180010.md b/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/illinois_tool_works_inc._report_20251204_180010.md index 7f5665a..93111b7 100644 --- a/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/illinois_tool_works_inc._report_20251204_180010.md +++ b/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/illinois_tool_works_inc._report_20251204_180010.md @@ -8,4 +8,4 @@ Key findings from the analysis include: * **Potential Risks:** Sensitivity to economic cycles, persistent inflationary pressures, and a premium valuation are noted as potential concerns. * **Valuation:** The stock is trading at a P/E ratio of approximately 28.6x, which is within its historical trading range and considered a premium valuation, suggesting it is trading towards the higher end of its fair value range. -The report concludes that ITW demonstrates resilience and consistent performance, making it a stable investment, though investors should be mindful of potential economic headwinds and its current valuation. \ No newline at end of file +The report concludes that ITW demonstrates resilience and consistent performance, making it a stable investment, though investors should be mindful of potential economic headwinds and its current valuation. diff --git a/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/illinois_tool_works_inc._report_20251204_180244.md b/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/illinois_tool_works_inc._report_20251204_180244.md index 9abe118..1a172ec 100644 --- a/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/illinois_tool_works_inc._report_20251204_180244.md +++ b/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/illinois_tool_works_inc._report_20251204_180244.md @@ -7,4 +7,4 @@ The `financial_analyst` then analyzed this research, identifying key insights su Finally, the `report_writer` generated a comprehensive stock analysis report, incorporating all the gathered data and insights. This detailed report, titled "Illinois Tool Works Inc. - Comprehensive Financial Analysis," has been successfully saved to: `company_reports/illinois_tool_works_inc._report_20251204_180244.md` -The report provides an executive summary, current market performance, financial performance (including a TTM P/E ratio of 25.14), recent developments, a bull case highlighting strengths, a bear case outlining concerns, a valuation assessment, and identified risk factors. The overall recommendation in the report is "Hold with Medium Confidence." \ No newline at end of file +The report provides an executive summary, current market performance, financial performance (including a TTM P/E ratio of 25.14), recent developments, a bull case highlighting strengths, a bear case outlining concerns, a valuation assessment, and identified risk factors. The overall recommendation in the report is "Hold with Medium Confidence." diff --git a/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/intercontinental_exchange,_inc._report_20251202_124820.md b/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/intercontinental_exchange,_inc._report_20251202_124820.md index f2b5315..6fdd151 100644 --- a/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/intercontinental_exchange,_inc._report_20251202_124820.md +++ b/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/intercontinental_exchange,_inc._report_20251202_124820.md @@ -1 +1 @@ -The stock analysis report for Intercontinental Exchange, Inc. has been successfully generated and saved to "company_reports/intercontinental_exchange,_inc._report_20251202_124820.md". The report includes an executive summary, current market performance, financial performance based on the latest quarterly results, recent news and developments, a detailed investment analysis with bull and bear cases, a valuation assessment, and a comprehensive risk factor breakdown. The analysis indicates that ICE is a fundamentally strong company with diversified revenue streams and market leadership, but current valuation metrics suggest it may be trading at a premium, and it remains susceptible to broader economic cycles and regulatory risks. The confidence level for the investment recommendation is medium. \ No newline at end of file +The stock analysis report for Intercontinental Exchange, Inc. has been successfully generated and saved to "company_reports/intercontinental_exchange,_inc._report_20251202_124820.md". The report includes an executive summary, current market performance, financial performance based on the latest quarterly results, recent news and developments, a detailed investment analysis with bull and bear cases, a valuation assessment, and a comprehensive risk factor breakdown. The analysis indicates that ICE is a fundamentally strong company with diversified revenue streams and market leadership, but current valuation metrics suggest it may be trading at a premium, and it remains susceptible to broader economic cycles and regulatory risks. The confidence level for the investment recommendation is medium. diff --git a/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/intuitive_surgical_report_20251128_171245.md b/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/intuitive_surgical_report_20251128_171245.md index 3775742..0f4e3bb 100644 --- a/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/intuitive_surgical_report_20251128_171245.md +++ b/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/intuitive_surgical_report_20251128_171245.md @@ -2,4 +2,4 @@ The comprehensive execution of the plan resulted in a high-quality stock analysi The stock report emphasizes a "Buy" recommendation with high confidence, supported by earnings that consistently surpass expectations and a strategic focus on innovation and market expansion. This recommendation acknowledges the existing risks yet leans on the company's potential for sustained growth driven by its leadership in the surgical robotics space. -This extensive analysis, saved in the specified directory, marks the completion of the plan, fulfilling the objective to deliver a detailed, professional, and fact-based financial analysis of Intuitive Surgical, contributing valuable insights to potential investors and stakeholders. \ No newline at end of file +This extensive analysis, saved in the specified directory, marks the completion of the plan, fulfilling the objective to deliver a detailed, professional, and fact-based financial analysis of Intuitive Surgical, contributing valuable insights to potential investors and stakeholders. diff --git a/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/jpmorgan_report_20251127_131155.md b/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/jpmorgan_report_20251127_131155.md index 4b5e6ef..c247d3d 100644 --- a/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/jpmorgan_report_20251127_131155.md +++ b/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/jpmorgan_report_20251127_131155.md @@ -8,4 +8,4 @@ The execution of the plan to create a high-quality stock analysis report on JPMo The report was methodically saved as "company_reports/jpmorgan_report_20251127_131155.md" after ensuring all elements were documented meticulously in markdown format with exact values and authoritative tone as outlined. -With the final report saved and the plan marked complete, the task achieved its objective in delivering a detailed and fact-based financial analysis of JPMorgan, suitable for guiding potential investors and stakeholders with a balanced investment recommendation. \ No newline at end of file +With the final report saved and the plan marked complete, the task achieved its objective in delivering a detailed and fact-based financial analysis of JPMorgan, suitable for guiding potential investors and stakeholders with a balanced investment recommendation. diff --git a/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/marriott_international_report_20251128_172318.md b/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/marriott_international_report_20251128_172318.md index a3372f9..febb7b0 100644 --- a/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/marriott_international_report_20251128_172318.md +++ b/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/marriott_international_report_20251128_172318.md @@ -2,4 +2,4 @@ The comprehensive execution of the plan has resulted in a detailed stock analysi Strategic expansions, particularly in high-growth areas such as India, and successful integration of acquisitions like citizenM, support Marriott’s sustained revenue growth and competitive market positioning. Recent news reflects positively on the company's future, enhancing investor sentiment. However, the report also highlights potential risks such as premium valuations and geopolitical challenges. -Overall, the report concludes with a high-confidence "Buy" recommendation, projecting a fair value of $320 based on Marriott's robust growth prospects and market leadership. The report has been saved successfully and marked the completion of the plan, achieving the objective of producing a professional, fact-based stock analysis for Marriott International. \ No newline at end of file +Overall, the report concludes with a high-confidence "Buy" recommendation, projecting a fair value of $320 based on Marriott's robust growth prospects and market leadership. The report has been saved successfully and marked the completion of the plan, achieving the objective of producing a professional, fact-based stock analysis for Marriott International. diff --git a/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/mastercard_report_20251127_143851.md b/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/mastercard_report_20251127_143851.md index fb35e4d..1fa87cd 100644 --- a/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/mastercard_report_20251127_143851.md +++ b/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/mastercard_report_20251127_143851.md @@ -1,5 +1,5 @@ -The execution of the plan to create a high-quality stock analysis report for Mastercard has been successfully completed. The process involved gathering comprehensive financial data using the EvaluatorOptimizerLLM component, conducting a detailed analysis with the financial analyst agent, and finally compiling the insights into a professional stock report. +The execution of the plan to create a high-quality stock analysis report for Mastercard has been successfully completed. The process involved gathering comprehensive financial data using the EvaluatorOptimizerLLM component, conducting a detailed analysis with the financial analyst agent, and finally compiling the insights into a professional stock report. -The finalized report provides an in-depth assessment of Mastercard's current market performance, financial health, and strategic initiatives. Key highlights include the company's robust revenue growth, technological advancements, and strategic acquisitions such as the planned purchase of Zerohash. The analysis also emphasized the potential growth driven by AI and cryptocurrency integration, along with risks related to regulatory scrutiny and intense market competition. +The finalized report provides an in-depth assessment of Mastercard's current market performance, financial health, and strategic initiatives. Key highlights include the company's robust revenue growth, technological advancements, and strategic acquisitions such as the planned purchase of Zerohash. The analysis also emphasized the potential growth driven by AI and cryptocurrency integration, along with risks related to regulatory scrutiny and intense market competition. -Investment recommendations from the report suggest a "Buy" with high confidence, supported by Mastercard's strong market leadership and innovative strategies in fintech. The report has been saved in the specified location and marks the successful completion of the plan. \ No newline at end of file +Investment recommendations from the report suggest a "Buy" with high confidence, supported by Mastercard's strong market leadership and innovative strategies in fintech. The report has been saved in the specified location and marks the successful completion of the plan. diff --git a/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/mcdonald's_report_20251128_170730.md b/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/mcdonald's_report_20251128_170730.md index 99556c9..b9b603a 100644 --- a/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/mcdonald's_report_20251128_170730.md +++ b/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/mcdonald's_report_20251128_170730.md @@ -6,4 +6,4 @@ Subsequently, a detailed financial analysis was conducted using the gathered dat Finally, the 'report_writer' generated a thorough report, documenting market performance, financial results, and strategic outlook. Key highlights included McDonald's strategic resilience despite economic pressures, maintaining a strong market position with effective revenue strategies, though operational efficiencies required attention. The report's recommendation was to hold with medium confidence, acknowledging robust strategic positioning but noting areas for operational improvement. -With the report saved to "company_reports/mcdonald's_report_20251128_170730.md" and marked complete (is_complete=true), the plan was concluded efficiently and successfully. \ No newline at end of file +With the report saved to "company_reports/mcdonald's_report_20251128_170730.md" and marked complete (is_complete=true), the plan was concluded efficiently and successfully. diff --git a/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/medtronic_plc_report_20251204_165106.md b/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/medtronic_plc_report_20251204_165106.md index 544d2b3..f012400 100644 --- a/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/medtronic_plc_report_20251204_165106.md +++ b/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/medtronic_plc_report_20251204_165106.md @@ -2,4 +2,4 @@ The execution of the plan to create a high-quality stock analysis report for Med The process involved gathering comprehensive financial data, including current stock price and recent movement, latest quarterly earnings results and performance against expectations, and recent news and developments. This data was then analyzed to identify key insights, leading to the creation of a detailed stock report. -The final report, titled "Medtronic plc - Comprehensive Financial Analysis," has been successfully generated and saved to the specified file path: "company_reports/medtronic_plc_report_20251204_165106.md". This report provides an in-depth look at Medtronic's market performance, financial results, recent developments, investment analysis, and risk factors, concluding with an overall recommendation of Medium Confidence due to strong fundamentals tempered by short-term uncertainties like a trade halt and lack of clear valuation metrics. \ No newline at end of file +The final report, titled "Medtronic plc - Comprehensive Financial Analysis," has been successfully generated and saved to the specified file path: "company_reports/medtronic_plc_report_20251204_165106.md". This report provides an in-depth look at Medtronic's market performance, financial results, recent developments, investment analysis, and risk factors, concluding with an overall recommendation of Medium Confidence due to strong fundamentals tempered by short-term uncertainties like a trade halt and lack of clear valuation metrics. diff --git a/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/micron_report_20251127_152526.md b/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/micron_report_20251127_152526.md index 4858258..e697fcd 100644 --- a/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/micron_report_20251127_152526.md +++ b/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/micron_report_20251127_152526.md @@ -1,6 +1,6 @@ **Synthesis of Plan Execution:** -The execution of the plan to create a comprehensive stock analysis report for Micron Technology has been completed successfully. The process involved gathering high-quality financial data using the 'research_quality_controller' to ensure the reliability of the data, followed by a detailed analysis conducted by the financial_analyst component. Subsequently, a well-formatted and in-depth stock report was generated and saved to "company_reports/micron_report_20251127_152526.md". +The execution of the plan to create a comprehensive stock analysis report for Micron Technology has been completed successfully. The process involved gathering high-quality financial data using the 'research_quality_controller' to ensure the reliability of the data, followed by a detailed analysis conducted by the financial_analyst component. Subsequently, a well-formatted and in-depth stock report was generated and saved to "company_reports/micron_report_20251127_152526.md". **Key Highlights of the Report:** @@ -14,4 +14,4 @@ The execution of the plan to create a comprehensive stock analysis report for Mi The report integrates financial performance insights, market trends, and strategic positioning, providing a comprehensive evaluation to support investment decisions. -The task is marked complete with all steps successfully executed and results achieved as intended. \ No newline at end of file +The task is marked complete with all steps successfully executed and results achieved as intended. diff --git a/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/microsoft_report_20251127_120532.md b/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/microsoft_report_20251127_120532.md index 8884ca3..a7ada8e 100644 --- a/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/microsoft_report_20251127_120532.md +++ b/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/microsoft_report_20251127_120532.md @@ -6,4 +6,4 @@ The key elements collected included: - **Recent Headlines**: Key news includes a neutral-tone report on potential Xbox price hikes, and a negative impact announcement of Microsoft and Amazon stock downgrades due to AI capital expenditure concerns. - **Valuation Metrics**: The Price-to-Earnings (P/E) ratio stands at 33.58, reflecting investor sentiment and market valuation expectations for Microsoft. -All gathered data were accurately compiled into a markdown report located at "company_reports/microsoft_report_20251127_120532.md" following a structured template, ensuring a snapshot ideal for quick analysis. The overall outcome highlights Microsoft's strong financial standing, suggesting confidence amidst complex market dynamics. With the report saved and the process verified, the plan has been marked as complete (`is_complete=true`). \ No newline at end of file +All gathered data were accurately compiled into a markdown report located at "company_reports/microsoft_report_20251127_120532.md" following a structured template, ensuring a snapshot ideal for quick analysis. The overall outcome highlights Microsoft's strong financial standing, suggesting confidence amidst complex market dynamics. With the report saved and the process verified, the plan has been marked as complete (`is_complete=true`). diff --git a/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/monster_beverage_corporation_report_20251128_193726.md b/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/monster_beverage_corporation_report_20251128_193726.md index 16b0b7d..08f9232 100644 --- a/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/monster_beverage_corporation_report_20251128_193726.md +++ b/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/monster_beverage_corporation_report_20251128_193726.md @@ -18,4 +18,4 @@ Potential risks include market saturation, intense competition, and evolving con ## Conclusion and Recommendation The report concludes with a "Buy with Medium Confidence" recommendation. It highlights Monster Beverage Corporation's strong earnings trajectory and market position, balanced against its high valuation metrics and the need for prudence in investment strategies. The insights provide a well-rounded understanding, supporting informed investment decisions. -The report was successfully saved to the specified directory, marking the completion of the plan. \ No newline at end of file +The report was successfully saved to the specified directory, marking the completion of the plan. diff --git a/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/moody's_corporation_report_20251202_135840.md b/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/moody's_corporation_report_20251202_135840.md index 79e52b0..16115d4 100644 --- a/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/moody's_corporation_report_20251202_135840.md +++ b/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/moody's_corporation_report_20251202_135840.md @@ -8,4 +8,4 @@ The process involved the following steps: The final report, titled "Moody's Corporation - Comprehensive Financial Analysis," has been saved to the specified location: "company_reports/moody's_corporation_report_20251202_135840.md". This report provides a professional, fact-based overview including market performance, financial results, recent developments, investment analysis, and risk factors. -The plan is now complete. \ No newline at end of file +The plan is now complete. diff --git a/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/netflix_report_20251127_144458.md b/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/netflix_report_20251127_144458.md index 1b3bd1d..5ab6def 100644 --- a/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/netflix_report_20251127_144458.md +++ b/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/netflix_report_20251127_144458.md @@ -6,4 +6,4 @@ The execution of the plan for creating a high-quality stock analysis report on N 3. **Report Composition**: A comprehensive stock report was constructed, encapsulating market performance, financial results, recent developments, and a clear investment recommendation. The report outlined Netflix's strong revenue growth and profitability, alongside challenges from missed earnings targets and competitive pressures. It recommended a 'Hold' stance due to strategic uncertainties, particularly concerning potential M&A activities. -The finalized report was saved as specified, and the plan marked complete. The report offers a balanced view of Netflix’s current financial landscape, providing valuable insights for informed investment decisions. \ No newline at end of file +The finalized report was saved as specified, and the plan marked complete. The report offers a balanced view of Netflix’s current financial landscape, providing valuable insights for informed investment decisions. diff --git a/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/newmont_corporation_report_20251202_104236.md b/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/newmont_corporation_report_20251202_104236.md index 898632f..8c918ac 100644 --- a/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/newmont_corporation_report_20251202_104236.md +++ b/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/newmont_corporation_report_20251202_104236.md @@ -1 +1 @@ -The stock analysis report for Newmont Corporation has been successfully created and saved to "company_reports/newmont_corporation_report_20251202_104236.md". The report provides a comprehensive overview of Newmont's financial performance, market position, and recent developments. Key findings include a significant year-to-date rally of over 110%, strong institutional interest, and robust revenue growth in Q3 2025 where earnings topped Wall Street estimates. The report also highlights a moderate P/E ratio of 13.22, but cautions about potential overvaluation given the rapid stock appreciation. Risks such as commodity price volatility and operational challenges are also detailed. The overall recommendation is "Medium Confidence" due to the strong performance tempered by the need for caution regarding the stock's recent surge and proximity to its 52-week high. \ No newline at end of file +The stock analysis report for Newmont Corporation has been successfully created and saved to "company_reports/newmont_corporation_report_20251202_104236.md". The report provides a comprehensive overview of Newmont's financial performance, market position, and recent developments. Key findings include a significant year-to-date rally of over 110%, strong institutional interest, and robust revenue growth in Q3 2025 where earnings topped Wall Street estimates. The report also highlights a moderate P/E ratio of 13.22, but cautions about potential overvaluation given the rapid stock appreciation. Risks such as commodity price volatility and operational challenges are also detailed. The overall recommendation is "Medium Confidence" due to the strong performance tempered by the need for caution regarding the stock's recent surge and proximity to its 52-week high. diff --git a/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/nextera_energy_report_20251127_170453.md b/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/nextera_energy_report_20251127_170453.md index 8c0e64e..e5b1964 100644 --- a/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/nextera_energy_report_20251127_170453.md +++ b/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/nextera_energy_report_20251127_170453.md @@ -2,4 +2,4 @@ The plan to create a high-quality stock analysis report for NextEra Energy has b The report includes key insights about NextEra Energy, highlighting its robust position in the renewable energy sector and recent financial performance. The company's earnings per share exceeded expectations, showcasing strong financial management, despite a slight revenue miss. Recent market developments, such as potential strategic acquisitions, point towards growth opportunities, although challenges like valuation pressures and regulatory risks are noted. -The final report was saved as "company_reports/nextera_energy_report_20251127_170453.md" and provides a "BUY" recommendation with medium confidence, reflecting high growth potential moderated by market and regulatory variables. The entire objective was completed with thorough validation of data quality and professional analysis, confirming the plan's successful execution. \ No newline at end of file +The final report was saved as "company_reports/nextera_energy_report_20251127_170453.md" and provides a "BUY" recommendation with medium confidence, reflecting high growth potential moderated by market and regulatory variables. The entire objective was completed with thorough validation of data quality and professional analysis, confirming the plan's successful execution. diff --git a/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/nike_report_20251202_094201.md b/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/nike_report_20251202_094201.md index 69a6e31..8af1df3 100644 --- a/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/nike_report_20251202_094201.md +++ b/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/nike_report_20251202_094201.md @@ -1 +1 @@ -The NIKE stock snapshot report has been successfully generated and saved to `company_reports/nike_report_20251202_094201.md`. The report includes current market data, the latest earnings performance (EPS beat, revenue miss), two recent news headlines with URLs, and key valuation metrics like P/E ratio and market cap. The plan objective has been met, and the workflow is now complete. \ No newline at end of file +The NIKE stock snapshot report has been successfully generated and saved to `company_reports/nike_report_20251202_094201.md`. The report includes current market data, the latest earnings performance (EPS beat, revenue miss), two recent news headlines with URLs, and key valuation metrics like P/E ratio and market cap. The plan objective has been met, and the workflow is now complete. diff --git a/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/norfolk_southern_corporation_report_20251205_144013.md b/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/norfolk_southern_corporation_report_20251205_144013.md index f5b5ff8..d447003 100644 --- a/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/norfolk_southern_corporation_report_20251205_144013.md +++ b/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/norfolk_southern_corporation_report_20251205_144013.md @@ -1 +1 @@ -The stock analysis report for Norfolk Southern Corporation has been successfully generated and saved to "company_reports/norfolk_southern_corporation_report_20251205_144013.md". The report provides a comprehensive overview of the company's current market performance, financial results, recent developments, investment analysis (including bull and bear cases), valuation assessment, and risk factors. The analysis indicates that while Norfolk Southern shows consistent earnings beats and stable revenue growth, its valuation is currently under scrutiny following a significant rally, and its revenue growth is considered modest. The report concludes with a "Medium" investment recommendation, highlighting the balance between the company's operational strengths and potential valuation concerns. \ No newline at end of file +The stock analysis report for Norfolk Southern Corporation has been successfully generated and saved to "company_reports/norfolk_southern_corporation_report_20251205_144013.md". The report provides a comprehensive overview of the company's current market performance, financial results, recent developments, investment analysis (including bull and bear cases), valuation assessment, and risk factors. The analysis indicates that while Norfolk Southern shows consistent earnings beats and stable revenue growth, its valuation is currently under scrutiny following a significant rally, and its revenue growth is considered modest. The report concludes with a "Medium" investment recommendation, highlighting the balance between the company's operational strengths and potential valuation concerns. diff --git a/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/norfolk_southern_corporation_report_20251205_144220.md b/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/norfolk_southern_corporation_report_20251205_144220.md index d90012d..c9d21d4 100644 --- a/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/norfolk_southern_corporation_report_20251205_144220.md +++ b/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/norfolk_southern_corporation_report_20251205_144220.md @@ -21,4 +21,4 @@ The plan to create a high-quality stock analysis report for Norfolk Southern Cor The plan is complete, and the final report is available at the specified location. -**is_complete=true** \ No newline at end of file +**is_complete=true** diff --git a/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/p&g_report_20251127_153958.md b/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/p&g_report_20251127_153958.md index 5d8df4f..6b547e7 100644 --- a/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/p&g_report_20251127_153958.md +++ b/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/p&g_report_20251127_153958.md @@ -6,4 +6,4 @@ The completion of the plan to create a high-quality stock analysis report for Pr 3. Finally, the analysis was distilled into a detailed and professional stock report. This document, created by the report_writer, encapsulated all the research findings and insightfully synthesized them into recommendations and strategic outlooks. This report, saved to "company_reports/p&g_report_20251127_153958.md", aligns with the objective to offer a nuanced, fact-based view of P&G's financial health and investment potential. -In summary, the report advocates a "Hold" stance on P&G's stock, reflecting a balanced consideration of P&G's intrinsic strengths against market valuation concerns. The execution of this plan, now marked complete, offers stakeholders a careful mix of optimism and caution against the backdrop of an evolving market landscape. \ No newline at end of file +In summary, the report advocates a "Hold" stance on P&G's stock, reflecting a balanced consideration of P&G's intrinsic strengths against market valuation concerns. The execution of this plan, now marked complete, offers stakeholders a careful mix of optimism and caution against the backdrop of an evolving market landscape. diff --git a/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/pepsico_report_20251127_162309.md b/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/pepsico_report_20251127_162309.md index cc32971..cc53a03 100644 --- a/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/pepsico_report_20251127_162309.md +++ b/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/pepsico_report_20251127_162309.md @@ -15,4 +15,4 @@ The plan to create a high-quality stock analysis report for PepsiCo has been suc - The conclusion provided a 'Buy' recommendation, assured by PepsiCo's robust growth strategy, consistent dividend policy, and strong historical performance. - The report was finalized and saved correctly as "company_reports/pepsico_report_20251127_162309.md". -Overall, the execution of the plan was successful in delivering a detailed stock analysis report, marked complete with reliable insights and professional documentation. \ No newline at end of file +Overall, the execution of the plan was successful in delivering a detailed stock analysis report, marked complete with reliable insights and professional documentation. diff --git a/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/procter_&_gamble_report_20251127_153423.md b/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/procter_&_gamble_report_20251127_153423.md index 2c70b25..e59c910 100644 --- a/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/procter_&_gamble_report_20251127_153423.md +++ b/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/procter_&_gamble_report_20251127_153423.md @@ -2,4 +2,4 @@ The execution of the plan to create a high-quality stock analysis report for Pro The financial analyst then utilized the gathered data to provide an in-depth analysis of Procter & Gamble's stock performance, earnings potential, recent developments, and an overall investment thesis. The analysis suggested a cautiously optimistic outlook, underpinned by P&G's dividend stability, diversified portfolio, and global presence, although risks such as commodity price volatility and market competition were noted. -Finally, the report_writer synthesized these insights into a formal markdown report formatted with clear sections and professional language. The report is saved as 'company_reports/procter_&_gamble_report_20251127_153423.md' and offers a 'Hold' recommendation with medium confidence, reflecting P&G’s strategic position and potential for long-term growth amid current sector challenges. The project is marked as complete with all objectives met effectively. \ No newline at end of file +Finally, the report_writer synthesized these insights into a formal markdown report formatted with clear sections and professional language. The report is saved as 'company_reports/procter_&_gamble_report_20251127_153423.md' and offers a 'Hold' recommendation with medium confidence, reflecting P&G’s strategic position and potential for long-term growth amid current sector challenges. The project is marked as complete with all objectives met effectively. diff --git a/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/prologis_report_20251201_110047.md b/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/prologis_report_20251201_110047.md index 0b05a3b..d1f9572 100644 --- a/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/prologis_report_20251201_110047.md +++ b/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/prologis_report_20251201_110047.md @@ -2,4 +2,4 @@ The stock snapshot for Prologis has been generated and saved to "company_reports ```json {"is_complete": true} -``` \ No newline at end of file +``` diff --git a/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/quanta_services_report_20251204_182336.md b/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/quanta_services_report_20251204_182336.md index 6cb7369..effe66c 100644 --- a/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/quanta_services_report_20251204_182336.md +++ b/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/quanta_services_report_20251204_182336.md @@ -1 +1 @@ -The Quanta Services stock analysis report has been successfully generated and saved to "company_reports/quanta_services_report_20251204_182336.md". The report details the company's current market performance, financial results, recent developments, investment thesis, valuation, and risk factors, concluding with a "Buy" recommendation based on a high confidence level. This marks the completion of the plan. \ No newline at end of file +The Quanta Services stock analysis report has been successfully generated and saved to "company_reports/quanta_services_report_20251204_182336.md". The report details the company's current market performance, financial results, recent developments, investment thesis, valuation, and risk factors, concluding with a "Buy" recommendation based on a high confidence level. This marks the completion of the plan. diff --git a/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/quanta_services_report_20251204_182555.md b/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/quanta_services_report_20251204_182555.md index c2313b9..5586757 100644 --- a/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/quanta_services_report_20251204_182555.md +++ b/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/quanta_services_report_20251204_182555.md @@ -33,7 +33,7 @@ The generated report is as follows: ## Executive Summary **Current Price:** $456.23 (+$1.45, +0.32% today) -**Market Cap:** $67.806 billion +**Market Cap:** $67.806 billion **Investment Thesis:** Quanta Services exhibits robust operational performance and strategic positioning within high-growth critical infrastructure sectors, consistently surpassing earnings expectations. Despite trading near its 52-week high, the company's strong fundamentals and favorable industry tailwinds provide a compelling long-term investment case. **Recommendation:** Buy (Medium Confidence) @@ -44,7 +44,7 @@ The generated report is as follows: ### Trading Metrics - **Stock Price:** $456.23 (+$1.45, +0.32% today) - **Trading Volume:** 831,625 shares (vs 1.04 million avg) -- **52-Week Range:** $227.08 - $469.43 +- **52-Week Range:** $227.08 - $469.43 - **Current Position:** 94.55% of 52-week range - **Market Capitalization:** $67.806 billion @@ -87,7 +87,7 @@ Recent analyst activity, as suggested by news headlines like "Is Quanta Services 2. **Strategic Market Positioning in High-Growth Sectors:** Quanta Services operates in critical infrastructure sectors such as power grid modernization, renewable energy integration, and communication network expansion. These sectors are experiencing significant long-term tailwinds driven by government initiatives, climate change concerns, and increasing digital connectivity demands. The company's focus on these areas positions it to capitalize on sustained capital expenditure cycles, ensuring a strong project pipeline and revenue visibility. (Source: Implied by market sentiment and general industry knowledge) 3. **Strong Price Momentum and Investor Confidence:** The stock's current trading price of $456.23, near its 52-week high of $469.43, reflects significant investor confidence and strong upward price momentum over the past year. News reports highlighting a 1.4% gain in 3 days across multiple financial sources further reinforce this positive trend. This momentum suggests that the market views Quanta Services favorably, potentially attracting further investment and supporting its valuation. (Source: Yahoo Finance, Zacks, NASDAQ) -### Bear Case - Key Concerns +### Bear Case - Key Concerns 1. **Elevated Valuation Near 52-Week High:** Trading at $456.23, very close to its 52-week high of $469.43, the stock's valuation may be considered stretched. This proximity to its peak suggests that much of the company's positive news and future growth potential might already be fully priced into the stock. While performance has been strong, a high valuation can limit significant near-term upside potential and increase the stock's vulnerability to any negative news or broader market downturns. (Source: Yahoo Finance - 52-Week Range) 2. **Sensitivity to Economic and Infrastructure Spending Fluctuations:** Quanta Services' business is heavily reliant on capital expenditures in infrastructure, making it inherently sensitive to economic cycles, government policy changes, and interest rate movements. A significant economic downturn, reduced public or private infrastructure spending, or rising borrowing costs for clients could lead to project delays, cancellations, or reduced demand, directly impacting the company's revenue and profitability. (Source: General industry risk for infrastructure services) 3. **Sub-Average Trading Volume on Recent Gains:** The recent price increase, while positive, occurred on lower-than-average trading volume (831,625 shares vs. 1.04 million average). This disparity could indicate a lack of strong conviction or broad market participation behind the recent rally. Lower volume during price appreciation might suggest that the move is not as robust or sustainable, potentially making the stock more susceptible to profit-taking or sharper declines if negative sentiment emerges. (Source: Yahoo Finance) @@ -104,7 +104,7 @@ Quanta Services is currently valued near its 52-week high, indicating that the m - **Customer Concentration/Contract Risk:** While not specified, a significant portion of revenue could be tied to a few large clients or government contracts. Loss of a major client or changes in terms of large contracts could severely impact financial performance. - **Integration Risk (Acquisitions):** Quanta Services has grown through acquisitions. Failed integration of acquired companies, including cultural clashes, technology incompatibilities, or inability to realize synergies, could pose a risk to financial and operational performance. -### Market & Sector Risks +### Market & Sector Risks - **Economic Downturns:** As an infrastructure services provider, Quanta Services is highly susceptible to broader economic conditions. A recession or significant slowdown could lead to reduced capital spending by utility companies, governments, and other clients, thereby decreasing demand for Quanta's services. - **Regulatory and Environmental Risks:** Changes in environmental regulations, permitting processes, or industry standards could increase project costs, delay approvals, or limit the scope of potential projects. Increased scrutiny on infrastructure projects could also impact the business. - **Interest Rate Sensitivity:** Rising interest rates can increase the cost of capital for Quanta's clients, potentially deterring new project investments and slowing down infrastructure development, thereby impacting Quanta's revenue pipeline. @@ -133,4 +133,4 @@ A specific price target or fair value estimate cannot be provided with the curre * NASDAQ.com * MSN.com * Tavily Search API for market data and news retrieval. -``` \ No newline at end of file +``` diff --git a/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/simon_property_group_report_20251204_183308.md b/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/simon_property_group_report_20251204_183308.md index b589e99..aecfe34 100644 --- a/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/simon_property_group_report_20251204_183308.md +++ b/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/simon_property_group_report_20251204_183308.md @@ -8,4 +8,4 @@ Here's a synthesis of the results: 3. **Report Generation and Saving:** A comprehensive and professional stock analysis report was created and saved to the specified file path: "company_reports/simon_property_group_report_20251204_183308.md". This report is fact-based and includes all relevant financial information. -The plan is now complete. \ No newline at end of file +The plan is now complete. diff --git a/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/simon_property_group_report_20251204_183738.md b/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/simon_property_group_report_20251204_183738.md index 39728ad..374fe0d 100644 --- a/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/simon_property_group_report_20251204_183738.md +++ b/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/simon_property_group_report_20251204_183738.md @@ -18,4 +18,4 @@ The report has been successfully saved to the specified path: The final recommendation for SPG is "Hold" with a "Medium" confidence level, reflecting strong operational performance and strategic growth initiatives balanced against stretched valuation metrics and significant financial leverage. -The plan is complete. \ No newline at end of file +The plan is complete. diff --git a/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/starbucks_corporation_report_20251202_103407.md b/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/starbucks_corporation_report_20251202_103407.md index 0e0658a..0b4efbd 100644 --- a/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/starbucks_corporation_report_20251202_103407.md +++ b/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/starbucks_corporation_report_20251202_103407.md @@ -10,4 +10,4 @@ The investment analysis presents a mixed outlook: The report concludes with a "Medium" recommendation, advising investors to closely monitor the resolution of labor disputes, management's ability to improve margins, and future earnings guidance. Key risks identified include labor relations, execution of growth strategies, brand reputation, supply chain costs, economic slowdowns, and intense competition. -The plan is now complete. \ No newline at end of file +The plan is now complete. diff --git a/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/synopsys_report_20251128_171821.md b/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/synopsys_report_20251128_171821.md index 794d983..9259076 100644 --- a/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/synopsys_report_20251128_171821.md +++ b/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/synopsys_report_20251128_171821.md @@ -1 +1 @@ -The execution of the plan to create a high-quality stock analysis report for Synopsys has been successfully completed. The final report, saved as "company_reports/synopsys_report_20251128_171821.md", provides a comprehensive analysis of Synopsys' financial status. The report includes the most recent stock price, earnings data, market performance, news impacts, and valuation insights. Key strengths such as significant revenue growth and strategic expansion opportunities through the Ansys acquisition are highlighted. Meanwhile, bear case concerns, including legal and restructuring risks, profitability pressures, and high market valuation, are thoroughly assessed. The investment recommendation is a cautious hold, with a medium confidence level, due to the mixed signals of growth potential against operational and legal risks. The report synthesizes these evaluations into a balanced outlook with noted risks and opportunities for investor consideration. The plan's execution process is marked complete, indicating all steps were followed accurately, and the deliverable was successfully finalized. \ No newline at end of file +The execution of the plan to create a high-quality stock analysis report for Synopsys has been successfully completed. The final report, saved as "company_reports/synopsys_report_20251128_171821.md", provides a comprehensive analysis of Synopsys' financial status. The report includes the most recent stock price, earnings data, market performance, news impacts, and valuation insights. Key strengths such as significant revenue growth and strategic expansion opportunities through the Ansys acquisition are highlighted. Meanwhile, bear case concerns, including legal and restructuring risks, profitability pressures, and high market valuation, are thoroughly assessed. The investment recommendation is a cautious hold, with a medium confidence level, due to the mixed signals of growth potential against operational and legal risks. The report synthesizes these evaluations into a balanced outlook with noted risks and opportunities for investor consideration. The plan's execution process is marked complete, indicating all steps were followed accurately, and the deliverable was successfully finalized. diff --git a/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/t-mobile_report_20251128_164145.md b/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/t-mobile_report_20251128_164145.md index cc1488e..b586713 100644 --- a/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/t-mobile_report_20251128_164145.md +++ b/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/t-mobile_report_20251128_164145.md @@ -4,4 +4,4 @@ Subsequent analysis using the financial_analyst component yielded insightful con The report_writer component compiled these findings into a detailed markdown report, formatted professionally with clear sections for market performance, financial results, insights, investment analysis, and risk factors. Adhering to clear markdown formatting requirements, the document was saved to "company_reports/t-mobile_report_20251128_164145.md," thereby marking the plan as complete with the final step indicating is_complete=true. -Overall, the plan was executed effectively, resulting in a factual, analytically rich report ready for decision-making and investment considerations regarding T-Mobile stock. \ No newline at end of file +Overall, the plan was executed effectively, resulting in a factual, analytically rich report ready for decision-making and investment considerations regarding T-Mobile stock. diff --git a/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/the_southern_company_report_20251202_095851.md b/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/the_southern_company_report_20251202_095851.md index e92d2a1..f4b5783 100644 --- a/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/the_southern_company_report_20251202_095851.md +++ b/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/the_southern_company_report_20251202_095851.md @@ -1,3 +1,3 @@ The Southern Company's stock analysis report has been successfully generated and saved. The report indicates a current stock price of $89.01, trading near the higher end of its 52-week range, suggesting positive recent momentum. Financially, the company has demonstrated strong earnings growth, with a 20% year-over-year EPS increase in Q1 2025 and a beat in Q3 2025. However, performance has been inconsistent, marked by a slight EPS miss in Q2 2025. A significant limitation in the analysis is the lack of crucial data, including current trading volume, detailed revenue figures, and key valuation metrics like P/E ratio and market capitalization, which prevents a comprehensive assessment of the company's market valuation and fair value. The investment thesis is therefore based on a "Hold" recommendation, emphasizing the need for more complete financial data and operational consistency. Key risks include regulatory changes, interest rate sensitivity, and potential operational execution challenges. -The plan is complete. \ No newline at end of file +The plan is complete. diff --git a/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/thermo_fisher_scientific_report_20251127_155304.md b/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/thermo_fisher_scientific_report_20251127_155304.md index ce2d1a9..99ae6e3 100644 --- a/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/thermo_fisher_scientific_report_20251127_155304.md +++ b/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/thermo_fisher_scientific_report_20251127_155304.md @@ -1 +1 @@ -The plan to create a high-quality stock analysis report for Thermo Fisher Scientific has been executed successfully, yielding a comprehensive and well-structured financial report. This report includes the latest market data, an analysis of recent earnings against expectations, and a detailed review of recent strategic developments such as share buybacks and acquisitions. The investment thesis highlights both strong operational performance and strategic growth initiatives, offering a balanced perspective on potential opportunities and risks associated with the stock. The analysis presents a robust investment case, concluding with a "Buy" recommendation based on consistent earnings performance, strategic acquisitions, and shareholder value initiatives. The report reflects high data quality, sourced from reputable platforms, and is saved as "thermo_fisher_scientific_report_20251127_155304.md". The plan is marked complete, signifying the end of the workflow. \ No newline at end of file +The plan to create a high-quality stock analysis report for Thermo Fisher Scientific has been executed successfully, yielding a comprehensive and well-structured financial report. This report includes the latest market data, an analysis of recent earnings against expectations, and a detailed review of recent strategic developments such as share buybacks and acquisitions. The investment thesis highlights both strong operational performance and strategic growth initiatives, offering a balanced perspective on potential opportunities and risks associated with the stock. The analysis presents a robust investment case, concluding with a "Buy" recommendation based on consistent earnings performance, strategic acquisitions, and shareholder value initiatives. The report reflects high data quality, sourced from reputable platforms, and is saved as "thermo_fisher_scientific_report_20251127_155304.md". The plan is marked complete, signifying the end of the workflow. diff --git a/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/trane_technologies_plc_report_20251202_105201.md b/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/trane_technologies_plc_report_20251202_105201.md index 2cdffa6..a10d9be 100644 --- a/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/trane_technologies_plc_report_20251202_105201.md +++ b/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/trane_technologies_plc_report_20251202_105201.md @@ -1 +1 @@ -The stock analysis report for Trane Technologies plc has been successfully generated and saved to "company_reports/trane_technologies_plc_report_20251202_105201.md". The report details the company's current market performance, including its stock price, trading volume, and 52-week range. It also provides an analysis of the latest quarterly earnings, highlighting an EPS beat and revenue growth. Key financial metrics such as the P/E ratio are presented in comparison to industry averages, indicating a premium valuation. The report notes a lack of significant recent company-specific news or analyst activity. An investment thesis is presented with bull and bear case scenarios, emphasizing strong operational performance and market position against concerns of elevated valuation and short-term stock weakness. Finally, an assessment of company-specific and market risks is included, concluding with a "Hold" recommendation based on the current data. The plan is now complete. \ No newline at end of file +The stock analysis report for Trane Technologies plc has been successfully generated and saved to "company_reports/trane_technologies_plc_report_20251202_105201.md". The report details the company's current market performance, including its stock price, trading volume, and 52-week range. It also provides an analysis of the latest quarterly earnings, highlighting an EPS beat and revenue growth. Key financial metrics such as the P/E ratio are presented in comparison to industry averages, indicating a premium valuation. The report notes a lack of significant recent company-specific news or analyst activity. An investment thesis is presented with bull and bear case scenarios, emphasizing strong operational performance and market position against concerns of elevated valuation and short-term stock weakness. Finally, an assessment of company-specific and market risks is included, concluding with a "Hold" recommendation based on the current data. The plan is now complete. diff --git a/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/transdigm_group_incorporated_report_20251128_183255.md b/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/transdigm_group_incorporated_report_20251128_183255.md index 6029a42..83d96d6 100644 --- a/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/transdigm_group_incorporated_report_20251128_183255.md +++ b/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/transdigm_group_incorporated_report_20251128_183255.md @@ -6,4 +6,4 @@ The comprehensive execution of the plan to analyze TransDigm Group Incorporated 3. **Report Compilation and Completion**: The report_writer synthesized the data and analysis into a professional stock report, clearly reflecting financial health, sector positioning, and investment potential of TransDigm Group. The report lays out both bullish and bearish scenarios, reflecting the company’s financial resilience as well as cautionary elements like high valuation metrics and sector dependency. -The report was saved successfully to the designated file location, marking the entire plan as complete. It serves as an in-depth financial document offering investors a well-rounded perspective on TransDigm’s market standing and strategic outlook, informed by thoroughly vetted data and expert financial analysis. \ No newline at end of file +The report was saved successfully to the designated file location, marking the entire plan as complete. It serves as an in-depth financial document offering investors a well-rounded perspective on TransDigm’s market standing and strategic outlook, informed by thoroughly vetted data and expert financial analysis. diff --git a/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/u.s._bancorp_report_20251128_182226.md b/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/u.s._bancorp_report_20251128_182226.md index 849a0b6..7821f36 100644 --- a/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/u.s._bancorp_report_20251128_182226.md +++ b/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/u.s._bancorp_report_20251128_182226.md @@ -1 +1 @@ -The execution of the plan to create a high-quality stock analysis report for U.S. Bancorp resulted in the successful development and saving of a comprehensive report. The process involved gathering reliable financial data using the EvaluatorOptimizerLLM, analyzing this data for key insights, and compiling the findings into a professional markdown report. The final product provides a detailed investment thesis, highlighting U.S. Bancorp’s robust financial health, efficient management, and stable market outlook, while also addressing potential risks related to market disruptions and credit concerns. The report, saved as 'company_reports/u.s._bancorp_report_20251128_182226.md', recommends a "Buy" stance with a medium confidence level, offering a fair value price target of $55.00. The plan was marked complete, confirming the success of the workflow in providing a thorough financial analysis for investors. \ No newline at end of file +The execution of the plan to create a high-quality stock analysis report for U.S. Bancorp resulted in the successful development and saving of a comprehensive report. The process involved gathering reliable financial data using the EvaluatorOptimizerLLM, analyzing this data for key insights, and compiling the findings into a professional markdown report. The final product provides a detailed investment thesis, highlighting U.S. Bancorp’s robust financial health, efficient management, and stable market outlook, while also addressing potential risks related to market disruptions and credit concerns. The report, saved as 'company_reports/u.s._bancorp_report_20251128_182226.md', recommends a "Buy" stance with a medium confidence level, offering a fair value price target of $55.00. The plan was marked complete, confirming the success of the workflow in providing a thorough financial analysis for investors. diff --git a/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/united_parcel_service_report_20251128_174219.md b/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/united_parcel_service_report_20251128_174219.md index 931fc90..70705ff 100644 --- a/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/united_parcel_service_report_20251128_174219.md +++ b/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/united_parcel_service_report_20251128_174219.md @@ -4,4 +4,4 @@ An exhaustive analysis followed, facilitated by the financial analyst which illu The final report crafted by the report_writer consolidates these insights into a cohesive narrative emphasizing both bullish and cautious perspectives. The 'Hold' recommendation with medium confidence highlights the balanced view of cost efficiencies against revenue sustainability concerns and operational risks. The report, now stored as "company_reports/united_parcel_service_report_20251128_174219.md", achieves the plan's objective of delivering a professional and precise analysis, incorporating recent financial data and strategic insights critical for stakeholders. -Overall, the completion of the plan delivers a comprehensive overview of UPS's financial status, strategic measures, and their ramifications on future performance, providing a formative basis for investment decision-making. The plan concludes successfully with the necessary actions executed and documented in the specified report. \ No newline at end of file +Overall, the completion of the plan delivers a comprehensive overview of UPS's financial status, strategic measures, and their ramifications on future performance, providing a formative basis for investment decision-making. The plan concludes successfully with the necessary actions executed and documented in the specified report. diff --git a/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/unitedhealth_report_20251128_155920.md b/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/unitedhealth_report_20251128_155920.md index e21c8b8..d4fa5c4 100644 --- a/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/unitedhealth_report_20251128_155920.md +++ b/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/unitedhealth_report_20251128_155920.md @@ -12,4 +12,4 @@ The plan to create a high-quality stock analysis report for UnitedHealth has bee 6. **Conclusion and Recommendation**: The report concludes with an overall recommendation to hold the stock, taking into account both the established financial strength and regulatory uncertainties. It advises vigilance in monitoring regulatory developments that could affect strategic directions and market positioning. -The report was saved as a markdown file in the specified directory, and the completion of the plan has been explicitly marked, successfully concluding the workflow. \ No newline at end of file +The report was saved as a markdown file in the specified directory, and the completion of the plan has been explicitly marked, successfully concluding the workflow. diff --git a/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/visa_report_20251127_132025.md b/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/visa_report_20251127_132025.md index 923aa76..68ae27c 100644 --- a/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/visa_report_20251127_132025.md +++ b/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/visa_report_20251127_132025.md @@ -10,4 +10,4 @@ The execution of the plan culminated in the successful creation of a comprehensi 5. **Completion Status**: The task was marked complete, ensuring the workflow concluded as planned, with the result ready for stakeholder review and decision-making processes. -Overall, the plan achieved its objectives, providing valuable insights into Visa's financial health and strategic outlook. \ No newline at end of file +Overall, the plan achieved its objectives, providing valuable insights into Visa's financial health and strategic outlook. diff --git a/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/walmart_report_20251127_125312.md b/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/walmart_report_20251127_125312.md index 3de7e44..7d806d4 100644 --- a/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/walmart_report_20251127_125312.md +++ b/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/walmart_report_20251127_125312.md @@ -6,4 +6,4 @@ The execution of the plan resulted in the successful creation and documentation 3. **Report Generation**: A professional and fact-based markdown report was produced by the report_writer, summarizing the analysis and insights into a cohesive document containing an executive summary, financial performance details, recent noteworthy developments, investment analysis, risk factors, and a comprehensive valuation assessment. -Ultimately, the report, titled "Walmart - Comprehensive Financial Analysis," was saved successfully under the specified file path, marking the completion of the plan (is_complete=true). This report provides stakeholders with an in-depth understanding of Walmart's current market standing, strategic positioning, and potential investment outlook. \ No newline at end of file +Ultimately, the report, titled "Walmart - Comprehensive Financial Analysis," was saved successfully under the specified file path, marking the completion of the plan (is_complete=true). This report provides stakeholders with an in-depth understanding of Walmart's current market standing, strategic positioning, and potential investment outlook. diff --git a/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/waste_management,_inc._report_20251202_131842.md b/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/waste_management,_inc._report_20251202_131842.md index a782379..9b72d9f 100644 --- a/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/waste_management,_inc._report_20251202_131842.md +++ b/examples/mcp-agent/mcp_financial_analyzer_benchmark/company_reports/waste_management,_inc._report_20251202_131842.md @@ -1 +1 @@ -The stock analysis report for Waste Management, Inc. has been successfully created and saved to "company_reports/waste_management,_inc._report_20251202_131842.md". The report provides a comprehensive overview of the company's financial performance, market position, and investment potential, highlighting both its strengths in essential services and renewable energy growth, and its weaknesses in terms of high valuation and recent earnings misses. The analysis indicates a "Medium" recommendation, advising caution due to the premium valuation despite a stable business model and growth prospects. \ No newline at end of file +The stock analysis report for Waste Management, Inc. has been successfully created and saved to "company_reports/waste_management,_inc._report_20251202_131842.md". The report provides a comprehensive overview of the company's financial performance, market position, and investment potential, highlighting both its strengths in essential services and renewable energy growth, and its weaknesses in terms of high valuation and recent earnings misses. The analysis indicates a "Medium" recommendation, advising caution due to the premium valuation despite a stable business model and growth prospects. diff --git a/examples/mcp-agent/mcp_financial_analyzer_benchmark/main.py b/examples/mcp-agent/mcp_financial_analyzer_benchmark/main.py index d973862..dd26b04 100644 --- a/examples/mcp-agent/mcp_financial_analyzer_benchmark/main.py +++ b/examples/mcp-agent/mcp_financial_analyzer_benchmark/main.py @@ -634,7 +634,7 @@ async def main(): for server in context.config.mcp.servers.values(): if server.args: server.args = [os.path.expandvars(arg) for arg in server.args] - + requested_chain = _parse_search_provider_chain(REQUESTED_SEARCH_PROVIDERS) available_search_servers: list[str] = [] unavailable_providers: list[str] = [] @@ -645,7 +645,7 @@ async def main(): available_search_servers.append(server_name) else: unavailable_providers.append(provider) - + if not available_search_servers: logger.error( "No search MCP servers were configured for the requested providers (%s). " @@ -653,33 +653,33 @@ async def main(): ", ".join(requested_chain), ) return False - + if unavailable_providers: logger.warning( "Skipping unavailable search providers: %s", ", ".join(unavailable_providers), ) - + logger.info( "Search providers enabled (in order): %s", ", ".join(_search_provider_label(name) for name in available_search_servers), ) - + search_provider_description = _describe_search_chain(available_search_servers) search_tool_usage = _describe_search_tool_usage(available_search_servers) research_server_names = list(available_search_servers) if "fetch" not in research_server_names: research_server_names.append("fetch") - + run_metadata = None # Filled in after the LLM backend loads successfully. - + canonical_backend: str | None = None try: llm_factory, canonical_backend = _load_llm_factory(LLM_BACKEND) except (ImportError, AttributeError, ValueError) as err: logger.error(f"Unable to load LLM backend '{LLM_BACKEND}': {err}") return False - + model_name = LLM_MODEL_OVERRIDE or _default_model_for_backend( context.config, canonical_backend ) @@ -689,12 +689,12 @@ async def main(): "Use --llm-model to provide one." ) return False - + logger.info( f"Using LLM backend '{LLM_BACKEND}' " f"(canonical: {canonical_backend or 'custom'}) with model '{model_name}'" ) - + run_metadata = _build_run_metadata( run_id=timestamp, llm_backend=LLM_BACKEND, @@ -708,16 +708,16 @@ async def main(): run_metadata["otel_exporter"] = ( f"remote_otlp:{OTEL_REMOTE_ENDPOINT}" if OTEL_REMOTE_ENDPOINT else "local_file" ) - + # Configure filesystem server to use current directory if "filesystem" in context.config.mcp.servers: context.config.mcp.servers["filesystem"].args.extend([os.getcwd()]) logger.info("Filesystem server configured") else: logger.warning("Filesystem server not configured - report saving may fail") - + # --- SPECIALIZED AGENT DEFINITIONS --- - + scope_note = ( "This run is a QUICK sanity check. Gather just enough factual data to " "prove the workflow works—prioritize accuracy over volume, stop once the " @@ -725,71 +725,71 @@ async def main(): if SANITY_MODE else "Collect the full data pack required for a comprehensive briefing." ) - + news_scope = ( f"Provide {NEWS_ITEMS_REQUIRED} recent, well-sourced headlines." ) - + # Data collection agent that gathers comprehensive financial information research_agent = Agent( name="data_collector", instruction=f"""You are a comprehensive financial data collector for {COMPANY_NAME}. {scope_note} - + Use {search_provider_description} together with fetch to gather the requested facts in the order listed. Stick to the available MCP functions exactly as named: {search_tool_usage} - + Prefer the most recent data and stop once each section has concrete numbers. - + **REQUIRED DATA TO COLLECT:** - + 1. **Current Market Data**: Search: "{COMPANY_NAME} stock price today current" Search: "{COMPANY_NAME} trading volume market data" Extract: Current price, daily change ($ and %), trading volume, 52-week range - + 2. **Latest Earnings Information**: Search: "{COMPANY_NAME} latest quarterly earnings results" Search: "{COMPANY_NAME} earnings vs estimates beat miss" Extract: EPS actual vs estimate, revenue actual vs estimate, beat/miss percentages - + 3. **Recent Financial News**: Search: "{COMPANY_NAME} financial news latest week" Search: "{COMPANY_NAME} analyst ratings upgrade downgrade" Extract: {news_scope} - + 4. **Financial Metrics**: Search: "{COMPANY_NAME} PE ratio market cap financial metrics" Extract: P/E ratio, market cap, key financial ratios - + **OUTPUT FORMAT:** Organize your findings in these exact sections: - + ## CURRENT MARKET DATA - Stock Price: $XXX.XX (±X.XX, ±X.X%) - Trading Volume: X.X million (vs avg X.X million) - 52-Week Range: $XXX.XX - $XXX.XX - Market Cap: $XXX billion - Source: [URL and date] - + ## LATEST EARNINGS - EPS: $X.XX actual vs $X.XX estimate (beat/miss by X%) - Revenue: $XXX billion actual vs $XXX billion estimate (beat/miss by X%) - Year-over-Year Growth: X% - Quarter: QX YYYY - Source: [URL and date] - + ## RECENT NEWS (Last 7 Days) 1. [Headline] - [Date] - [Source] - [Impact: Positive/Negative/Neutral] 2. [Headline] - [Date] - [Source] - [Impact: Positive/Negative/Neutral] 3. [Continue until you reach the required number of items] - + ## KEY FINANCIAL METRICS - P/E Ratio: XX.X - Market Cap: $XXX billion - [Other available metrics] - Source: [URL and date] - + **CRITICAL REQUIREMENTS:** - Use EXACT figures, not approximations - Include source URLs for verification @@ -799,7 +799,7 @@ async def main(): server_names=research_server_names, ) # research_agent.attach_llm(GoogleAugmentedLLM) - + # Quality control agent that enforces strict data standards evaluator_scope = ( "Since this is a sanity-check run, treat the output as GOOD if the essential " @@ -807,76 +807,76 @@ async def main(): if SANITY_MODE else "Hold the research to the full EXCELLENT standard." ) - + research_evaluator = Agent( name="data_evaluator", instruction=f"""You are a strict financial data quality evaluator for {COMPANY_NAME} research. {evaluator_scope} - + **EVALUATION CRITERIA:** - + 1. **COMPLETENESS CHECK** (Must have ALL of these): ✓ Current stock price with exact dollar amount and percentage change ✓ Latest quarterly EPS with actual vs estimate comparison - ✓ Latest quarterly revenue with actual vs estimate comparison + ✓ Latest quarterly revenue with actual vs estimate comparison ✓ At least 3 recent financial news items with dates and sources ✓ Key financial metrics (P/E ratio, market cap) ✓ All data has proper source citations with URLs - + 2. **ACCURACY CHECK**: ✓ Numbers are specific (not "around" or "approximately") ✓ Dates are recent and clearly stated ✓ Sources are credible financial websites ✓ No conflicting information without explanation - + 3. **CURRENCY CHECK**: ✓ Stock price data is from today or latest trading day ✓ Earnings data is from most recent quarter ✓ News items are from last 7 days (or most recent available) - + **RATING GUIDELINES:** - + - **EXCELLENT**: All criteria met perfectly, comprehensive data, multiple source verification - **GOOD**: All required data present, good quality sources, minor gaps acceptable - **FAIR**: Most required data present but missing some elements or has quality issues - **POOR**: Missing critical data (stock price, earnings, or major sources), unreliable sources - + **EVALUATION OUTPUT FORMAT:** - + COMPLETENESS: [EXCELLENT/GOOD/FAIR/POOR] - Stock price data: [Present/Missing] - [Details] - - Earnings data: [Present/Missing] - [Details] + - Earnings data: [Present/Missing] - [Details] - News coverage: [Present/Missing] - [Details] - Financial metrics: [Present/Missing] - [Details] - Source quality: [Excellent/Good/Fair/Poor] - [Details] - + ACCURACY: [EXCELLENT/GOOD/FAIR/POOR] - Data specificity: [Comments] - Source credibility: [Comments] - Data consistency: [Comments] - + CURRENCY: [EXCELLENT/GOOD/FAIR/POOR] - Stock data recency: [Comments] - Earnings recency: [Comments] - News recency: [Comments] - + OVERALL RATING: [EXCELLENT/GOOD/FAIR/POOR] - + **IMPROVEMENT FEEDBACK:** [Specific instructions for what needs to be improved, added, or fixed] [If rating is below GOOD, provide exact search queries needed] [List any missing data points that must be found] - + **CRITICAL RULE**: If ANY of these are missing, overall rating cannot exceed FAIR: - Exact current stock price with change - - Latest quarterly EPS actual vs estimate + - Latest quarterly EPS actual vs estimate - Latest quarterly revenue actual vs estimate - At least 2 credible news sources from recent period """, server_names=[], ) # research_evaluator.attach_llm(GoogleAugmentedLLM) - + # Create the research quality control component research_quality_controller = EvaluatorOptimizerLLM( optimizer=research_agent, @@ -885,7 +885,7 @@ async def main(): min_rating=RESEARCH_MIN_RATING, max_refinements=RESEARCH_MAX_REFINEMENTS, ) - + # Financial analysis agent that provides investment insights analyst_scope = ( "Keep this analysis brief (2 short paragraphs max) and highlight only the " @@ -893,51 +893,51 @@ async def main(): if SANITY_MODE else "Provide the full analysis outlined below." ) - + analyst_agent = Agent( name="financial_analyst", instruction=f"""You are a senior financial analyst providing investment analysis for {COMPANY_NAME}. - + Based on the verified, high-quality data provided, create a comprehensive analysis. {analyst_scope} - + **1. STOCK PERFORMANCE ANALYSIS** - Analyze current price movement and trading patterns - Compare to historical performance and volatility - Assess volume trends and market sentiment indicators - - **2. EARNINGS ANALYSIS** + + **2. EARNINGS ANALYSIS** - Evaluate earnings beat/miss significance - Analyze revenue growth trends and sustainability - Compare to guidance and analyst expectations - Identify key performance drivers - + **3. NEWS IMPACT ASSESSMENT** - Synthesize how recent news affects investment outlook - Identify market sentiment shifts - Highlight potential catalysts or risk factors - + **4. INVESTMENT THESIS DEVELOPMENT** - + **BULL CASE (Top 3 Strengths)**: 1. [Strength with supporting data and metrics] 2. [Strength with supporting data and metrics] 3. [Strength with supporting data and metrics] - + **BEAR CASE (Top 3 Concerns)**: 1. [Risk with supporting evidence and impact assessment] - 2. [Risk with supporting evidence and impact assessment] + 2. [Risk with supporting evidence and impact assessment] 3. [Risk with supporting evidence and impact assessment] - + **5. VALUATION PERSPECTIVE** - Current valuation metrics analysis (P/E, etc.) - Historical valuation context - Fair value assessment based on fundamentals - + **6. RISK ASSESSMENT** - Company-specific operational risks - Market/sector risks and headwinds - Regulatory or competitive threats - + **OUTPUT REQUIREMENTS:** - Support all conclusions with specific data points - Use exact numbers and percentages from the research @@ -948,147 +948,147 @@ async def main(): server_names=[], ) # analyst_agent.attach_llm(GoogleAugmentedLLM) - + # Report generation agent that creates institutional-quality documents report_date = datetime.now().strftime("%B %d, %Y at %I:%M %p EST") - + if SANITY_MODE: report_instruction = f"""Create a concise, sanity-check markdown snapshot for {COMPANY_NAME}. - + **GOAL:** Confirm that the workflow surfaced real data. Use numbered lists or short paragraphs—keep the entire document under 400 words. - + # {COMPANY_NAME} Quick Financial Snapshot **Report Date:** {report_date} **Mode:** Sanity Check - + ## Market Pulse - Price + intraday change - Volume vs average - 52-week range position - + ## Earnings Pulse - Latest EPS actual vs estimate (state beat/miss) - Latest revenue actual vs estimate - YOY growth callout - + ## Headlines to Watch - Bullet the {NEWS_ITEMS_REQUIRED} news items with source + impact - + ## Key Metrics & Takeaways - P/E, market cap, or other notable ratios - 2 brief bullets on bullish/concern items - - Wrap up with a single-sentence overall assessment plus confidence level (High/Med/Low). + + Wrap up with a single-sentence overall assessment plus confidence level (High/Med/Low). Save the markdown to: {output_path} """ else: report_instruction = f"""Create a comprehensive, institutional-quality financial report for {COMPANY_NAME}. - + **REPORT STRUCTURE** (Use exactly this format): - + # {COMPANY_NAME} - Comprehensive Financial Analysis **Report Date:** {report_date} **Analyst:** AI Financial Research Team - + ## Executive Summary **Current Price:** $XXX.XX (±$X.XX, ±X.X% today) - **Market Cap:** $XXX.X billion + **Market Cap:** $XXX.X billion **Investment Thesis:** [2-3 sentence summary of key investment outlook] **Recommendation:** [Overall assessment with confidence level: High/Medium/Low] - + --- - + ## Current Market Performance - + ### Trading Metrics - **Stock Price:** $XXX.XX (±$X.XX, ±X.X% today) - **Trading Volume:** X.X million shares (vs X.X million avg) - - **52-Week Range:** $XXX.XX - $XXX.XX + - **52-Week Range:** $XXX.XX - $XXX.XX - **Current Position:** XX% of 52-week range - **Market Capitalization:** $XXX.X billion - + ### Technical Analysis [Analysis of price trends, volume patterns, momentum indicators] - + --- - + ## Financial Performance - + ### Latest Quarterly Results - **Earnings Per Share:** $X.XX actual vs $X.XX estimated (beat/miss by X.X%) - **Revenue:** $XXX.X billion actual vs $XXX.X billion estimated (beat/miss by X.X%) - **Year-over-Year Growth:** Revenue +/-X.X%, EPS +/-X.X% - **Quarter:** QX YYYY results - + ### Key Financial Metrics - **Price-to-Earnings Ratio:** XX.X - **Market Valuation:** [Analysis of current valuation vs historical/peers] - + --- - + ## Recent Developments - + ### Market-Moving News (Last 7 Days) [List 3-5 key news items with dates, sources, and impact analysis] - + ### Analyst Activity [Recent upgrades/downgrades, price target changes, consensus outlook] - + --- - + ## Investment Analysis - + ### Bull Case - Key Strengths 1. **[Strength Title]:** [Detailed explanation with supporting data] - 2. **[Strength Title]:** [Detailed explanation with supporting data] + 2. **[Strength Title]:** [Detailed explanation with supporting data] 3. **[Strength Title]:** [Detailed explanation with supporting data] - - ### Bear Case - Key Concerns + + ### Bear Case - Key Concerns 1. **[Risk Title]:** [Detailed explanation with potential impact] 2. **[Risk Title]:** [Detailed explanation with potential impact] 3. **[Risk Title]:** [Detailed explanation with potential impact] - + ### Valuation Assessment [Current valuation analysis, fair value estimate, historical context] - + --- - + ## Risk Factors - + ### Company-Specific Risks - [Operational, competitive, management risks] - - ### Market & Sector Risks + + ### Market & Sector Risks - [Economic, industry, regulatory risks] - + --- - + ## Investment Conclusion - + ### Summary Assessment [Balanced summary of key investment points] - + ### Overall Recommendation [Clear recommendation with rationale and confidence level] - + ### Price Target/Fair Value [If sufficient data available for valuation estimate] - + --- - + ## Data Sources & Methodology - + ### Sources Used [List all data sources with URLs and timestamps] - - ### Data Quality Notes + + ### Data Quality Notes [Any limitations, assumptions, or data quality considerations] - + ### Report Disclaimers *This report is for informational purposes only and should not be considered as personalized investment advice. Past performance does not guarantee future results. Please consult with a qualified financial advisor before making investment decisions.* - + --- - + **FORMATTING REQUIREMENTS:** - Use clean markdown formatting with proper headers - Include exact dollar amounts ($XXX.XX) and percentages (XX.X%) @@ -1096,71 +1096,71 @@ async def main(): - Maintain professional, objective tone - Length: 1200-1800 words - Save to file: {output_path} - + **CRITICAL:** Ensure all data comes directly from the verified research. Do not add speculative information not supported by the collected data. """ - + report_writer = Agent( name="report_writer", instruction=report_instruction, server_names=[], ) # report_writer.attach_llm(GoogleAugmentedLLM) - + # --- CREATE THE ORCHESTRATOR --- run_mode = "sanity-check mode" if SANITY_MODE else "full-report mode" logger.info( f"Initializing stock analysis workflow for {COMPANY_NAME} ({run_mode})" ) - + # Configure the orchestrator with our specialized agents pipeline_agents = [research_quality_controller] if not SANITY_MODE: pipeline_agents.append(analyst_agent) pipeline_agents.append(report_writer) - + orchestrator = Orchestrator( llm_factory=llm_factory, available_agents=pipeline_agents, plan_type="iterative" if SANITY_MODE else "full", ) - + # Define the comprehensive analysis task if SANITY_MODE: task = f"""Create a quick sanity-check stock snapshot for {COMPANY_NAME}: - + 1. Use 'research_quality_controller' once (with automatic evaluation) to gather: - Today's stock price, change %, and volume vs average - Latest EPS + revenue actual vs estimate - {NEWS_ITEMS_REQUIRED} timely headlines with URLs - Two valuation metrics (ex: P/E, market cap) - + 2. Pass the verified notes to 'report_writer' so it produces a concise markdown file at "{output_path}" following the quick snapshot template. - + The goal is to produce trustworthy data with minimal latency—skip deep dives, but do include precise figures and citations. - + After you save the markdown report to "{output_path}", explicitly mark the plan complete (is_complete=true) so the workflow can stop.""" else: task = f"""Create a high-quality stock analysis report for {COMPANY_NAME} by following these steps: - - 1. Use the EvaluatorOptimizerLLM component (named 'research_quality_controller') to gather high-quality - financial data about {COMPANY_NAME}. This component will automatically evaluate + + 1. Use the EvaluatorOptimizerLLM component (named 'research_quality_controller') to gather high-quality + financial data about {COMPANY_NAME}. This component will automatically evaluate and improve the research until it reaches GOOD quality. - + Ask for: - Current stock price and recent movement - Latest quarterly earnings results and performance vs expectations - Recent news and developments - + 2. Use the financial_analyst to analyze this research data and identify key insights. - + 3. Use the report_writer to create a comprehensive stock report and save it to: "{output_path}" - + The final report should be professional, fact-based, and include all relevant financial information. - + After you save the markdown report to "{output_path}", explicitly mark the plan complete (is_complete=true) so the workflow can stop.""" - + # Execute the analysis workflow logger.info("Starting the stock analysis workflow") run_succeeded = False @@ -1173,25 +1173,25 @@ async def main(): temperature=0.4 if SANITY_MODE else 0.7, use_history=False, ) - + report_markdown = await orchestrator.generate_str( message=task, request_params=orchestrator_params, ) - + # Persist report locally since filesystem tool is not used directly with open(output_path, "w", encoding="utf-8") as report_file: report_file.write(report_markdown) - + if not report_markdown or not report_markdown.strip(): raise ValueError( "Generated report was empty; treating workflow as a failure." ) # TODO: Add semantic validation to ensure the report meets benchmark criteria. - + logger.info(f"Report successfully generated: {output_path}") run_succeeded = True - + except Exception as e: workflow_error = str(e) logger.error(f"Error during workflow execution: {workflow_error}") @@ -1217,7 +1217,7 @@ async def main(): run_metadata["workflow_status"], ) _write_trace_metadata(logger, new_trace_logs, run_metadata, output_path) - + return run_succeeded diff --git a/examples/mcp-agent/mcp_financial_analyzer_benchmark/mcp_agent.config.yaml b/examples/mcp-agent/mcp_financial_analyzer_benchmark/mcp_agent.config.yaml index 81463df..865bd65 100644 --- a/examples/mcp-agent/mcp_financial_analyzer_benchmark/mcp_agent.config.yaml +++ b/examples/mcp-agent/mcp_financial_analyzer_benchmark/mcp_agent.config.yaml @@ -26,7 +26,7 @@ mcp: # Fetch server for basic web retrieval fetch: command: "mcp-server-fetch" - + # Google Search MCP server g-search: command: "./bin/g-search-headless.mjs" @@ -35,7 +35,7 @@ mcp: PLAYWRIGHT_HEADLESS: "1" NODE_PATH: "/home/cheny0y/.nvm/versions/node/v24.11.1/lib/node_modules" G_SEARCH_GLOBAL_NODE_ROOT: "/home/cheny0y/.nvm/versions/node/v24.11.1/lib/node_modules" - + # Filesystem server for writing reports filesystem: command: "mcp-server-filesystem" diff --git a/examples/mcp-agent/mcp_financial_analyzer_benchmark/sample_report.md b/examples/mcp-agent/mcp_financial_analyzer_benchmark/sample_report.md index 051f2fb..18bd5ee 100644 --- a/examples/mcp-agent/mcp_financial_analyzer_benchmark/sample_report.md +++ b/examples/mcp-agent/mcp_financial_analyzer_benchmark/sample_report.md @@ -4,7 +4,7 @@ ## Executive Summary **Current Price:** $360.67 (±$17.54, ±4.7% today) -**Market Cap:** $16.62 billion +**Market Cap:** $16.62 billion **Investment Thesis:** Duolingo presents a compelling growth potential with strong revenue and earnings performance, driven by increased user engagement and product diversification. However, its high P/E ratio indicates significant growth expectations already priced in, warranting careful consideration. **Recommendation:** Cautious optimism given high market valuation, with a Medium confidence level due to strong financials balanced by valuation concerns. @@ -15,7 +15,7 @@ ### Trading Metrics - **Stock Price:** $360.67 (±$17.54, ±4.7% today) - **Trading Volume:** 829.02K shares (vs 841.06K avg) -- **52-Week Range:** $145.05 - $544.93 +- **52-Week Range:** $145.05 - $544.93 - **Current Position:** 66% of 52-week range - **Market Capitalization:** $16.62 billion @@ -57,7 +57,7 @@ Recent analyst downgrade has impacted Duolingo's stock, but buoyant earnings and 2. **Expanding User Base:** Continued growth in user engagement and monetization suggests a sustained revenue trajectory. 3. **Strong Financial Health:** Low debt-to-equity ratio of 0.06 underscores financial stability. -### Bear Case - Key Concerns +### Bear Case - Key Concerns 1. **High P/E Ratio:** At 188.95, Duolingo's valuation may not be sustainable if growth slows, posing a risk of correction. 2. **Declining Trading Volume:** The marked drop in trading volume could indicate waning investor interest. 3. **Sensitivity to Analyst Opinions:** The stock's recent decline following a downgrade demonstrates vulnerability to external analyst perceptions. @@ -73,7 +73,7 @@ Duolingo's current valuation, with a P/E of 188.95, reflects high growth expecta - Operational risks from reliance on sustained user engagement. - Competitive pressures in the online education space. -### Market & Sector Risks +### Market & Sector Risks - Regulatory changes affecting the online education landscape. - Economic downturns impacting consumer discretionary spending. @@ -95,16 +95,16 @@ No fair value estimate provided, given the high variability and market condition ## Data Sources & Methodology ### Sources Used -- [Yahoo Finance](https://finance.yahoo.com/news/duolingo-stock-posing-attractive-entry-182029389.html) - Jul 16, 2025 +- [Yahoo Finance](https://finance.yahoo.com/news/duolingo-stock-posing-attractive-entry-182029389.html) - Jul 16, 2025 - [Yahoo Finance](https://finance.yahoo.com/news/duolingo-inc-duol-q1-earnings-211507492.html) - Date of report - [AInvest](https://www.ainvest.com/news/duolingo-trading-volume-drops-42-77-223-million-ranks-454th-stock-price-gain-2507/) - [MarketBeat](https://www.marketbeat.com/instant-alerts/duolingo-nasdaqduol-trading-down-46-following-analyst-downgrade-2025-07-08/) - [Robinhood](https://robinhood.com/stocks/DUOL/) -### Data Quality Notes +### Data Quality Notes Information is based on up-to-date and verified sources for accuracy. Limitations may exist due to market volatility and data gathering timings. ### Report Disclaimers *This report is for informational purposes only and should not be considered as personalized investment advice. Past performance does not guarantee future results. Please consult with a qualified financial advisor before making investment decisions.* ---- \ No newline at end of file +--- diff --git a/examples/mcp-agent/temporal/.gitignore b/examples/mcp-agent/temporal/.gitignore index 6a81b5a..6e25f61 100644 --- a/examples/mcp-agent/temporal/.gitignore +++ b/examples/mcp-agent/temporal/.gitignore @@ -1 +1 @@ -mcp_agent.secrets.yaml \ No newline at end of file +mcp_agent.secrets.yaml diff --git a/plot/.gitignore b/plot/.gitignore index f81463a..cbc3a56 100644 --- a/plot/.gitignore +++ b/plot/.gitignore @@ -1,3 +1,3 @@ figures_bak/ figures/ -deprecated/ \ No newline at end of file +deprecated/ diff --git a/plot/README.md b/plot/README.md index 6c55594..a447002 100644 --- a/plot/README.md +++ b/plot/README.md @@ -10,7 +10,7 @@ Paths below are relative to that directory. Install dependencies from `requireme ## Data download -Figures in our paper can be reproduced using our public dataset on HuggingFace, which +Figures in our paper can be reproduced using our public dataset on HuggingFace, which is available at: https://huggingface.co/datasets/kaust-generative-ai/maestro-mas-benchmark. Before running any plotting scripts, in this directory, run: diff --git a/plot/lib/analyzer.py b/plot/lib/analyzer.py index 1c969b9..4874388 100644 --- a/plot/lib/analyzer.py +++ b/plot/lib/analyzer.py @@ -25,7 +25,7 @@ class MetricsAnalyzer: """Main class for analyzing and visualizing OpenTelemetry metrics.""" - + def __init__( self, trace_file: Optional[str] = None, @@ -38,7 +38,7 @@ def __init__( tags: Optional[Iterable[str]] = None, ): """Initialize the analyzer. - + Args: trace_file: Path to trace JSONL file (if None, uses traces_dir based on analysis_mode) metrics_file: Path to metrics JSONL file (if None, uses metrics_dir based on analysis_mode) @@ -52,18 +52,18 @@ def __init__( # Use current working directory as base if not specified if base_dir is None: base_dir = str(Path.cwd()) - + # Set default directories relative to base_dir if traces_dir is None: traces_dir = str(Path(base_dir) / "traces") else: traces_dir = str(Path(base_dir) / traces_dir) if not Path(traces_dir).is_absolute() else traces_dir - + if metrics_dir is None: metrics_dir = str(Path(base_dir) / "metrics") else: metrics_dir = str(Path(base_dir) / metrics_dir) if not Path(metrics_dir).is_absolute() else metrics_dir - + self.traces_dir = traces_dir self.metrics_dir = metrics_dir self.analysis_mode = analysis_mode @@ -71,7 +71,7 @@ def __init__( self.tags = list(tags) if tags is not None else None self._traces_parquet = Path(traces_dir).suffix.lower() == ".parquet" self._metrics_parquet = Path(metrics_dir).suffix.lower() == ".parquet" - + if trace_file: self.trace_files = [trace_file] elif self._traces_parquet: @@ -93,20 +93,20 @@ def sort_key(p): else: # Fallback to mtime for files without timestamp return (0, Path(p).stat().st_mtime) - + all_trace_files = sorted(all_trace_files_raw, key=sort_key, reverse=True) if not all_trace_files: raise ValueError( f"No trace files found in {traces_dir}/ (expected .jsonl or .json files)" ) - + if analysis_mode == 'latest': self.trace_files = [all_trace_files[0]] # Use latest only elif analysis_mode == 'per_run': self.trace_files = all_trace_files # Analyze each separately else: raise ValueError(f"Invalid analysis_mode: {analysis_mode}. Must be 'latest' or 'per_run'") - + if metrics_file: self.metrics_files = [metrics_file] elif self._metrics_parquet: @@ -120,15 +120,15 @@ def sort_key(p): self.metrics_files = all_metrics_files # Analyze each separately else: self.metrics_files = [] - + # For backward compatibility self.trace_file = self.trace_files[0] if self.trace_files else None self.metrics_file = self.metrics_files[0] if self.metrics_files else None - + self.traces: List[Dict] = [] self.metrics: List[Dict] = [] self._loaded = False - + def load_data(self): """Load trace and metrics data.""" if self._traces_parquet or self._metrics_parquet: @@ -196,12 +196,12 @@ def load_data(self): # For per-run analysis, we need to handle two cases: # 1. Monolithic system (marketing-agency): one trace file = one run # 2. Distributed system (dist_vs_monolithic_without_kagent): multiple trace files per run (one per agent) - + # Group trace files by timestamp (extracted from filename or file modification time) # Pattern: agent-name_YYYYMMDD_HHMMSS.(jsonl|json) from collections import defaultdict import re - + # Group by timestamp (extracted from filename) runs_by_timestamp = defaultdict(list) for trace_file in self.trace_files: @@ -217,15 +217,15 @@ def load_data(self): mtime = Path(trace_file).stat().st_mtime timestamp = str(int(mtime // 60)) # Round to minute runs_by_timestamp[timestamp].append(trace_file) - + # If we have multiple files per timestamp, it's a distributed system # Otherwise, it's a monolithic system (one file per run) is_distributed = any(len(files) > 1 for files in runs_by_timestamp.values()) - + # For distributed systems, files with the same timestamp belong to the same run # No need to merge by time window - unified timestamp ensures accuracy # (The unified timestamp is set via OTEL_RUN_TIMESTAMP environment variable at startup) - + if is_distributed: # Distributed system: merge files with same/similar timestamp into one run print(f"Detected distributed system: grouping {len(self.trace_files)} trace files into runs by timestamp...") @@ -237,7 +237,7 @@ def load_data(self): # Monolithic system: one file per run (marketing-agency case) print(f"Detected monolithic system: {len(self.trace_files)} trace files (one per run)...") self.trace_files_grouped = [[f] for f in self.trace_files] - + # Group metrics files similarly if self.metrics_files: metrics_by_timestamp = defaultdict(list) @@ -251,11 +251,11 @@ def load_data(self): mtime = Path(metrics_file).stat().st_mtime timestamp = str(int(mtime // 60)) metrics_by_timestamp[timestamp].append(metrics_file) - + # For distributed systems, files with the same timestamp belong to the same run # No need to merge by time window - unified timestamp ensures accuracy # (The unified timestamp is set via OTEL_RUN_TIMESTAMP environment variable) - + if is_distributed: self.metrics_files_grouped = [] # Use the same order as trace_files_grouped to ensure alignment @@ -291,11 +291,11 @@ def load_data(self): self.metrics_files_grouped.append(matching_metrics if matching_metrics else []) else: self.metrics_files_grouped = [[]] * len(self.trace_files_grouped) - + # Load and merge files per run self.run_traces = [] self.run_metrics = [] - + print(f"\nLoading {len(self.trace_files_grouped)} runs for per-run analysis...") for run_idx, (trace_files_group, metrics_files_group) in enumerate(zip(self.trace_files_grouped, self.metrics_files_grouped)): # Merge all trace files for this run @@ -303,20 +303,20 @@ def load_data(self): for trace_file in trace_files_group: traces = load_traces(trace_file) merged_traces.extend(traces) - + # Merge all metrics files for this run merged_metrics = [] for metrics_file in metrics_files_group: metrics = load_metrics(metrics_file) merged_metrics.extend(metrics) - + # Use first trace file name as identifier run_id = Path(trace_files_group[0]).stem if trace_files_group else f"run_{run_idx+1}" self.run_traces.append((run_id, merged_traces)) self.run_metrics.append((run_id, merged_metrics)) - + print(f" Run {run_idx + 1}: {len(merged_traces)} spans, {len(merged_metrics)} metrics") - + total_spans = sum(len(traces) for _, traces in self.run_traces) total_metrics = sum(len(metrics) for _, metrics in self.run_metrics) print(f"Loaded {total_spans} total trace spans across {len(self.run_traces)} run(s)") @@ -324,7 +324,7 @@ def load_data(self): print(f"Loaded {total_metrics} total metric records across {len(self.run_metrics)} run(s)") else: print("Warning: No metrics data found") - + # For backward compatibility, also set merged data self.traces = [] for _, traces in self.run_traces: @@ -336,11 +336,11 @@ def load_data(self): # Latest mode - should only have one file if len(self.trace_files) != 1: raise ValueError(f"Expected 1 trace file in 'latest' mode, got {len(self.trace_files)}") - + print(f"Loading trace file: {self.trace_files[0]}") self.traces = load_traces(self.trace_files[0]) print(f"Loaded {len(self.traces)} trace spans") - + # Load metrics file self.metrics = [] if self.metrics_files: @@ -351,31 +351,31 @@ def load_data(self): print(f"Loaded {len(self.metrics)} metric records") else: print("Warning: No metrics file found") - + self._loaded = True - + def analyze(self, output_dir: str = "visualizations"): """Run complete analysis and generate all visualizations. - + Args: output_dir: Directory to save visualization files """ if not self._loaded: self.load_data() - + output_path = Path(output_dir) output_path.mkdir(exist_ok=True) - + print("\nGenerating visualizations...") print() - + if self.analysis_mode == 'per_run': # Per-run analysis mode (recommended for academic papers) print("=" * 80) print("Per-Run Analysis Mode (Recommended for Academic Papers)") print("=" * 80) print() - + # Analyze each run separately and compute statistics # For distributed systems, we've already grouped files by run in load_data() # Extract the merged trace/metrics data per run @@ -404,10 +404,10 @@ def analyze(self, output_dir: str = "visualizations"): print(f" Trace start: {first_trace_start_sec:.2f}, Metric start: {first_metric_ts:.2f}") run_trace_data.append(traces) run_metrics_data.append(matching_metrics if matching_metrics else None) - + # Use the new method that works with pre-loaded and grouped data results = self._analyze_per_run_from_data(run_trace_data, run_metrics_data) - + # Generate per-run visualizations with statistics plot_per_run_statistics(results, str(output_path)) else: @@ -419,17 +419,17 @@ def analyze(self, output_dir: str = "visualizations"): output_path=output_path, suffix="" ) - + print() print("=" * 80) print("All visualizations generated successfully!") print(f"Output directory: {output_path.absolute()}") print("=" * 80) - + def _generate_standard_visualizations(self, traces: List[Dict], metrics: Optional[List[Dict]], extracted_data: Optional[Dict], output_path: Path, suffix: str = ""): """Generate standard visualizations for a single run. - + Args: traces: List of trace spans metrics: List of metric records (can be None) @@ -444,13 +444,13 @@ def _generate_standard_visualizations(self, traces: List[Dict], metrics: Optiona total_tokens, per_agent_tokens = extract_token_consumption(traces) plot_token_consumption(total_tokens, per_agent_tokens, str(output_path / f"1_token_consumption{suffix}.pdf")) - + # 2. Delay Breakdown print("2. Delay Breakdown...") delays, per_agent_delays, inter_agent_delays, agent_llm_delays = extract_delay_breakdown(traces) plot_delay_breakdown(delays, per_agent_delays, inter_agent_delays, agent_llm_delays, str(output_path / f"2_delay_breakdown{suffix}.pdf")) - + # 3. CPU/Memory Usage if metrics: print("3. CPU/Memory Usage...") @@ -458,18 +458,18 @@ def _generate_standard_visualizations(self, traces: List[Dict], metrics: Optiona plot_cpu_memory_usage(usage, str(output_path / f"3_cpu_memory_usage{suffix}.pdf")) else: print("3. CPU/Memory Usage... (skipped - no metrics data)") - + # 4. Message Sizes print("4. Message Sizes...") sizes, per_agent_sizes, inter_agent_sizes, agent_llm_sizes = extract_message_sizes(traces) plot_message_sizes(sizes, per_agent_sizes, inter_agent_sizes, agent_llm_sizes, str(output_path / f"4_message_sizes{suffix}.pdf")) - + # 5. Events/Spans with CPU/Memory print("5. Events/Spans with CPU/Memory...") if metrics: usage = extract_cpu_memory_usage(metrics) - plot_events_with_cpu_memory(traces, usage, + plot_events_with_cpu_memory(traces, usage, str(output_path / f"6_events_with_cpu_memory{suffix}.pdf")) else: # Plot spans even without CPU/memory data @@ -501,7 +501,7 @@ def _generate_standard_visualizations(self, traces: List[Dict], metrics: Optiona extracted_data['tokens']['per_agent'], str(output_path / f"1_token_consumption{suffix}.pdf") ) - + # 2. Delay Breakdown print(f"2. Delay Breakdown{suffix.replace('_', ' ')}...") plot_delay_breakdown( @@ -511,7 +511,7 @@ def _generate_standard_visualizations(self, traces: List[Dict], metrics: Optiona extracted_data['delays']['agent_llm'], str(output_path / f"2_delay_breakdown{suffix}.pdf") ) - + # 3. CPU/Memory Usage if extracted_data['resource_usage']: print(f"3. CPU/Memory Usage{suffix.replace('_', ' ')}...") @@ -521,7 +521,7 @@ def _generate_standard_visualizations(self, traces: List[Dict], metrics: Optiona ) else: print(f"3. CPU/Memory Usage{suffix.replace('_', ' ')}... (skipped - no metrics data)") - + # 4. Message Sizes print(f"4. Message Sizes{suffix.replace('_', ' ')}...") plot_message_sizes( @@ -531,13 +531,13 @@ def _generate_standard_visualizations(self, traces: List[Dict], metrics: Optiona extracted_data['message_sizes']['agent_llm'], str(output_path / f"4_message_sizes{suffix}.pdf") ) - + # 5. Events/Spans with CPU/Memory print(f"5. Events/Spans with CPU/Memory{suffix.replace('_', ' ')}...") # Always use metrics if available to ensure traces and resource_usage are from the same run # This is critical for time alignment usage = None - + # First, try to use metrics directly if available # This ensures traces and metrics are from the same run if metrics and len(metrics) > 0: @@ -564,7 +564,7 @@ def _generate_standard_visualizations(self, traces: List[Dict], metrics: Optiona # resource_usage exists but is empty, this shouldn't happen but handle gracefully print(f" Warning: resource_usage exists but is empty (CPU: {len(cpu_data)}, Memory: {len(mem_data)})") usage = None - + if usage: plot_events_with_cpu_memory(traces, usage, str(output_path / f"6_events_with_cpu_memory{suffix}.pdf")) @@ -573,18 +573,18 @@ def _generate_standard_visualizations(self, traces: List[Dict], metrics: Optiona empty_usage = {'cpu': {'process': []}, 'memory': {'process': []}} plot_events_with_cpu_memory(traces, empty_usage, str(output_path / f"6_events_with_cpu_memory{suffix}.pdf")) - + def _analyze_per_run_from_data(self, run_trace_data: List[List[Dict]], run_metrics_data: List[Optional[List[Dict]]]) -> Dict: """Analyze per-run data that's already loaded and grouped. - + This is a wrapper around analyze_per_run that works with pre-loaded data instead of file paths. This is needed for distributed systems where multiple trace files need to be merged per run. - + Args: run_trace_data: List of trace data lists, one per run run_metrics_data: List of metrics data lists, one per run (can be None) - + Returns: Same format as analyze_per_run """ @@ -593,12 +593,12 @@ def _analyze_per_run_from_data(self, run_trace_data: List[List[Dict]], run_metri extract_call_graph, extract_call_sequence, ) - + results = { 'runs': [], 'statistics': {} } - + # Analyze each run for i, (traces, metrics) in enumerate(zip(run_trace_data, run_metrics_data)): # Extract metrics for this run @@ -608,14 +608,14 @@ def _analyze_per_run_from_data(self, run_trace_data: List[List[Dict]], run_metri 'traces': traces, 'metrics': metrics if metrics else [], } - + # Extract token consumption total_tokens, per_agent_tokens = extract_token_consumption(traces) run_data['tokens'] = { 'total': total_tokens, 'per_agent': per_agent_tokens } - + # Extract delay breakdown delays, per_agent_delays, inter_agent_delays, agent_llm_delays = extract_delay_breakdown(traces) run_data['delays'] = { @@ -624,7 +624,7 @@ def _analyze_per_run_from_data(self, run_trace_data: List[List[Dict]], run_metri 'inter_agent': inter_agent_delays, 'agent_llm': agent_llm_delays } - + # Extract message sizes sizes, per_agent_sizes, inter_agent_sizes, agent_llm_sizes = extract_message_sizes(traces) run_data['message_sizes'] = { @@ -633,25 +633,25 @@ def _analyze_per_run_from_data(self, run_trace_data: List[List[Dict]], run_metri 'inter_agent': inter_agent_sizes, 'agent_llm': agent_llm_sizes } - + # Extract CPU/memory usage if metrics and len(metrics) > 0: usage = extract_cpu_memory_usage(metrics) run_data['resource_usage'] = usage else: run_data['resource_usage'] = None - + # Extract call graph edges (set, unordered) call_graph_edges = extract_call_graph(traces) run_data['call_graph'] = call_graph_edges - + # Extract call sequence (ordered list, preserves calling order) call_sequence = extract_call_sequence(traces) run_data['call_sequence'] = call_sequence - + results['runs'].append(run_data) - + # Compute statistics across runs results['statistics'] = compute_cross_run_statistics(results['runs']) - + return results diff --git a/plot/lib/comparison.py b/plot/lib/comparison.py index 023bb8c..08dba0c 100644 --- a/plot/lib/comparison.py +++ b/plot/lib/comparison.py @@ -22,7 +22,7 @@ class ExampleMetrics: """Metrics data container for a single example.""" - + def __init__(self, example_name: str): self.example_name = example_name self.tokens = None @@ -43,7 +43,7 @@ def __init__(self, example_name: str): self.pairwise_jaccard = [] # All pairwise Jaccard similarity values self.traces = [] self.metrics = [] - + def extract_all_metrics(self, traces: List[Dict], metrics: Optional[List[Dict]] = None): """Extract all metrics from traces and metrics (latest mode).""" # Token consumption @@ -52,7 +52,7 @@ def extract_all_metrics(self, traces: List[Dict], metrics: Optional[List[Dict]] 'total': total_tokens, 'per_agent': per_agent_tokens } - + # Delay breakdown delays, per_agent_delays, inter_agent_delays, agent_llm_delays = extract_delay_breakdown(traces) self.delays = { @@ -61,7 +61,7 @@ def extract_all_metrics(self, traces: List[Dict], metrics: Optional[List[Dict]] 'inter_agent': inter_agent_delays, 'agent_llm': agent_llm_delays } - + # Message sizes sizes, per_agent_sizes, inter_agent_sizes, agent_llm_sizes = extract_message_sizes(traces) self.message_sizes = { @@ -70,19 +70,19 @@ def extract_all_metrics(self, traces: List[Dict], metrics: Optional[List[Dict]] 'inter_agent': inter_agent_sizes, 'agent_llm': agent_llm_sizes } - + # CPU/Memory usage if metrics: self.resource_usage = extract_cpu_memory_usage(metrics) else: self.resource_usage = None - + # Call graph self.call_graph = extract_call_graph(traces) - + self.traces = traces self.metrics = metrics or [] - + def extract_call_graphs_from_runs( self, run_traces: List[Tuple[str, List[Dict]]], @@ -118,7 +118,7 @@ def extract_call_graphs_from_runs( if len(edges_preview) > 5 else f" Graph edges: {edges_preview}" ) - + # Compute average pairwise normalized GED across all runs within this example if compute_ged: if len(self.call_graphs_per_run) >= 2: @@ -134,7 +134,7 @@ def extract_call_graphs_from_runs( pairwise_distances.append(normalized_ged) print(f" Pair ({i+1}, {j+1}): graph_ged={normalized_ged:.4f}") self.pairwise_normalized_ged = pairwise_distances - + # Average = mean of all pairwise distances # Total number of pairs: n*(n-1)/2 self.average_pairwise_normalized_ged = np.mean(pairwise_distances) if pairwise_distances else 0.0 @@ -203,10 +203,10 @@ def extract_call_graphs_from_runs( class MultiExampleCollector: """Multi-example metrics collector.""" - + def __init__(self): self.examples: Dict[str, ExampleMetrics] = {} - + def add_example( self, example_name: str, @@ -219,14 +219,14 @@ def add_example( tags: Optional[List[str]] = None, ) -> ExampleMetrics: """Add an example and extract its metrics. - + Args: example_name: Example name traces_dir: Path to traces directory metrics_dir: Path to metrics directory (optional) base_dir: Base directory (optional) analysis_mode: Analysis mode ('latest' or 'per_run') - + Returns: ExampleMetrics object """ @@ -240,10 +240,10 @@ def add_example( tags=tags, ) analyzer.load_data() - + # Create metrics object and extract data example_metrics = ExampleMetrics(example_name) - + if analysis_mode == 'per_run' and hasattr(analyzer, 'run_traces'): # per_run mode: extract call graph for each run and compute average pairwise normalized GED example_metrics.extract_call_graphs_from_runs( @@ -258,26 +258,26 @@ def add_example( else: # latest mode: extract only one call graph example_metrics.extract_all_metrics(analyzer.traces, analyzer.metrics) - + self.examples[example_name] = example_metrics return example_metrics - + def get_metric(self, example_name: str, metric_type: str): """Get specified metric for specified example. - + Args: example_name: Example name metric_type: Metric type ('tokens', 'delays', 'message_sizes', 'resource_usage', 'call_graph') - + Returns: Metric data """ if example_name not in self.examples: raise ValueError(f"Example '{example_name}' not found") - + example = self.examples[example_name] return getattr(example, metric_type, None) - + def get_all_examples(self) -> Dict[str, ExampleMetrics]: """Get all examples.""" return self.examples @@ -285,22 +285,22 @@ def get_all_examples(self) -> Dict[str, ExampleMetrics]: def compute_graph_size(edges: Set[Tuple[str, str]] | Dict[Tuple[str, str], int]) -> int: """Compute graph size (number of nodes). - + Args: edges: Edge set or weighted edge map (source, target) -> weight - + Returns: Graph size = number of nodes """ if not edges: return 0 - + # Collect all nodes nodes = set() for source, target in edges: nodes.add(source) nodes.add(target) - + return len(nodes) @@ -375,13 +375,13 @@ def compute_graph_edit_distance( graph2: Set[Tuple[str, str]] | Dict[Tuple[str, str], int] ) -> int: """Compute weighted graph edit distance between two call graphs. - + Graph edit distance = edge weight L1 distance + node insert/delete cost. - + Args: graph1: Edge set or weighted edge map for the first graph graph2: Edge set or weighted edge map for the second graph - + Returns: Graph edit distance (integer) """ @@ -402,18 +402,18 @@ def compute_normalized_graph_edit_distance( graph2: Set[Tuple[str, str]] | Dict[Tuple[str, str], int] ) -> float: """Compute normalized graph edit distance. - + Normalization = graph edit distance / total_call_weight - + Args: graph1: Edge set of the first graph graph2: Edge set of the second graph - + Returns: Normalized graph edit distance (0 means identical) """ edit_distance = compute_graph_edit_distance(graph1, graph2) - + weights1 = _to_weight_map(graph1) weights2 = _to_weight_map(graph2) total_edge_weight = sum(weights1.values()) + sum(weights2.values()) @@ -422,11 +422,11 @@ def compute_normalized_graph_edit_distance( nodes2 = {node for edge in weights2 for node in edge} total_node_weight = len(nodes1) + len(nodes2) total_weight = total_edge_weight + total_node_weight - + if total_weight == 0: # Both graphs are empty, consider them identical return 0.0 - + return edit_distance / total_weight @@ -434,7 +434,7 @@ def compute_normalized_graph_edit_distance( class CrossExampleVisualizer: """Cross-example comparison visualizer.""" - + def __init__(self, collector: MultiExampleCollector): self.collector = collector @@ -570,11 +570,11 @@ def _collect_call_graph_similarity_data( ) -> Optional[Dict[str, object]]: if examples is None: examples = list(self.collector.examples.keys()) - + if len(examples) < 1: print("Warning: Need at least 1 example for comparison") return None - + # Collect average pairwise normalized graph edit distance for each example example_avg_ged = {} example_avg_lcs = {} @@ -588,10 +588,10 @@ def _collect_call_graph_similarity_data( example_num_runs = {} example_graph_sizes = {} example_graph_details = {} - + for name in examples: example = self.collector.examples[name] - + # Get average pairwise normalized graph edit distance if example.average_pairwise_normalized_ged is not None: example_avg_ged[name] = example.average_pairwise_normalized_ged @@ -604,7 +604,7 @@ def _collect_call_graph_similarity_data( example_pairwise_lcs[name] = example.pairwise_lcs if example.pairwise_lcs else [0.0] example_pairwise_jaccard[name] = example.pairwise_jaccard if example.pairwise_jaccard else [0.0] example_num_runs[name] = len(example.call_graphs_per_run) - + # Compute average graph size across all runs if example.call_graphs_per_run: # Collect detailed information (node and edge counts) @@ -637,11 +637,11 @@ def _collect_call_graph_similarity_data( example_graph_details[name] = {'avg_nodes': 0, 'avg_edges': 0, 'unique_nodes': 0, 'unique_edges': 0} else: print(f"Warning: No call graph similarity data found for example '{name}' (may need per_run mode)") - + if not example_avg_ged: print("Warning: No examples with call graph similarity data found") return None - + example_names = list(example_avg_ged.keys()) return { 'example_names': example_names, @@ -671,7 +671,7 @@ def plot_call_graph_similarity_violins( data = self._collect_call_graph_similarity_data(examples) if data is None: return - + example_names = data['example_names'] pairwise_ged_values = data['pairwise_ged_values'] pairwise_lcs_values = data['pairwise_lcs_values'] @@ -777,17 +777,17 @@ def plot_call_graph_similarity_pairwise_panels( "Overall mean across examples: " f"Jaccard={mean_jaccard:.4f}, LCS={mean_lcs:.4f}" ) - + def plot_call_graph_similarity( self, output_file: str, examples: Optional[List[str]] = None ): """Plot call graph similarity comparison. - + Computes call graph similarity for each example (average pairwise normalized graph edit distance across all runs within each example) and displays results in a single figure. - + Args: output_file: Output file path examples: List of examples to compare (None means all examples) @@ -795,7 +795,7 @@ def plot_call_graph_similarity( data = self._collect_call_graph_similarity_data(examples) if data is None: return - + # Prepare data for visualization example_names = data['example_names'] avg_ged_values = data['avg_ged_values'] @@ -812,7 +812,7 @@ def plot_call_graph_similarity( avg_nodes = data['avg_nodes'] avg_edges = data['avg_edges'] example_graph_details = data['example_graph_details'] - + style = self._similarity_plot_style() panel_specs = self._similarity_panel_specs() scope_label = 'Example' @@ -950,25 +950,25 @@ def plot_call_graph_similarity( fig.savefig(output_file, dpi=300, bbox_inches='tight') plt.close(fig) print(f"Call graph similarity comparison plot saved to: {output_file}") - + def plot_token_comparison(self, output_file: str, examples: Optional[List[str]] = None): """Compare token consumption across multiple examples. - + Args: output_file: Output file path examples: List of examples to compare (None means all examples) """ if examples is None: examples = list(self.collector.examples.keys()) - + fig, axes = plt.subplots(2, 2, figsize=(16, 10)) - + # Collect data example_names = [] total_tokens = [] prompt_tokens = [] completion_tokens = [] - + for name in examples: tokens = self.collector.get_metric(name, 'tokens') if tokens: @@ -976,17 +976,17 @@ def plot_token_comparison(self, output_file: str, examples: Optional[List[str]] total_tokens.append(tokens['total']['total']) prompt_tokens.append(tokens['total']['prompt']) completion_tokens.append(tokens['total']['completion']) - + if not example_names: print("Warning: No token data found for comparison") return - + # Chart 1: Total token comparison (stacked bar chart) ax1 = axes[0, 0] x = np.arange(len(example_names)) width = 0.6 ax1.bar(x, prompt_tokens, width, label='Prompt Tokens', color='#3498db', alpha=0.8) - ax1.bar(x, completion_tokens, width, bottom=prompt_tokens, + ax1.bar(x, completion_tokens, width, bottom=prompt_tokens, label='Completion Tokens', color='#e74c3c', alpha=0.8) ax1.set_xlabel('Example', fontsize=12, fontweight='bold') ax1.set_ylabel('Token Count', fontsize=12, fontweight='bold') @@ -995,7 +995,7 @@ def plot_token_comparison(self, output_file: str, examples: Optional[List[str]] ax1.set_xticklabels(example_names, rotation=45, ha='right') ax1.legend() ax1.grid(axis='y', alpha=0.3) - + # Chart 2: Total token comparison (side-by-side bar chart) ax2 = axes[0, 1] width = 0.35 @@ -1008,7 +1008,7 @@ def plot_token_comparison(self, output_file: str, examples: Optional[List[str]] ax2.set_xticklabels(example_names, rotation=45, ha='right') ax2.legend() ax2.grid(axis='y', alpha=0.3) - + # Chart 3 removed (per-agent token comparison was too crowded) # Chart 3 (now statistics table) ax3 = axes[1, 0] @@ -1029,16 +1029,16 @@ def plot_token_comparison(self, output_file: str, examples: Optional[List[str]] table.set_fontsize(9) table.scale(1, 2) ax3.set_title('Token Statistics Summary', fontsize=14, fontweight='bold', pad=20) - + # Chart 4 left blank to avoid overcrowding ax4 = axes[1, 1] ax4.axis('off') - + plt.tight_layout() plt.savefig(output_file, dpi=300, bbox_inches='tight') plt.close() print(f"Token comparison plot saved to: {output_file}") - + def plot_custom_comparison( self, metric_extractor: callable, @@ -1047,7 +1047,7 @@ def plot_custom_comparison( examples: Optional[List[str]] = None ): """Custom comparison visualization. - + Args: metric_extractor: Function that takes ExampleMetrics object and returns metric value to compare plot_function: Function that takes (metric_values_dict, output_path) and performs plotting @@ -1056,12 +1056,12 @@ def plot_custom_comparison( """ if examples is None: examples = list(self.collector.examples.keys()) - + # Extract metrics for each example metric_values = {} for name in examples: example = self.collector.examples[name] metric_values[name] = metric_extractor(example) - + # Call custom plotting function plot_function(metric_values, output_file) diff --git a/plot/lib/extractors.py b/plot/lib/extractors.py index 430c3be..1047cf7 100644 --- a/plot/lib/extractors.py +++ b/plot/lib/extractors.py @@ -24,7 +24,7 @@ def _safe_int(value: object) -> int: def extract_token_consumption(traces: List[Dict]) -> Tuple[Dict[str, int], Dict[str, Dict[str, int]]]: """Extract token consumption from traces. - + Returns: (total_tokens, per_agent_tokens) total_tokens: {'prompt': int, 'completion': int, 'total': int} @@ -32,19 +32,19 @@ def extract_token_consumption(traces: List[Dict]) -> Tuple[Dict[str, int], Dict[ """ # Build span dictionary for parent lookup span_dict = {span['span_id']: span for span in traces} - + total_tokens = {'prompt': 0, 'completion': 0, 'total': 0} per_agent_tokens = defaultdict(lambda: {'prompt': 0, 'completion': 0, 'total': 0}) - + for span in traces: span_name = _safe_lower(span.get('name')) attrs = span.get('attributes', {}) - + # Check if this is an LLM call span # Priority 1: Use gen_ai.operation.name if available (per template) operation_name = _safe_lower(attrs.get('gen_ai.operation.name')) is_llm_span = operation_name == 'call_llm' - + # Priority 2: Fallback to span name pattern matching (for compatibility) if not is_llm_span: is_llm_span = ( @@ -54,13 +54,13 @@ def extract_token_consumption(traces: List[Dict]) -> Tuple[Dict[str, int], Dict[ attrs.get('gen_ai.usage.input_tokens') or attrs.get('gen_ai.usage.output_tokens') ) - + if is_llm_span: # Use prompt_token_count and candidates_token_count if available llm_response = attrs.get('gcp.vertex.agent.llm_response', '') prompt_tokens = attrs.get('gen_ai.usage.input_tokens', 0) or 0 completion_tokens = attrs.get('gen_ai.usage.output_tokens', 0) or 0 - + # Try to extract from llm_response if available if llm_response: try: @@ -71,13 +71,13 @@ def extract_token_consumption(traces: List[Dict]) -> Tuple[Dict[str, int], Dict[ completion_tokens = usage_metadata.get('candidates_token_count', completion_tokens) except: pass - + if prompt_tokens or completion_tokens: # Find agent name from span attributes or parent spans (search up the tree) agent_name = 'unknown' # First try to get from current span attributes agent_name = attrs.get('gen_ai.agent.name') or span.get('agent_name') or 'unknown' - + # If not found, search up parent chain if agent_name == 'unknown': current = span @@ -86,42 +86,42 @@ def extract_token_consumption(traces: List[Dict]) -> Tuple[Dict[str, int], Dict[ parent_id = current.get('parent_span_id') if not parent_id or parent_id not in span_dict: break - + parent_span = span_dict[parent_id] parent_name = _safe_lower(parent_span.get('name')) parent_attrs = parent_span.get('attributes', {}) - + # Check if parent has agent name parent_agent_name = parent_attrs.get('gen_ai.agent.name') or parent_span.get('agent_name') if parent_agent_name: agent_name = parent_agent_name break - + # Check if parent is an agent span (OpenAIAugmentedLLM, Agent, etc.) - if ('invoke_agent' in parent_name or + if ('invoke_agent' in parent_name or 'augmentedllm' in parent_name or ('agent' in parent_name and 'generate' in parent_name)): # Try to extract from parent name or attributes if parent_agent_name: agent_name = parent_agent_name break - + current = parent_span - + total_tokens['prompt'] += prompt_tokens total_tokens['completion'] += completion_tokens total_tokens['total'] += (prompt_tokens + completion_tokens) - + per_agent_tokens[agent_name]['prompt'] += prompt_tokens per_agent_tokens[agent_name]['completion'] += completion_tokens per_agent_tokens[agent_name]['total'] += (prompt_tokens + completion_tokens) - + return total_tokens, dict(per_agent_tokens) def extract_delay_breakdown(traces: List[Dict]) -> Tuple[Dict[str, List[float]], Dict[str, List[float]], Dict[Tuple[str, str], List[float]], Dict[str, List[float]]]: """Extract delay breakdown from traces. - + Returns: (component_delays, per_agent_delays, inter_agent_delays, agent_llm_delays) component_delays: { @@ -140,27 +140,27 @@ def extract_delay_breakdown(traces: List[Dict]) -> Tuple[Dict[str, List[float]], 'agent_processing_delay': [], 'e2e_delay': [] } - + # Build span tree span_dict = {span['span_id']: span for span in traces} root_spans = [span for span in traces if span.get('parent_span_id') is None] - + per_agent_delays = defaultdict(list) # {agent_name: [durations]} inter_agent_delays = defaultdict(list) # {(source, target): [durations]} agent_llm_delays = defaultdict(list) # {agent_name: [durations]} agent_llm_delays_detail = defaultdict(list) # {agent_name: [durations]} for detailed breakdown - + # Agent-LLM delay: LLM call spans with agent attribution # Support multiple span name patterns: call_llm, openai.chat, generate, etc. for span in traces: span_name = _safe_lower(span.get('name')) attrs = span.get('attributes', {}) - + # Check if this is an LLM call span # Priority 1: Use gen_ai.operation.name if available (per template) operation_name = _safe_lower(attrs.get('gen_ai.operation.name')) is_llm_span = operation_name == 'call_llm' - + # Priority 2: Fallback to span name pattern matching (for compatibility) if not is_llm_span: is_llm_span = ( @@ -170,17 +170,17 @@ def extract_delay_breakdown(traces: List[Dict]) -> Tuple[Dict[str, List[float]], attrs.get('gen_ai.usage.input_tokens') or attrs.get('gen_ai.usage.output_tokens') ) - + if is_llm_span: duration_ns = _safe_int(span.get('duration_ns')) duration_ms = duration_ns / 1e6 delays['agent_llm_delay'].append(duration_ms) - + # Find agent name from span attributes or parent span agent_name = 'unknown' # First try to get from current span attributes agent_name = attrs.get('gen_ai.agent.name') or span.get('agent_name', 'unknown') - + # If not found, try parent span if agent_name == 'unknown': parent_id = span.get('parent_span_id') @@ -188,14 +188,14 @@ def extract_delay_breakdown(traces: List[Dict]) -> Tuple[Dict[str, List[float]], parent_span = span_dict[parent_id] parent_name = _safe_lower(parent_span.get('name')) # Support multiple parent span patterns - if ('invoke_agent' in parent_name or + if ('invoke_agent' in parent_name or 'agent' in parent_name or 'augmentedllm' in parent_name): parent_attrs = parent_span.get('attributes', {}) agent_name = parent_attrs.get('gen_ai.agent.name') or parent_span.get('agent_name', 'unknown') agent_llm_delays[agent_name].append(duration_ms) agent_llm_delays_detail[agent_name].append(duration_ms) - + # Inter-agent delay: execute_tool spans with source/target (AgentTool) or invoke_agent spans with parent invoke_agent (SequentialAgent/LoopAgent) for span in traces: span_name = _safe_lower(span.get('name')) @@ -204,20 +204,20 @@ def extract_delay_breakdown(traces: List[Dict]) -> Tuple[Dict[str, List[float]], is_agent_comm = False source_agent = 'unknown' target_agent = 'unknown' - + # Method 1: Check execute_tool spans (for AgentTool in marketing-agency or send_message in distributed systems) if 'execute_tool' in span_name: tool_type = attrs.get('gen_ai.tool.type', '') - + # Check if it's agent communication: either marked as such, or is AgentTool type is_agent_comm = comm.get('is_agent_communication', False) or tool_type == 'AgentTool' - + # Also check if it's a send_message tool (common in distributed systems) if not is_agent_comm: tool_name = _safe_lower(attrs.get('gen_ai.tool.name')) if 'message' in tool_name or 'agent' in tool_name: is_agent_comm = True - + if is_agent_comm: # Source: traverse up to find invoke_agent span current_id = span.get('parent_span_id') @@ -233,10 +233,10 @@ def extract_delay_breakdown(traces: List[Dict]) -> Tuple[Dict[str, List[float]], depth += 1 else: break - + # Target: from span name or attributes target_agent = attrs.get('gen_ai.agent.name') or attrs.get('adk.agent.name') or span.get('agent_name', 'unknown') - + # For send_message tool in distributed systems, extract agent_name from tool_call_args if target_agent == 'unknown' and 'send_message' in span_name: tool_call_args = attrs.get('gcp.vertex.agent.tool_call_args', '') @@ -248,13 +248,13 @@ def extract_delay_breakdown(traces: List[Dict]) -> Tuple[Dict[str, List[float]], target_agent = args['agent_name'] except: pass - + # Fallback: try to extract from span name if not found in attributes if target_agent == 'unknown': match = re.search(r'(?:execute_tool|invoke_agent)\s+(\w+)', span.get('name', ''), re.IGNORECASE) if match: target_agent = match.group(1) - + # Method 2: Check invoke_agent spans with parent invoke_agent (for SequentialAgent/LoopAgent in image-scoring) # NOTE: For monolithic systems with SequentialAgent/LoopAgent, invoke_agent spans represent agent-to-agent calls. # The duration includes the entire sub-agent execution (LLM calls, tool calls, etc.), similar to AgentTool in marketing-agency. @@ -279,50 +279,50 @@ def extract_delay_breakdown(traces: List[Dict]) -> Tuple[Dict[str, List[float]], parts = span.get('name', '').split() if len(parts) > 1: target_agent = parts[-1] - + if is_agent_comm and source_agent != 'unknown' and target_agent != 'unknown': # For inter-agent communication, we need to distinguish between: # 1. Communication overhead (actual time to pass data between agents) - should be < 1ms for in-process calls # 2. Total sub-agent execution time (including LLM calls) - this is what the span duration represents - + # For in-process calls (monolithic), the communication overhead is minimal (< 1ms) # The span duration includes the entire sub-agent execution, which is not just communication overhead # However, for the purpose of "inter-agent delay", we should use a small fixed value for in-process calls # or calculate the actual communication overhead by subtracting child spans - + span_id = span.get('span_id') duration_ns = _safe_int(span.get('duration_ns')) - + # Calculate communication overhead: duration minus child spans (LLM calls, tool calls, etc.) # This gives us the actual communication overhead + agent processing time child_duration = 0 for child_span in traces: if child_span.get('parent_span_id') == span_id: child_duration += _safe_int(child_span.get('duration_ns')) - + # Communication overhead = total duration - child spans duration # For in-process calls, this should be very small (< 1ms) # For distributed calls, this includes network latency communication_overhead_ns = max(0, duration_ns - child_duration) communication_overhead_ms = communication_overhead_ns / 1e6 - + # Use communication overhead as inter-agent delay (not the full duration) # This represents the actual time spent in communication, not the sub-agent execution time if communication_overhead_ms > 0: delays['inter_agent_delay'].append(communication_overhead_ms) inter_agent_delays[(source_agent, target_agent)].append(communication_overhead_ms) - + # Agent processing delay: invoke_agent spans minus child spans # Also support other agent span patterns (e.g., OpenAIAugmentedLLM.*.generate spans, tot.* spans) for span in traces: span_name = _safe_lower(span.get('name')) attrs = span.get('attributes', {}) - + # Check if this is an agent span # Priority 1: Check gen_ai.operation.name if available operation_name = _safe_lower(attrs.get('gen_ai.operation.name')) is_agent_span = operation_name == 'invoke_agent' - + # Priority 2: Check span name patterns if not is_agent_span: is_agent_span = ( @@ -330,7 +330,7 @@ def extract_delay_breakdown(traces: List[Dict]) -> Tuple[Dict[str, List[float]], ('augmentedllm' in span_name and 'generate' in span_name) or (attrs.get('gen_ai.agent.name') and 'generate' in span_name) ) - + # Priority 3: Check if span has gen_ai.agent.name and has child spans (for tot.*, etc.) # This handles cases where agent spans don't follow standard naming patterns if not is_agent_span: @@ -341,22 +341,22 @@ def extract_delay_breakdown(traces: List[Dict]) -> Tuple[Dict[str, List[float]], has_children = any(s.get('parent_span_id') == span_id for s in traces) if has_children: is_agent_span = True - + if is_agent_span: span_id = span.get('span_id') duration_ns = _safe_int(span.get('duration_ns')) - + # Find child spans child_duration = 0 for child_span in traces: if child_span.get('parent_span_id') == span_id: child_duration += _safe_int(child_span.get('duration_ns')) - + # Processing delay = total duration - child spans duration processing_delay = max(0, duration_ns - child_duration) / 1e6 # Convert to ms if processing_delay > 0: delays['agent_processing_delay'].append(processing_delay) - + # Attribute to agent agent_name = attrs.get('gen_ai.agent.name') or span.get('agent_name', 'unknown') if agent_name == 'unknown': @@ -364,17 +364,17 @@ def extract_delay_breakdown(traces: List[Dict]) -> Tuple[Dict[str, List[float]], parts = span.get('name', '').split('.') if len(parts) >= 2: agent_name = parts[1] # Extract agent name from span name - + if agent_name != 'unknown': per_agent_delays[agent_name].append(processing_delay) - + # E2E delay: sum of all invocation spans or top-level spans (total E2E delay for all requests) # Note: In distributed systems, invocation spans may have parent_span_id (they're not root spans) # So we need to find all invocation spans, not just root spans # Also support other top-level span patterns if invocation spans don't exist - invocation_spans = [span for span in traces if span.get('name') == 'invocation' or + invocation_spans = [span for span in traces if span.get('name') == 'invocation' or (_safe_lower(span.get('name')) == 'invocation')] - + if not invocation_spans: # Fallback 1: use root spans (spans with no parent) as E2E delay root_spans = [span for span in traces if not span.get('parent_span_id')] @@ -384,8 +384,8 @@ def extract_delay_breakdown(traces: List[Dict]) -> Tuple[Dict[str, List[float]], # Fallback 2: use top-level spans (spans whose parent is not in traces) # This handles cases where parent spans are in different trace files span_dict_for_lookup = {span['span_id']: span for span in traces} - top_level_spans = [span for span in traces - if span.get('parent_span_id') and + top_level_spans = [span for span in traces + if span.get('parent_span_id') and span.get('parent_span_id') not in span_dict_for_lookup] if top_level_spans: # Group by trace_id to get one span per trace/request @@ -398,18 +398,18 @@ def extract_delay_breakdown(traces: List[Dict]) -> Tuple[Dict[str, List[float]], elif _safe_int(span.get('duration_ns')) > _safe_int(trace_groups[trace_id].get('duration_ns')): trace_groups[trace_id] = span invocation_spans = list(trace_groups.values()) - + if invocation_spans: # Calculate total E2E delay as sum of all individual request delays total_e2e_delay = sum(_safe_int(span.get('duration_ns')) for span in invocation_spans) / 1e6 delays['e2e_delay'].append(total_e2e_delay) # Single value: total of all requests - + return delays, dict(per_agent_delays), dict(inter_agent_delays), dict(agent_llm_delays_detail) def extract_message_sizes(traces: List[Dict]) -> Tuple[Dict[str, List[float]], Dict[str, Dict[str, List[float]]], Dict[str, Dict[str, List[float]]], Dict[str, Dict[str, List[float]]]]: """Extract message sizes from traces. - + Returns: (component_sizes, per_agent_sizes, inter_agent_sizes, agent_llm_sizes) component_sizes: { @@ -428,35 +428,35 @@ def extract_message_sizes(traces: List[Dict]) -> Tuple[Dict[str, List[float]], D 'agent_llm_input': [], 'agent_llm_output': [] } - + span_dict = {span['span_id']: span for span in traces} per_agent_sizes = defaultdict(lambda: {'input': [], 'output': []}) inter_agent_sizes = defaultdict(lambda: {'input': [], 'output': []}) agent_llm_sizes = defaultdict(lambda: {'input': [], 'output': []}) - + for span in traces: comm = span.get('communication', {}) input_size = _safe_int(comm.get('input_message_size_bytes')) output_size = _safe_int(comm.get('output_message_size_bytes')) - + if input_size > 0 or output_size > 0: span_name = _safe_lower(span.get('name')) attrs = span.get('attributes', {}) is_agent_comm = False source_agent = 'unknown' target_agent = 'unknown' - + # Method 1: Check execute_tool spans (for AgentTool in marketing-agency or send_message in distributed systems) if 'execute_tool' in span_name: tool_type = attrs.get('gen_ai.tool.type', '') is_agent_comm = comm.get('is_agent_communication', False) or tool_type == 'AgentTool' - + # Also check if it's a send_message tool (common in distributed systems) if not is_agent_comm: tool_name = _safe_lower(attrs.get('gen_ai.tool.name')) if 'message' in tool_name or 'agent' in tool_name: is_agent_comm = True - + if is_agent_comm: # Find source agent by traversing up to find invoke_agent span current_id = span.get('parent_span_id') @@ -472,10 +472,10 @@ def extract_message_sizes(traces: List[Dict]) -> Tuple[Dict[str, List[float]], D depth += 1 else: break - + # Find target agent from span name or attributes target_agent = attrs.get('gen_ai.agent.name') or attrs.get('adk.agent.name') or span.get('agent_name', 'unknown') - + # For send_message tool in distributed systems, extract agent_name from tool_call_args if target_agent == 'unknown' and 'send_message' in span_name: tool_call_args = attrs.get('gcp.vertex.agent.tool_call_args', '') @@ -487,13 +487,13 @@ def extract_message_sizes(traces: List[Dict]) -> Tuple[Dict[str, List[float]], D target_agent = args['agent_name'] except: pass - + # Fallback: try to extract from span name if not found in attributes if target_agent == 'unknown': match = re.search(r'execute_tool\s+(\w+)', span.get('name', ''), re.IGNORECASE) if match: target_agent = match.group(1) - + # Method 2: Check invoke_agent spans with parent invoke_agent (for SequentialAgent/LoopAgent in image-scoring) elif 'invoke_agent' in span_name: # Check if parent span is also an invoke_agent (indicating agent-to-agent call) @@ -514,25 +514,25 @@ def extract_message_sizes(traces: List[Dict]) -> Tuple[Dict[str, List[float]], D parts = span.get('name', '').split() if len(parts) > 1: target_agent = parts[-1] - + # For invoke_agent spans, try to get message size from communication metrics # Similar to execute_tool spans, but for invoke_agent spans in monolithic systems comm = span.get('communication', {}) input_size = _safe_int(comm.get('input_message_size_bytes')) output_size = _safe_int(comm.get('output_message_size_bytes')) - + # If not found in communication metrics, try to get from attributes (same as execute_tool) if input_size == 0 and output_size == 0: tool_call_args = attrs.get('gcp.vertex.agent.tool_call_args', '') if tool_call_args and isinstance(tool_call_args, str) and tool_call_args != "{}": input_size = len(tool_call_args.encode('utf-8')) - + tool_response = attrs.get('gcp.vertex.agent.tool_response', '') if tool_response and isinstance(tool_response, str) and tool_response != "{}": output_size = len(tool_response.encode('utf-8')) input_size = _safe_int(input_size) output_size = _safe_int(output_size) - + # For in-process calls (SequentialAgent/LoopAgent), message size is typically 0 # because data is passed as Python objects, not serialized messages # We still record the communication for completeness, even if size is 0 @@ -540,25 +540,25 @@ def extract_message_sizes(traces: List[Dict]) -> Tuple[Dict[str, List[float]], D # This is an in-process call with no serialized message data # We still want to track it as inter-agent communication pass - + if is_agent_comm: # Inter-agent communication input_size = _safe_int(input_size) output_size = _safe_int(output_size) input_kb = input_size / 1024 output_kb = output_size / 1024 - + if input_size > 0: sizes['inter_agent_input'].append(input_kb) if output_size > 0: sizes['inter_agent_output'].append(output_kb) - + if source_agent != 'unknown': if input_size > 0: per_agent_sizes[source_agent]['output'].append(input_kb) if output_size > 0: per_agent_sizes[target_agent]['input'].append(output_kb) - + if source_agent != 'unknown' and target_agent != 'unknown': # Record inter-agent communication even if message size is 0 # For in-process calls, this represents the communication relationship @@ -567,19 +567,19 @@ def extract_message_sizes(traces: List[Dict]) -> Tuple[Dict[str, List[float]], D elif comm.get('is_in_process_call', False): # For in-process calls with no serialized message, record 0 inter_agent_sizes[(source_agent, target_agent)]['input'].append(0.0) - + if output_size > 0: inter_agent_sizes[(source_agent, target_agent)]['output'].append(output_kb) elif comm.get('is_in_process_call', False): # For in-process calls with no serialized message, record 0 inter_agent_sizes[(source_agent, target_agent)]['output'].append(0.0) - + else: # Check if this is an LLM call span # Priority 1: Use gen_ai.operation.name if available (per template) operation_name = _safe_lower(attrs.get('gen_ai.operation.name')) is_llm_span = operation_name == 'call_llm' - + # Priority 2: Fallback to span name pattern matching (for compatibility) if not is_llm_span: span_name = _safe_lower(span.get('name')) @@ -590,24 +590,24 @@ def extract_message_sizes(traces: List[Dict]) -> Tuple[Dict[str, List[float]], D attrs.get('gen_ai.usage.input_tokens') or attrs.get('gen_ai.usage.output_tokens') ) - + if is_llm_span: # Agent-LLM communication input_size = _safe_int(input_size) output_size = _safe_int(output_size) input_kb = input_size / 1024 output_kb = output_size / 1024 - + if input_size > 0: sizes['agent_llm_input'].append(input_kb) if output_size > 0: sizes['agent_llm_output'].append(output_kb) - + # Find agent name from span attributes or parent spans (search up the tree) agent_name = 'unknown' # First try current span agent_name = attrs.get('gen_ai.agent.name') or span.get('agent_name', 'unknown') - + # If not found, search up parent chain if agent_name == 'unknown': current = span @@ -616,34 +616,34 @@ def extract_message_sizes(traces: List[Dict]) -> Tuple[Dict[str, List[float]], D parent_id = current.get('parent_span_id') if not parent_id or parent_id not in span_dict: break - + parent_span = span_dict[parent_id] parent_name = _safe_lower(parent_span.get('name')) parent_attrs = parent_span.get('attributes', {}) - + # Check if parent has agent name parent_agent_name = parent_attrs.get('gen_ai.agent.name') or parent_span.get('agent_name') if parent_agent_name: agent_name = parent_agent_name break - + # Check if parent is an agent span - if ('invoke_agent' in parent_name or + if ('invoke_agent' in parent_name or 'agent' in parent_name or 'augmentedllm' in parent_name): # Try to extract from parent name or attributes if parent_agent_name: agent_name = parent_agent_name break - + current = parent_span - + if agent_name != 'unknown': if input_size > 0: agent_llm_sizes[agent_name]['input'].append(input_kb) if output_size > 0: agent_llm_sizes[agent_name]['output'].append(output_kb) - + return sizes, dict(per_agent_sizes), dict(inter_agent_sizes), dict(agent_llm_sizes) @@ -730,7 +730,7 @@ def extract_call_graph(traces: List[Dict]) -> set: """ edges = set() span_dict = {span['span_id']: span for span in traces} - + # Method 0: Extract call graph based on agent name hierarchy (most reliable for all systems) # This method works even when gen_ai.tool.type or communication.is_agent_communication are missing for span in traces: @@ -743,14 +743,14 @@ def extract_call_graph(traces: List[Dict]) -> set: candidate = span.get('agent_name') if candidate and _is_real_agent(candidate): current_agent = candidate - + if current_agent and _is_real_agent(current_agent): # Find parent agent by traversing up the span tree # Traverse until we find a span with a real agent name parent_id = span.get('parent_span_id') depth = 0 parent_agent_found = None - + while parent_id and depth < 15: # Increased depth limit if parent_id in span_dict: parent = span_dict[parent_id] @@ -762,7 +762,7 @@ def extract_call_graph(traces: List[Dict]) -> set: candidate = parent.get('agent_name') if candidate and _is_real_agent(candidate): parent_agent = candidate - + # If we found a real parent agent, check if it's different from current if parent_agent and _is_real_agent(parent_agent): if parent_agent != current_agent: @@ -777,14 +777,14 @@ def extract_call_graph(traces: List[Dict]) -> set: depth += 1 else: break - + # If we found a valid parent agent, add the edge # But only if both source and target are real agents (not system-level) if parent_agent_found: # Normalize agent names source_agent = parent_agent_found.replace('_agent', '').replace('-agent', '').replace('_', '-').replace(' ', '-').lower() target_agent = current_agent.replace('_agent', '').replace('-agent', '').replace('_', '-').replace(' ', '-').lower() - + # Double-check both are real agents (not system/service names) # Skip system-level sources (like 'autogen-multi-system') if 'system' in source_agent or 'service' in source_agent: @@ -793,39 +793,39 @@ def extract_call_graph(traces: List[Dict]) -> set: # Skip self-loops elif source_agent != target_agent: edges.add((source_agent, target_agent)) - + # Continue with existing methods for additional edges (execute_tool, invoke_agent) for span in traces: span_name = _safe_lower(span.get('name')) comm = span.get('communication', {}) attrs = span.get('attributes', {}) - + # Method 1: Check execute_tool spans (for monolithic systems with AgentTool) if 'execute_tool' in span_name: tool_type = attrs.get('gen_ai.tool.type', '') tool_name = _safe_lower(attrs.get('gen_ai.tool.name')) - + # Explicitly exclude FunctionTool - these are tools, not agents # Exception: send_message in distributed systems (even if marked as FunctionTool) if tool_type == 'FunctionTool': # Only allow send_message tool for agent communication in distributed systems if 'send_message' not in tool_name: continue - + # Check if it's agent communication: either marked as such, or is AgentTool type is_agent_comm = comm.get('is_agent_communication', False) or tool_type == 'AgentTool' - + # Also check if it's a send_message tool (common in distributed systems) # Even if tool_type is FunctionTool, send_message is used for agent communication if not is_agent_comm: if 'message' in tool_name or 'agent' in tool_name: is_agent_comm = True - + # For send_message tools in distributed systems, always treat as agent communication # even if tool_type is FunctionTool if tool_type == 'FunctionTool' and 'send_message' in tool_name: is_agent_comm = True - + # Fallback: If tool_type is missing, check if we can find agent-to-agent relationship # by checking parent-child agent names if not is_agent_comm and not tool_type: @@ -843,7 +843,7 @@ def extract_call_graph(traces: List[Dict]) -> set: break current_id = parent_span.get('parent_span_id') depth += 1 - + # Try to find target agent from current span or child spans target_agent = attrs.get('gen_ai.agent.name') or span.get('agent_name') if not target_agent: @@ -856,20 +856,20 @@ def extract_call_graph(traces: List[Dict]) -> set: if child_agent: target_agent = child_agent break - + # If we found both source and target agents, and they're different, it's agent communication if source_agent and target_agent and source_agent != target_agent: is_agent_comm = True - + # Final check: if still not agent communication, skip if not is_agent_comm: continue - + if is_agent_comm: # Find source and target agents source_agent = 'unknown' target_agent = 'unknown' - + # Find source agent by traversing up to find invoke_agent span # This is the most reliable method for monolithic systems (marketing-agency) # Continue traversing even if we find call_llm spans in between @@ -890,7 +890,7 @@ def extract_call_graph(traces: List[Dict]) -> set: depth += 1 else: break - + # For send_message tools in distributed systems, use service.name as source agent # This is more reliable than traversing parent spans for distributed systems # Only use service.name if we couldn't find invoke_agent in parent chain @@ -899,11 +899,11 @@ def extract_call_graph(traces: List[Dict]) -> set: service_name = resource_attrs.get('service.name', '') if service_name and service_name not in ['unknown', '']: source_agent = service_name.replace('_agent', '').replace('-agent', '').replace('_', '-').replace(' ', '-').lower() - + # Find target agent - for AgentTool, extract from span name (e.g., "execute_tool domain_create_agent") # For send_message, extract from tool_call_args span_name_full = span.get('name', '') - + # For send_message tool in distributed systems, extract agent_name from tool_call_args if 'send_message' in span_name or 'send_message' in tool_name: tool_call_args = attrs.get('gcp.vertex.agent.tool_call_args', '') @@ -917,7 +917,7 @@ def extract_call_graph(traces: List[Dict]) -> set: target_agent = target_agent.replace(' Agent', '').replace('_', '-').replace(' ', '-').lower() except: pass - + # For AgentTool in monolithic systems, extract from span name if target_agent == 'unknown': # Format: "execute_tool " @@ -933,11 +933,11 @@ def extract_call_graph(traces: List[Dict]) -> set: target_agent = attrs.get('gen_ai.agent.name') or attrs.get('adk.agent.name') or span.get('agent_name', 'unknown') if target_agent != 'unknown': target_agent = target_agent.lower() - + # Normalize source agent name for consistency if source_agent != 'unknown': source_agent = source_agent.replace('_agent', '').replace('-agent', '').replace('_', '-').replace(' ', '-').lower() - + if source_agent != 'unknown' and target_agent != 'unknown' and source_agent != target_agent: # In distributed systems, 'routing' is typically an internal agent of 'coordinator' # Replace 'routing' with 'coordinator' to avoid duplicate nodes @@ -945,40 +945,40 @@ def extract_call_graph(traces: List[Dict]) -> set: target_agent = 'coordinator' if source_agent == 'routing': source_agent = 'coordinator' - + # Skip self-loops after normalization if source_agent == target_agent: continue - + edges.add((source_agent, target_agent)) - + # Method 2: Check invoke_agent spans # Two cases: # 1. Distributed systems: invoke_agent spans in remote agent traces (handled below) # 2. Monolithic systems with SequentialAgent/LoopAgent: invoke_agent spans with parent invoke_agent (handled here) elif 'invoke_agent' in span_name: target_agent = attrs.get('gen_ai.agent.name') or span.get('agent_name', 'unknown') - + # Try to extract from span name if not in attributes if target_agent == 'unknown': # Format: "invoke_agent " parts = span.get('name', '').split() if len(parts) > 1: target_agent = parts[-1] - + # Filter out internal operations if not _is_real_agent(target_agent): continue - + # Normalize target agent name normalized_target = target_agent.replace('_agent', '').replace('-agent', '').replace('_', '-').lower() - + # Find source agent from parent span source_agent = 'unknown' resource_attrs = span.get('resource', {}).get('attributes', {}) service_name = resource_attrs.get('service.name', '') normalized_service = service_name.replace('_agent', '').replace('-agent', '').replace('_', '-').lower() if service_name else '' - + # Check if parent span is also an invoke_agent (indicating agent-to-agent call in monolithic system) parent_id = span.get('parent_span_id') if parent_id and parent_id in span_dict: @@ -990,11 +990,11 @@ def extract_call_graph(traces: List[Dict]) -> set: parent_agent = parent_attrs.get('gen_ai.agent.name') or parent_span.get('agent_name', 'unknown') if parent_agent != 'unknown': source_agent = parent_agent.replace('_agent', '').replace('-agent', '').replace('_', '-').lower() - + # If we didn't find a parent invoke_agent, this might be: # 1. Entry point in monolithic system (marketing-agency) - skip it # 2. Remote agent invocation in distributed system - handle below - + # Skip if this is likely an entry point in monolithic system if source_agent == 'unknown': # If service.name and target agent name refer to the same agent, this is likely an entry point @@ -1007,13 +1007,13 @@ def extract_call_graph(traces: List[Dict]) -> set: # Only use service.name if it represents a real agent, not a system if service_name and service_name not in ['unknown', ''] and _is_real_agent(normalized_service): source_agent = normalized_service - + # Normalize both agent names for consistency if source_agent != 'unknown': source_agent = source_agent.replace('_agent', '').replace('-agent', '').replace('_', '-').replace(' ', '-').lower() if target_agent != 'unknown': target_agent = target_agent.replace('_agent', '').replace('-agent', '').replace('_', '-').replace(' ', '-').lower() - + # Only add edge if both agents are known and different if source_agent != 'unknown' and target_agent != 'unknown' and source_agent != target_agent: # In distributed systems, 'routing' is typically an internal agent of 'coordinator' @@ -1022,30 +1022,30 @@ def extract_call_graph(traces: List[Dict]) -> set: target_agent = 'coordinator' if source_agent == 'routing': source_agent = 'coordinator' - + # Skip self-loops after normalization if source_agent == target_agent: continue - + edges.add((source_agent, target_agent)) - + # If no edges were found but we have agents, we might have a system where agents don't call each other directly # (e.g., all agents are called by a coordinator). In this case, we still want to show the agents as nodes. # However, for call graph visualization, edges are required. So we'll return empty edges # and the visualization code can handle showing just nodes if needed. # For now, we return edges as-is. If edges is empty, it means no agent-to-agent calls were found, # which is valid for some system architectures (e.g., all agents called by a coordinator). - + return edges def extract_call_sequence(traces: List[Dict]) -> List[tuple]: """Extract call sequence (ordered list of agent-to-agent calls) from traces. - + Returns the sequence of (source_agent, target_agent) tuples in chronological order. This preserves the calling order, unlike extract_call_graph which returns a set. Supports both monolithic systems (AgentTool via execute_tool) and distributed systems (invoke_agent spans). - + Returns: List of tuples (source_agent, target_agent) in chronological order """ @@ -1053,10 +1053,10 @@ def extract_call_sequence(traces: List[Dict]) -> List[tuple]: span_dict = {span['span_id']: span for span in traces} filter_internal_agents = _should_filter_internal_agents(traces) used_sender_spans = set() - + # Collect all spans that represent agent communication agent_calls = [] - + if filter_internal_agents: for span in traces: attrs = span.get("attributes", {}) @@ -1085,7 +1085,7 @@ def extract_call_sequence(traces: List[Dict]) -> List[tuple]: continue attrs = span.get('attributes', {}) current_agent = attrs.get('gen_ai.agent.name') or span.get('agent_name') - + if current_agent: if filter_internal_agents and not _is_real_agent(current_agent): continue @@ -1107,52 +1107,52 @@ def extract_call_sequence(traces: List[Dict]) -> List[tuple]: # Normalize agent names source_agent = parent_agent.replace('_agent', '').replace('-agent', '').replace('_', '-').replace(' ', '-').lower() target_agent = current_agent.replace('_agent', '').replace('-agent', '').replace('_', '-').replace(' ', '-').lower() - + # Skip self-loops if source_agent != target_agent: # Store with timestamp for sorting start_time = _safe_int(span.get('start_time')) agent_calls.append((start_time, source_agent, target_agent)) break - + parent_id = parent.get('parent_span_id') depth += 1 else: break - + # Continue with existing methods for additional edges (execute_tool, invoke_agent) for span in traces: span_name = _safe_lower(span.get('name')) comm = span.get('communication', {}) attrs = span.get('attributes', {}) tool_name = _safe_lower(attrs.get('gen_ai.tool.name')) if attrs.get('gen_ai.tool.name') else '' - + # Method 1: Check execute_tool spans (for monolithic systems with AgentTool) if 'execute_tool' in span_name: if filter_internal_agents and span.get("span_id") in used_sender_spans: continue tool_type = attrs.get('gen_ai.tool.type', '') - + # Check if it's agent communication is_agent_comm = comm.get('is_agent_communication', False) or tool_type == 'AgentTool' - + if not is_agent_comm: if 'message' in tool_name or 'agent' in tool_name: is_agent_comm = True - + # For send_message tools in distributed systems, always treat as agent communication if tool_type == 'FunctionTool' and 'send_message' in tool_name: is_agent_comm = True - + # Skip non-agent FunctionTool calls (but allow send_message) if tool_type == 'FunctionTool' and not is_agent_comm: continue - + if is_agent_comm: # Find source and target agents (same logic as extract_call_graph) source_agent = 'unknown' target_agent = 'unknown' - + # For send_message tools in distributed systems, use service.name as source agent resource_attrs = span.get('resource', {}).get('attributes', {}) service_name = resource_attrs.get('service.name', '') @@ -1161,7 +1161,7 @@ def extract_call_sequence(traces: List[Dict]) -> List[tuple]: service_name = '' else: source_agent = service_name.replace('_agent', '').replace('-agent', '').replace('_', '-').replace(' ', '-').lower() - + # If service.name is not available, find source agent by traversing up to find invoke_agent span if source_agent == 'unknown': current_id = span.get('parent_span_id') @@ -1182,12 +1182,12 @@ def extract_call_sequence(traces: List[Dict]) -> List[tuple]: depth += 1 else: break - + # Find target agent target_agent = attrs.get('gen_ai.agent.name') or attrs.get('adk.agent.name') or span.get('agent_name', 'unknown') if filter_internal_agents and not _is_real_agent(target_agent): target_agent = 'unknown' - + if target_agent == 'unknown' and ('send_message' in span_name or 'send_message' in tool_name): tool_call_args = attrs.get('gcp.vertex.agent.tool_call_args', '') if tool_call_args: @@ -1200,30 +1200,30 @@ def extract_call_sequence(traces: List[Dict]) -> List[tuple]: target_agent = target_agent.replace(' Agent', '').replace('_', '-').lower() except: pass - + if target_agent == 'unknown' and span_name: match = re.search(r'execute_tool\s+(\w+)', span_name, re.IGNORECASE) if match: target_agent = match.group(1) target_agent = target_agent.replace('_agent', '').replace('-agent', '').replace('_', '-').lower() - + # Normalize source and target agent names for consistency if source_agent != 'unknown': source_agent = source_agent.replace('_agent', '').replace('-agent', '').replace('_', '-').lower() if target_agent != 'unknown': target_agent = target_agent.replace('_agent', '').replace('-agent', '').replace('_', '-').lower() - + if source_agent != 'unknown' and target_agent != 'unknown' and source_agent != target_agent: # Store with timestamp for sorting start_time = _safe_int(span.get('start_time')) agent_calls.append((start_time, source_agent, target_agent)) - + # Method 2: Check invoke_agent spans (for distributed systems or SequentialAgent/LoopAgent) elif 'invoke_agent' in span_name: if filter_internal_agents and span.get("span_id") in used_sender_spans: continue target_agent = attrs.get('gen_ai.agent.name') or span.get('agent_name', 'unknown') - + # Try to extract from span name if not in attributes if target_agent == 'unknown': parts = span.get('name', '').split() @@ -1231,7 +1231,7 @@ def extract_call_sequence(traces: List[Dict]) -> List[tuple]: target_agent = parts[-1] if filter_internal_agents and not _is_real_agent(target_agent): continue - + # Check if parent span is also an invoke_agent (indicating agent-to-agent call in monolithic system) source_agent = 'unknown' parent_id = span.get('parent_span_id') @@ -1247,44 +1247,44 @@ def extract_call_sequence(traces: List[Dict]) -> List[tuple]: parent_agent = 'unknown' if parent_agent != 'unknown': source_agent = parent_agent.replace('_agent', '').replace('-agent', '').replace('_', '-').lower() - + # If we didn't find a parent invoke_agent, this might be a distributed system call if source_agent == 'unknown': # Normalize target agent name normalized_target = target_agent.replace('_agent', '').replace('-agent', '').replace('_', '-').lower() - + # Find source agent from service.name resource_attrs = span.get('resource', {}).get('attributes', {}) service_name = resource_attrs.get('service.name', '') - + # Normalize service name normalized_service = service_name.replace('_agent', '').replace('-agent', '').replace('_', '-').lower() if service_name else '' - + # Skip if service.name and target agent name refer to the same agent (entry point) if normalized_service == normalized_target: continue - + # Skip marketing-agency entry point if normalized_service and normalized_service in ['marketing-agency', 'marketing-agency']: continue - + if service_name and service_name not in ['unknown', '']: if filter_internal_agents and not _is_real_agent(service_name): service_name = '' else: source_agent = normalized_service - + # Normalize both agent names for consistency if source_agent != 'unknown': source_agent = source_agent.replace('_agent', '').replace('-agent', '').replace('_', '-').replace(' ', '-').lower() if target_agent != 'unknown': target_agent = target_agent.replace('_agent', '').replace('-agent', '').replace('_', '-').replace(' ', '-').lower() - + if source_agent != 'unknown' and target_agent != 'unknown' and source_agent != target_agent: # Store with timestamp for sorting start_time = _safe_int(span.get('start_time')) agent_calls.append((start_time, source_agent, target_agent)) - + # Sort by timestamp and extract sequence agent_calls.sort(key=lambda x: x[0]) call_sequence = [ @@ -1292,16 +1292,16 @@ def extract_call_sequence(traces: List[Dict]) -> List[tuple]: for _, src, tgt in agent_calls if src != 'unknown' and tgt != 'unknown' ] - + return call_sequence def extract_cpu_memory_usage(metrics: List[Dict]) -> Dict: """Extract CPU and memory usage from metrics (process-level only). - + In distributed systems, each agent has its own metrics file, so we extract per-agent CPU/memory usage. In monolithic systems, all metrics are from one process. - + Returns: { 'cpu': { @@ -1318,20 +1318,20 @@ def extract_cpu_memory_usage(metrics: List[Dict]) -> Dict: 'cpu': {'process': [], 'per_agent': defaultdict(list)}, 'memory': {'process': [], 'per_agent': defaultdict(list)} } - + for metric in metrics: metric_name = metric.get('metric_name', '') data_points = metric.get('data_points', []) - + # Extract agent name from resource attributes resource = metric.get('resource', {}) resource_attrs = resource.get('attributes', {}) if resource else {} agent_name = resource_attrs.get('service.name', 'unknown') - + for dp in data_points: timestamp = dp.get('timestamp', 0) value = dp.get('value', 0) - + # Support both old and new metric name formats if metric_name == 'process.cpu.usage' or metric_name == 'process.cpu.utilization': usage['cpu']['process'].append((timestamp / 1e9, value)) # Convert to seconds @@ -1340,7 +1340,7 @@ def extract_cpu_memory_usage(metrics: List[Dict]) -> Dict: mem_mb = value / (1024 * 1024) # Convert to MB usage['memory']['process'].append((timestamp / 1e9, mem_mb)) usage['memory']['per_agent'][agent_name].append((timestamp / 1e9, mem_mb)) - + # Sort by timestamp for key in usage: for subkey in usage[key]: @@ -1349,9 +1349,9 @@ def extract_cpu_memory_usage(metrics: List[Dict]) -> Dict: elif isinstance(usage[key][subkey], dict): for agent_name in usage[key][subkey]: usage[key][subkey][agent_name].sort(key=lambda x: x[0]) - + # Convert defaultdict to dict for JSON serialization usage['cpu']['per_agent'] = dict(usage['cpu']['per_agent']) usage['memory']['per_agent'] = dict(usage['memory']['per_agent']) - + return usage diff --git a/plot/lib/per_run_analyzer.py b/plot/lib/per_run_analyzer.py index 6a3c19f..c514eee 100644 --- a/plot/lib/per_run_analyzer.py +++ b/plot/lib/per_run_analyzer.py @@ -19,11 +19,11 @@ def analyze_per_run(trace_files: List[str], metrics_files: Optional[List[str]] = None) -> Dict: """Analyze each run separately and compute statistics across runs. - + Args: trace_files: List of trace file paths metrics_files: Optional list of metrics file paths - + Returns: Dictionary containing per-run data and cross-run statistics """ @@ -31,12 +31,12 @@ def analyze_per_run(trace_files: List[str], metrics_files: Optional[List[str]] = 'runs': [], 'statistics': {} } - + # Analyze each run for i, trace_file in enumerate(trace_files): traces = load_traces(trace_file) metrics = load_metrics(metrics_files[i]) if metrics_files and i < len(metrics_files) else [] - + # Extract metrics for this run run_data = { 'run_id': i + 1, @@ -44,14 +44,14 @@ def analyze_per_run(trace_files: List[str], metrics_files: Optional[List[str]] = 'traces': traces, 'metrics': metrics, } - + # Extract token consumption total_tokens, per_agent_tokens = extract_token_consumption(traces) run_data['tokens'] = { 'total': total_tokens, 'per_agent': per_agent_tokens } - + # Extract delay breakdown delays, per_agent_delays, inter_agent_delays, agent_llm_delays = extract_delay_breakdown(traces) run_data['delays'] = { @@ -60,7 +60,7 @@ def analyze_per_run(trace_files: List[str], metrics_files: Optional[List[str]] = 'inter_agent': inter_agent_delays, 'agent_llm': agent_llm_delays } - + # Extract message sizes sizes, per_agent_sizes, inter_agent_sizes, agent_llm_sizes = extract_message_sizes(traces) run_data['message_sizes'] = { @@ -69,45 +69,45 @@ def analyze_per_run(trace_files: List[str], metrics_files: Optional[List[str]] = 'inter_agent': inter_agent_sizes, 'agent_llm': agent_llm_sizes } - + # Extract CPU/memory usage if metrics: usage = extract_cpu_memory_usage(metrics) run_data['resource_usage'] = usage else: run_data['resource_usage'] = None - + # Extract call graph edges (set, unordered) call_graph_edges = extract_call_graph(traces) run_data['call_graph'] = call_graph_edges - + # Extract call sequence (ordered list, preserves calling order) call_sequence = extract_call_sequence(traces) run_data['call_sequence'] = call_sequence - + results['runs'].append(run_data) - + # Compute statistics across runs results['statistics'] = compute_cross_run_statistics(results['runs']) - + return results def compute_cross_run_statistics(runs: List[Dict]) -> Dict: """Compute statistics across multiple runs. - + Returns: Dictionary with statistics for each metric type """ stats = {} - + # E2E Delay statistics e2e_delays = [] for run in runs: e2e_delay_list = run['delays']['component'].get('e2e_delay', []) if e2e_delay_list: e2e_delays.append(e2e_delay_list[0]) # Single value per run - + if e2e_delays: stats['e2e_delay'] = { 'values': e2e_delays, @@ -121,17 +121,17 @@ def compute_cross_run_statistics(runs: List[Dict]) -> Dict: 'cv': np.std(e2e_delays) / np.mean(e2e_delays) if np.mean(e2e_delays) > 0 else 0, 'n': len(e2e_delays) } - + # Token consumption statistics (total across all runs) total_prompt = sum(run['tokens']['total']['prompt'] for run in runs) total_completion = sum(run['tokens']['total']['completion'] for run in runs) total_tokens = sum(run['tokens']['total']['total'] for run in runs) - + # Per-run token totals run_token_totals = [run['tokens']['total']['total'] for run in runs] run_token_prompt = [run['tokens']['total']['prompt'] for run in runs] run_token_completion = [run['tokens']['total']['completion'] for run in runs] - + stats['tokens'] = { 'total_prompt': total_prompt, 'total_completion': total_completion, @@ -147,12 +147,12 @@ def compute_cross_run_statistics(runs: List[Dict]) -> Dict: 'std_completion': np.std(run_token_completion) if run_token_completion else 0, 'n': len(runs) } - + # Agent-LLM delay statistics (collect all delays across all runs) all_llm_delays = [] for run in runs: all_llm_delays.extend(run['delays']['component'].get('agent_llm_delay', [])) - + if all_llm_delays: stats['agent_llm_delay'] = { 'values': all_llm_delays, @@ -165,12 +165,12 @@ def compute_cross_run_statistics(runs: List[Dict]) -> Dict: 'p99': np.percentile(all_llm_delays, 99), 'n': len(all_llm_delays) } - + # Inter-agent delay statistics all_inter_delays = [] for run in runs: all_inter_delays.extend(run['delays']['component'].get('inter_agent_delay', [])) - + if all_inter_delays: stats['inter_agent_delay'] = { 'values': all_inter_delays, @@ -183,12 +183,12 @@ def compute_cross_run_statistics(runs: List[Dict]) -> Dict: 'p99': np.percentile(all_inter_delays, 99), 'n': len(all_inter_delays) } - + # Processing delay statistics all_proc_delays = [] for run in runs: all_proc_delays.extend(run['delays']['component'].get('agent_processing_delay', [])) - + if all_proc_delays: stats['agent_processing_delay'] = { 'values': all_proc_delays, @@ -201,7 +201,7 @@ def compute_cross_run_statistics(runs: List[Dict]) -> Dict: 'p99': np.percentile(all_proc_delays, 99), 'n': len(all_proc_delays) } - + # CPU and Memory usage statistics # For time-series data, we compute per-run statistics (mean, peak, min) first, # then analyze the distribution of these statistics across runs @@ -211,11 +211,11 @@ def compute_cross_run_statistics(runs: List[Dict]) -> Dict: memory_per_run_mean = [] memory_per_run_peak = [] memory_per_run_min = [] - + for run in runs: if run.get('resource_usage'): usage = run['resource_usage'] - + # CPU statistics per run cpu_data = usage.get('cpu', {}).get('process', []) if cpu_data: @@ -224,7 +224,7 @@ def compute_cross_run_statistics(runs: List[Dict]) -> Dict: cpu_per_run_mean.append(np.mean(cpu_values)) cpu_per_run_peak.append(np.max(cpu_values)) cpu_per_run_min.append(np.min(cpu_values)) - + # Memory statistics per run memory_data = usage.get('memory', {}).get('process', []) if memory_data: @@ -233,7 +233,7 @@ def compute_cross_run_statistics(runs: List[Dict]) -> Dict: memory_per_run_mean.append(np.mean(memory_values)) memory_per_run_peak.append(np.max(memory_values)) memory_per_run_min.append(np.min(memory_values)) - + if cpu_per_run_mean: stats['cpu_usage'] = { 'mean': { @@ -270,7 +270,7 @@ def compute_cross_run_statistics(runs: List[Dict]) -> Dict: 'n': len(cpu_per_run_min) } } - + if memory_per_run_mean: stats['memory_usage'] = { 'mean': { @@ -307,51 +307,51 @@ def compute_cross_run_statistics(runs: List[Dict]) -> Dict: 'n': len(memory_per_run_min) } } - + # Message size statistics # For each run, compute total message sizes, then analyze across runs inter_agent_input_totals = [] inter_agent_output_totals = [] agent_llm_input_totals = [] agent_llm_output_totals = [] - + # Per-agent message size totals per run agent_message_totals_per_run = defaultdict(list) # {agent_name: [total_size_per_run]} - + # Per agent-pair message size totals per run pair_message_totals_per_run = defaultdict(list) # {(src, tgt): [total_size_per_run]} - + # Per agent-LLM message size totals per run agent_llm_totals_per_run = defaultdict(list) # {agent_name: [total_size_per_run]} - + for run in runs: msg_sizes = run['message_sizes'] - + # Component totals comp = msg_sizes['component'] inter_agent_input_totals.append(sum(comp.get('inter_agent_input', []))) inter_agent_output_totals.append(sum(comp.get('inter_agent_output', []))) agent_llm_input_totals.append(sum(comp.get('agent_llm_input', []))) agent_llm_output_totals.append(sum(comp.get('agent_llm_output', []))) - + # Per-agent totals for agent_name, agent_data in msg_sizes['per_agent'].items(): total = sum(agent_data.get('input', [])) + sum(agent_data.get('output', [])) if total > 0: agent_message_totals_per_run[agent_name].append(total) - + # Per agent-pair totals for (src, tgt), pair_data in msg_sizes['inter_agent'].items(): total = sum(pair_data.get('input', [])) + sum(pair_data.get('output', [])) if total > 0: pair_message_totals_per_run[(src, tgt)].append(total) - + # Per agent-LLM totals for agent_name, llm_data in msg_sizes['agent_llm'].items(): total = sum(llm_data.get('input', [])) + sum(llm_data.get('output', [])) if total > 0: agent_llm_totals_per_run[agent_name].append(total) - + if inter_agent_input_totals or inter_agent_output_totals or agent_llm_input_totals or agent_llm_output_totals: stats['message_sizes'] = { 'inter_agent_input': { @@ -398,11 +398,11 @@ def compute_cross_run_statistics(runs: List[Dict]) -> Dict: 'per_pair_totals': dict(pair_message_totals_per_run), 'per_agent_llm_totals': dict(agent_llm_totals_per_run) } - + # Call Graph Similarity Statistics (Multi-Graph Metrics) if len(runs) > 1: call_graphs = [run['call_graph'] for run in runs] - + # 1. Consecutive pairwise similarities (for backward compatibility) consecutive_similarities = [] for i in range(len(call_graphs) - 1): @@ -415,7 +415,7 @@ def compute_cross_run_statistics(runs: List[Dict]) -> Dict: consecutive_similarities.append(similarity) else: consecutive_similarities.append(0.0) - + # 2. All pairwise similarities (for comprehensive analysis) all_pairwise_similarities = [] similarity_matrix = [] @@ -437,21 +437,21 @@ def compute_cross_run_statistics(runs: List[Dict]) -> Dict: if i < j: # Only store upper triangle to avoid duplicates all_pairwise_similarities.append(similarity) similarity_matrix.append(row) - + # 3. Overall consistency: intersection of all graphs / union of all graphs all_edges_intersection = set.intersection(*call_graphs) if call_graphs else set() all_edges_union = set.union(*call_graphs) if call_graphs else set() overall_consistency = len(all_edges_intersection) / len(all_edges_union) if all_edges_union else 0.0 - + # 4. Core edge ratio: edges that appear in all runs / total unique edges core_edge_ratio = len(all_edges_intersection) / len(all_edges_union) if all_edges_union else 0.0 - + # 5. Stability index: standard deviation of pairwise similarities (lower = more stable) stability_index = np.std(all_pairwise_similarities) if all_pairwise_similarities else 0.0 - + # 6. Mean pairwise similarity mean_pairwise_similarity = np.mean(all_pairwise_similarities) if all_pairwise_similarities else 0.0 - + stats['call_graph_similarity'] = { # Consecutive pairwise (for backward compatibility) 'jaccard_similarities': consecutive_similarities, @@ -469,17 +469,17 @@ def compute_cross_run_statistics(runs: List[Dict]) -> Dict: 'similarity_matrix': similarity_matrix, # Full similarity matrix 'n_runs': len(call_graphs) } - + # Call Sequence Similarity Statistics (LCS - Sequence Order) if len(runs) > 1: call_sequences = [run.get('call_sequence', []) for run in runs] - + # LCS similarity between consecutive runs lcs_similarities = [] for i in range(len(call_sequences) - 1): seq1 = call_sequences[i] seq2 = call_sequences[i + 1] - + if not seq1 and not seq2: lcs_similarities.append(1.0) # Both empty, consider identical elif not seq1 or not seq2: @@ -496,12 +496,12 @@ def lcs_length(s1, s2): else: dp[i][j] = max(dp[i-1][j], dp[i][j-1]) return dp[m][n] - + lcs_len = lcs_length(seq1, seq2) max_len = max(len(seq1), len(seq2)) similarity = lcs_len / max_len if max_len > 0 else 0.0 lcs_similarities.append(similarity) - + # All pairwise LCS similarities all_pairwise_lcs = [] lcs_matrix = [] @@ -535,7 +535,7 @@ def lcs_length(s1, s2): if i < j: all_pairwise_lcs.append(similarity) lcs_matrix.append(row) - + stats['call_sequence_similarity'] = { 'lcs_similarities': lcs_similarities, 'lcs_mean': np.mean(lcs_similarities) if lcs_similarities else 0.0, @@ -547,7 +547,7 @@ def lcs_length(s1, s2): 'lcs_matrix': lcs_matrix, 'n_runs': len(call_sequences) } - + return stats @@ -556,26 +556,26 @@ def plot_per_run_statistics(results: Dict, output_dir: str): import matplotlib matplotlib.use('Agg') # Use non-interactive backend import matplotlib.pyplot as plt - + stats = results['statistics'] runs = results['runs'] n_runs = len(runs) - + # 1. E2E Delay Distribution (Box Plot) if 'e2e_delay' in stats: fig, axes = plt.subplots(2, 2, figsize=(16, 12)) - + # Chart 1: E2E Delay Box Plot ax1 = axes[0, 0] e2e_data = stats['e2e_delay'] - bp = ax1.boxplot([e2e_data['values']], labels=['E2E Delay'], patch_artist=True, + bp = ax1.boxplot([e2e_data['values']], labels=['E2E Delay'], patch_artist=True, widths=0.6, showmeans=True, meanline=True) bp['boxes'][0].set_facecolor('#3498db') bp['boxes'][0].set_alpha(0.7) ax1.set_ylabel('Delay (ms)', fontsize=12, fontweight='bold') ax1.set_title(f'E2E Delay Distribution\n(n={e2e_data["n"]} runs)', fontsize=14, fontweight='bold') ax1.grid(axis='y', alpha=0.3) - + # Add statistics text stats_text = f'Mean: {e2e_data["mean"]:.2f} ms\n' stats_text += f'Std: {e2e_data["std"]:.2f} ms\n' @@ -584,10 +584,10 @@ def plot_per_run_statistics(results: Dict, output_dir: str): stats_text += f'p50: {e2e_data["p50"]:.2f} ms\n' stats_text += f'p95: {e2e_data["p95"]:.2f} ms\n' stats_text += f'CV: {e2e_data["cv"]:.3f}' - ax1.text(0.02, 0.98, stats_text, transform=ax1.transAxes, + ax1.text(0.02, 0.98, stats_text, transform=ax1.transAxes, verticalalignment='top', fontsize=10, family='monospace', bbox=dict(boxstyle='round', facecolor='wheat', alpha=0.5)) - + # Chart 2: E2E Delay per Run ax2 = axes[0, 1] run_ids = list(range(1, n_runs + 1)) @@ -599,28 +599,28 @@ def plot_per_run_statistics(results: Dict, output_dir: str): ax2.set_title('E2E Delay per Run', fontsize=14, fontweight='bold') ax2.grid(alpha=0.3, linestyle='--') ax2.set_xticks(run_ids) - + # Chart 3: Component Delay Distributions ax3 = axes[1, 0] delay_data = [] labels = [] colors_list = [] - + if 'agent_llm_delay' in stats: delay_data.append(stats['agent_llm_delay']['values']) labels.append('Agent-LLM') colors_list.append('#e74c3c') - + if 'inter_agent_delay' in stats: delay_data.append(stats['inter_agent_delay']['values']) labels.append('Inter-Agent') colors_list.append('#3498db') - + if 'agent_processing_delay' in stats: delay_data.append(stats['agent_processing_delay']['values']) labels.append('Processing') colors_list.append('#2ecc71') - + if delay_data: bp = ax3.boxplot(delay_data, labels=labels, patch_artist=True, widths=0.6, showmeans=True, meanline=True) @@ -630,13 +630,13 @@ def plot_per_run_statistics(results: Dict, output_dir: str): ax3.set_ylabel('Delay (ms)', fontsize=12, fontweight='bold') ax3.set_title('Component Delay Distributions\n(across all runs)', fontsize=14, fontweight='bold') ax3.grid(axis='y', alpha=0.3) - + # Chart 4: Statistics Table ax4 = axes[1, 1] ax4.axis('off') - + table_data = [['Metric', 'Mean ± Std', 'Min', 'Max', 'p95', 'CV']] - + if 'e2e_delay' in stats: e2e = stats['e2e_delay'] table_data.append([ @@ -647,7 +647,7 @@ def plot_per_run_statistics(results: Dict, output_dir: str): f'{e2e["p95"]:.2f}', f'{e2e["cv"]:.3f}' ]) - + if 'agent_llm_delay' in stats: llm = stats['agent_llm_delay'] table_data.append([ @@ -658,7 +658,7 @@ def plot_per_run_statistics(results: Dict, output_dir: str): f'{llm["p95"]:.2f}', '-' ]) - + if 'inter_agent_delay' in stats: inter = stats['inter_agent_delay'] table_data.append([ @@ -669,7 +669,7 @@ def plot_per_run_statistics(results: Dict, output_dir: str): f'{inter["p95"]:.2f}', '-' ]) - + if 'agent_processing_delay' in stats: proc = stats['agent_processing_delay'] table_data.append([ @@ -680,7 +680,7 @@ def plot_per_run_statistics(results: Dict, output_dir: str): f'{proc["p95"]:.2f}', '-' ]) - + if len(table_data) > 1: table = ax4.table(cellText=table_data[1:], colLabels=table_data[0], cellLoc='center', loc='center', bbox=[0, 0, 1, 1]) @@ -688,18 +688,18 @@ def plot_per_run_statistics(results: Dict, output_dir: str): table.set_fontsize(9) table.scale(1, 2) ax4.set_title('Delay Statistics Across Runs', fontsize=14, fontweight='bold', pad=20) - + plt.tight_layout() plt.savefig(str(Path(output_dir) / "2_delay_breakdown_per_run.pdf"), dpi=300, bbox_inches='tight') print(f"✓ Per-run delay breakdown chart saved to {Path(output_dir) / '2_delay_breakdown_per_run.pdf'}") plt.close() - + # 2. Token Consumption Statistics if 'tokens' in stats: fig, axes = plt.subplots(2, 2, figsize=(16, 10)) - + token_stats = stats['tokens'] - + # Chart 1: Total tokens across all runs ax1 = axes[0, 0] categories = ['Prompt', 'Completion', 'Total'] @@ -712,30 +712,30 @@ def plot_per_run_statistics(results: Dict, output_dir: str): for i, v in enumerate(values): if v > 0: ax1.text(i, v + max(values) * 0.01, f'{v:,}', ha='center', va='bottom', fontweight='bold') - + # Chart 2: Tokens per run (distribution) - Box plots for Prompt, Completion, and Total ax2 = axes[0, 1] box_data = [] box_labels = [] box_colors = [] - + if token_stats['per_run_prompt']: box_data.append(token_stats['per_run_prompt']) box_labels.append('Prompt') box_colors.append('#3498db') - + if token_stats['per_run_completion']: box_data.append(token_stats['per_run_completion']) box_labels.append('Completion') box_colors.append('#e74c3c') - + if token_stats['per_run_totals']: box_data.append(token_stats['per_run_totals']) box_labels.append('Total') box_colors.append('#2ecc71') - + if box_data: - bp = ax2.boxplot(box_data, labels=box_labels, patch_artist=True, + bp = ax2.boxplot(box_data, labels=box_labels, patch_artist=True, widths=0.6, showmeans=True, meanline=True) for patch, color in zip(bp['boxes'], box_colors): patch.set_facecolor(color) @@ -743,32 +743,32 @@ def plot_per_run_statistics(results: Dict, output_dir: str): ax2.set_ylabel('Token Count', fontsize=12, fontweight='bold') ax2.set_title('Token Consumption Distribution per Run', fontsize=14, fontweight='bold') ax2.grid(axis='y', alpha=0.3) - + # Chart 3: Tokens per run (line plot showing trend across runs) ax3 = axes[1, 0] run_ids = list(range(1, n_runs + 1)) - + # Plot prompt, completion, and total tokens as separate lines if token_stats['per_run_prompt']: - ax3.plot(run_ids, token_stats['per_run_prompt'], 'o-', color='#3498db', - linewidth=2, markersize=10, label='Prompt Tokens', markerfacecolor='#3498db', + ax3.plot(run_ids, token_stats['per_run_prompt'], 'o-', color='#3498db', + linewidth=2, markersize=10, label='Prompt Tokens', markerfacecolor='#3498db', markeredgecolor='white', markeredgewidth=1.5) if token_stats['per_run_completion']: - ax3.plot(run_ids, token_stats['per_run_completion'], 's-', color='#e74c3c', + ax3.plot(run_ids, token_stats['per_run_completion'], 's-', color='#e74c3c', linewidth=2, markersize=10, label='Completion Tokens', markerfacecolor='#e74c3c', markeredgecolor='white', markeredgewidth=1.5) if token_stats['per_run_totals']: - ax3.plot(run_ids, token_stats['per_run_totals'], '^-', color='#2ecc71', + ax3.plot(run_ids, token_stats['per_run_totals'], '^-', color='#2ecc71', linewidth=2, markersize=10, label='Total Tokens', markerfacecolor='#2ecc71', markeredgecolor='white', markeredgewidth=1.5) - + ax3.set_xlabel('Run ID', fontsize=12, fontweight='bold') ax3.set_ylabel('Token Count', fontsize=12, fontweight='bold') ax3.set_title('Token Consumption Trend Across Runs', fontsize=14, fontweight='bold') ax3.legend(loc='best', fontsize=10) ax3.grid(alpha=0.3, linestyle='--') ax3.set_xticks(run_ids) - + # Chart 4: Statistics table ax4 = axes[1, 1] ax4.axis('off') @@ -787,16 +787,16 @@ def plot_per_run_statistics(results: Dict, output_dir: str): table.set_fontsize(10) table.scale(1, 2) ax4.set_title('Token Consumption Statistics', fontsize=14, fontweight='bold', pad=20) - + plt.tight_layout() plt.savefig(str(Path(output_dir) / "1_token_consumption_per_run.pdf"), dpi=300, bbox_inches='tight') print(f"✓ Per-run token consumption chart saved to {Path(output_dir) / '1_token_consumption_per_run.pdf'}") plt.close() - + # 3. CPU/Memory Usage Statistics if 'cpu_usage' in stats or 'memory_usage' in stats: fig, axes = plt.subplots(2, 2, figsize=(16, 10)) - + # Chart 1: CPU Usage Distribution (Mean, Peak, Min) ax1 = axes[0, 0] if 'cpu_usage' in stats: @@ -804,7 +804,7 @@ def plot_per_run_statistics(results: Dict, output_dir: str): box_data = [] box_labels = [] box_colors = [] - + if cpu_stats['mean']['values']: box_data.append(cpu_stats['mean']['values']) box_labels.append('Mean') @@ -817,7 +817,7 @@ def plot_per_run_statistics(results: Dict, output_dir: str): box_data.append(cpu_stats['min']['values']) box_labels.append('Min') box_colors.append('#2ecc71') - + if box_data: bp = ax1.boxplot(box_data, labels=box_labels, patch_artist=True, widths=0.6, showmeans=True, meanline=True) @@ -827,7 +827,7 @@ def plot_per_run_statistics(results: Dict, output_dir: str): ax1.set_ylabel('CPU Usage (%)', fontsize=12, fontweight='bold') ax1.set_title('CPU Usage Distribution Across Runs', fontsize=14, fontweight='bold') ax1.grid(axis='y', alpha=0.3) - + # Chart 2: Memory Usage Distribution (Mean, Peak, Min) ax2 = axes[0, 1] if 'memory_usage' in stats: @@ -835,7 +835,7 @@ def plot_per_run_statistics(results: Dict, output_dir: str): box_data = [] box_labels = [] box_colors = [] - + if memory_stats['mean']['values']: box_data.append(memory_stats['mean']['values']) box_labels.append('Mean') @@ -848,7 +848,7 @@ def plot_per_run_statistics(results: Dict, output_dir: str): box_data.append(memory_stats['min']['values']) box_labels.append('Min') box_colors.append('#2ecc71') - + if box_data: bp = ax2.boxplot(box_data, labels=box_labels, patch_artist=True, widths=0.6, showmeans=True, meanline=True) @@ -858,7 +858,7 @@ def plot_per_run_statistics(results: Dict, output_dir: str): ax2.set_ylabel('Memory Usage (MB)', fontsize=12, fontweight='bold') ax2.set_title('Memory Usage Distribution Across Runs', fontsize=14, fontweight='bold') ax2.grid(axis='y', alpha=0.3) - + # Chart 3: CPU Usage per Run (Mean values) ax3 = axes[1, 0] if 'cpu_usage' in stats and stats['cpu_usage']['mean']['values']: @@ -877,7 +877,7 @@ def plot_per_run_statistics(results: Dict, output_dir: str): ax3.legend(loc='best', fontsize=10) ax3.grid(alpha=0.3, linestyle='--') ax3.set_xticks(run_ids) - + # Chart 4: Memory Usage per Run (Mean values) ax4 = axes[1, 1] if 'memory_usage' in stats and stats['memory_usage']['mean']['values']: @@ -896,23 +896,23 @@ def plot_per_run_statistics(results: Dict, output_dir: str): ax4.legend(loc='best', fontsize=10) ax4.grid(alpha=0.3, linestyle='--') ax4.set_xticks(run_ids) - + plt.tight_layout() plt.savefig(str(Path(output_dir) / "3_cpu_memory_usage_per_run.pdf"), dpi=300, bbox_inches='tight') print(f"✓ Per-run CPU/Memory usage chart saved to {Path(output_dir) / '3_cpu_memory_usage_per_run.pdf'}") plt.close() - + # 4. Message Size Statistics if 'message_sizes' in stats: msg_stats = stats['message_sizes'] fig, axes = plt.subplots(2, 3, figsize=(20, 12)) - + # Chart 1: Overall message size totals per run (box plots) ax1 = axes[0, 0] box_data = [] box_labels = [] box_colors = [] - + if msg_stats.get('inter_agent_input', {}).get('values'): box_data.append(msg_stats['inter_agent_input']['values']) box_labels.append('Inter-Agent\nInput') @@ -929,7 +929,7 @@ def plot_per_run_statistics(results: Dict, output_dir: str): box_data.append(msg_stats['agent_llm_output']['values']) box_labels.append('Agent-LLM\nOutput') box_colors.append('#f39c12') - + if box_data: bp = ax1.boxplot(box_data, labels=box_labels, patch_artist=True, widths=0.6, showmeans=True, meanline=True) @@ -939,7 +939,7 @@ def plot_per_run_statistics(results: Dict, output_dir: str): ax1.set_ylabel('Total Message Size (KB)', fontsize=12, fontweight='bold') ax1.set_title('Message Size Distribution Across Runs', fontsize=14, fontweight='bold') ax1.grid(axis='y', alpha=0.3) - + # Chart 2: Message size totals per run (line plot) ax2 = axes[0, 1] run_ids = list(range(1, n_runs + 1)) @@ -965,7 +965,7 @@ def plot_per_run_statistics(results: Dict, output_dir: str): ax2.legend(loc='best', fontsize=9) ax2.grid(alpha=0.3, linestyle='--') ax2.set_xticks(run_ids) - + # Chart 3: Per-agent message size totals (box plots) ax3 = axes[0, 2] if msg_stats.get('per_agent_totals'): @@ -974,12 +974,12 @@ def plot_per_run_statistics(results: Dict, output_dir: str): # Get top agents by average total agent_avgs = {agent: np.mean(totals) for agent, totals in agent_totals.items()} top_agents = sorted(agent_avgs.items(), key=lambda x: x[1], reverse=True)[:10] - + if top_agents: box_data = [agent_totals[agent] for agent, _ in top_agents] box_labels = [agent for agent, _ in top_agents] colors_list = plt.cm.Set3(np.linspace(0, 1, len(box_labels))) - + bp = ax3.boxplot(box_data, labels=box_labels, patch_artist=True, widths=0.6, showmeans=True, meanline=True) for patch, color in zip(bp['boxes'], colors_list): @@ -989,7 +989,7 @@ def plot_per_run_statistics(results: Dict, output_dir: str): ax3.set_title('Per-Agent Message Size (Top 10)', fontsize=12, fontweight='bold') ax3.set_xticklabels(box_labels, rotation=45, ha='right', fontsize=8) ax3.grid(axis='y', alpha=0.3) - + # Chart 4: Per agent-pair message size totals (box plots) ax4 = axes[1, 0] if msg_stats.get('per_pair_totals'): @@ -998,12 +998,12 @@ def plot_per_run_statistics(results: Dict, output_dir: str): # Get top pairs by average total pair_avgs = {pair: np.mean(totals) for pair, totals in pair_totals.items()} top_pairs = sorted(pair_avgs.items(), key=lambda x: x[1], reverse=True)[:15] # Show more pairs - + if top_pairs: box_data = [pair_totals[pair] for pair, _ in top_pairs] box_labels = [f'{src}→{tgt}' for (src, tgt), _ in top_pairs] colors_list = plt.cm.Pastel1(np.linspace(0, 1, len(box_labels))) - + bp = ax4.boxplot(box_data, labels=box_labels, patch_artist=True, widths=0.6, showmeans=True, meanline=True) for patch, color in zip(bp['boxes'], colors_list): @@ -1022,7 +1022,7 @@ def plot_per_run_statistics(results: Dict, output_dir: str): else: ax4.axis('off') ax4.text(0.5, 0.5, 'No agent-pair message size data', ha='center', va='center', fontsize=10) - + # Chart 5: Per agent-LLM message size totals (box plots) ax5 = axes[1, 1] if msg_stats.get('per_agent_llm_totals'): @@ -1031,12 +1031,12 @@ def plot_per_run_statistics(results: Dict, output_dir: str): # Get top agents by average total agent_llm_avgs = {agent: np.mean(totals) for agent, totals in agent_llm_totals.items()} top_agents_llm = sorted(agent_llm_avgs.items(), key=lambda x: x[1], reverse=True)[:10] - + if top_agents_llm: box_data = [agent_llm_totals[agent] for agent, _ in top_agents_llm] box_labels = [agent for agent, _ in top_agents_llm] colors_list = plt.cm.Set2(np.linspace(0, 1, len(box_labels))) - + bp = ax5.boxplot(box_data, labels=box_labels, patch_artist=True, widths=0.6, showmeans=True, meanline=True) for patch, color in zip(bp['boxes'], colors_list): @@ -1046,12 +1046,12 @@ def plot_per_run_statistics(results: Dict, output_dir: str): ax5.set_title('Per Agent-LLM Message Size (Top 10)', fontsize=12, fontweight='bold') ax5.set_xticklabels(box_labels, rotation=45, ha='right', fontsize=8) ax5.grid(axis='y', alpha=0.3) - + # Chart 6: Statistics table ax6 = axes[1, 2] ax6.axis('off') table_data = [['Metric', 'Mean ± Std (KB)', 'Min', 'Max', 'p95']] - + if msg_stats.get('inter_agent_input'): iai = msg_stats['inter_agent_input'] table_data.append([ @@ -1088,7 +1088,7 @@ def plot_per_run_statistics(results: Dict, output_dir: str): f'{alo["max"]:.2f}', f'{alo["p95"]:.2f}' ]) - + if len(table_data) > 1: table = ax6.table(cellText=table_data[1:], colLabels=table_data[0], cellLoc='center', loc='center', bbox=[0, 0, 1, 1]) @@ -1096,12 +1096,12 @@ def plot_per_run_statistics(results: Dict, output_dir: str): table.set_fontsize(9) table.scale(1, 2) ax6.set_title('Message Size Statistics Across Runs', fontsize=14, fontweight='bold', pad=20) - + plt.tight_layout() plt.savefig(str(Path(output_dir) / "4_message_sizes_per_run.pdf"), dpi=300, bbox_inches='tight') print(f"✓ Per-run message size chart saved to {Path(output_dir) / '4_message_sizes_per_run.pdf'}") plt.close() - + # 5. Generate statistics report report_file = Path(output_dir) / "statistics_report.txt" with open(report_file, 'w') as f: @@ -1109,7 +1109,7 @@ def plot_per_run_statistics(results: Dict, output_dir: str): f.write("Per-Run Analysis Statistics Report\n") f.write("=" * 80 + "\n\n") f.write(f"Number of runs analyzed: {n_runs}\n\n") - + if 'e2e_delay' in stats: e2e = stats['e2e_delay'] f.write("E2E Delay Statistics:\n") @@ -1118,7 +1118,7 @@ def plot_per_run_statistics(results: Dict, output_dir: str): f.write(f" Percentiles: p50={e2e['p50']:.2f} ms, p95={e2e['p95']:.2f} ms, p99={e2e['p99']:.2f} ms\n") f.write(f" Coefficient of Variation (CV): {e2e['cv']:.3f} ({e2e['cv']*100:.1f}%)\n") f.write(f" Number of runs: {e2e['n']}\n\n") - + if 'tokens' in stats: tok = stats['tokens'] f.write("Token Consumption Statistics:\n") @@ -1127,7 +1127,7 @@ def plot_per_run_statistics(results: Dict, output_dir: str): f.write(f" - Completion: {tok['total_completion']:,} tokens\n") f.write(f" Mean per run: {tok['mean_per_run']:.0f} tokens\n") f.write(f" Std per run: {tok['std_per_run']:.0f} tokens\n\n") - + if 'agent_llm_delay' in stats: llm = stats['agent_llm_delay'] f.write("Agent-LLM Delay Statistics:\n") @@ -1135,7 +1135,7 @@ def plot_per_run_statistics(results: Dict, output_dir: str): f.write(f" Min: {llm['min']:.2f} ms, Max: {llm['max']:.2f} ms\n") f.write(f" p95: {llm['p95']:.2f} ms\n") f.write(f" Total observations: {llm['n']}\n\n") - + if 'inter_agent_delay' in stats: inter = stats['inter_agent_delay'] f.write("Inter-Agent Delay Statistics:\n") @@ -1143,7 +1143,7 @@ def plot_per_run_statistics(results: Dict, output_dir: str): f.write(f" Min: {inter['min']:.2f} ms, Max: {inter['max']:.2f} ms\n") f.write(f" p95: {inter['p95']:.2f} ms\n") f.write(f" Total observations: {inter['n']}\n\n") - + if 'cpu_usage' in stats: cpu = stats['cpu_usage'] f.write("CPU Usage Statistics (Process-Level):\n") @@ -1156,7 +1156,7 @@ def plot_per_run_statistics(results: Dict, output_dir: str): f.write(f" Min: {cpu['peak']['min']:.2f}%, Max: {cpu['peak']['max']:.2f}%\n") f.write(f" p50: {cpu['peak']['p50']:.2f}%, p95: {cpu['peak']['p95']:.2f}%\n") f.write(f" Number of runs: {cpu['mean']['n']}\n\n") - + if 'memory_usage' in stats: mem = stats['memory_usage'] f.write("Memory Usage Statistics (Process-Level):\n") @@ -1169,13 +1169,13 @@ def plot_per_run_statistics(results: Dict, output_dir: str): f.write(f" Min: {mem['peak']['min']:.2f} MB, Max: {mem['peak']['max']:.2f} MB\n") f.write(f" p50: {mem['peak']['p50']:.2f} MB, p95: {mem['peak']['p95']:.2f} MB\n") f.write(f" Number of runs: {mem['mean']['n']}\n\n") - + # Call graph similarity statistics if 'call_graph_similarity' in stats: cg_stats = stats['call_graph_similarity'] f.write("Call Graph Similarity Statistics:\n") f.write("=" * 60 + "\n") - + # Multi-graph metrics (more comprehensive) f.write("Multi-Graph Similarity Metrics:\n") f.write(f" Overall Consistency: {cg_stats['overall_consistency']:.4f}\n") @@ -1189,7 +1189,7 @@ def plot_per_run_statistics(results: Dict, output_dir: str): f.write(f" (Average of all pairwise Jaccard similarities)\n") f.write(f" Stability Index: {cg_stats['stability_index']:.4f}\n") f.write(f" (Std dev of pairwise similarities, lower = more stable)\n\n") - + # Consecutive pairwise (for reference) f.write("Consecutive Pairwise Similarities (for reference):\n") f.write(f" Mean: {cg_stats['jaccard_mean']:.4f}\n") @@ -1202,7 +1202,7 @@ def plot_per_run_statistics(results: Dict, output_dir: str): union = len(edges1 | edges2) f.write(f" Run {i+1} vs Run {i+2}: {sim:.4f} ({intersection} common, {union} total)\n") f.write("\n") - + # Call sequence similarity statistics if 'call_sequence_similarity' in stats: f.write("\nCall Sequence Similarity Statistics (LCS - Sequence Order):\n") @@ -1215,7 +1215,7 @@ def plot_per_run_statistics(results: Dict, output_dir: str): seq2 = runs[i + 1].get('call_sequence', []) f.write(f" Run {i+1} vs Run {i+2}: {sim:.4f} (seq1 length: {len(seq1)}, seq2 length: {len(seq2)})\n") f.write("\n") - + if 'message_sizes' in stats: msg = stats['message_sizes'] f.write("Message Size Statistics:\n") @@ -1240,9 +1240,9 @@ def plot_per_run_statistics(results: Dict, output_dir: str): f.write(f" Min: {msg['agent_llm_output']['min']:.2f} KB, Max: {msg['agent_llm_output']['max']:.2f} KB\n") f.write(f" p50: {msg['agent_llm_output']['p50']:.2f} KB, p95: {msg['agent_llm_output']['p95']:.2f} KB\n") f.write("\n") - + print(f"✓ Statistics report saved to {report_file}") - + # 6. Call Graph Visualization and Similarity Analysis if len(runs) > 0: # Extract call graphs for each run @@ -1281,7 +1281,7 @@ def _normalize_sequence(seq: List[tuple]) -> List[tuple]: call_graphs = [_normalize_edge_set(edges) for edges in call_graphs] call_sequences = [_normalize_sequence(seq) for seq in call_sequences] - + # Generate call graph visualization even if there's only one run (show the call graph) # Similarity analysis requires at least 2 runs if call_graphs: @@ -1296,7 +1296,7 @@ def _normalize_sequence(seq: List[tuple]) -> List[tuple]: if ns and nt and ns != nt: core_edges_normalized.add((ns, nt)) max_sequence_len = max((len(seq) for seq in call_sequences), default=0) - + # Create visualization with multiple subplots # Layout: call graphs (top row) + similarity matrices (middle, side by side) + metrics (bottom) from matplotlib.gridspec import GridSpec @@ -1324,7 +1324,7 @@ def _normalize_sequence(seq: List[tuple]) -> List[tuple]: else: order_norm = None order_colorbar_added = False - + # Plot call graphs for each run # Extract agents from edges first all_agents = set() @@ -1337,7 +1337,7 @@ def _normalize_sequence(seq: List[tuple]) -> List[tuple]: for src, tgt in seq: all_agents.add(src) all_agents.add(tgt) - + # If no agents found from edges, extract from traces (for systems where agents don't call each other directly) if not all_agents and runs: from .extractors import extract_token_consumption @@ -1349,9 +1349,9 @@ def _normalize_sequence(seq: List[tuple]) -> List[tuple]: for agent_name in per_agent_tokens.keys(): normalized = agent_name.replace('_agent', '').replace('-agent', '').replace('_', '-').replace(' ', '-').lower() all_agents.add(normalized) - + all_agents = sorted(list(all_agents)) - + # Create node positions (circular layout) import math node_positions = {} @@ -1359,7 +1359,7 @@ def _normalize_sequence(seq: List[tuple]) -> List[tuple]: for i, agent in enumerate(all_agents): angle = 2 * math.pi * i / n_nodes if n_nodes > 0 else 0 node_positions[agent] = (math.cos(angle), math.sin(angle)) - + for i, edges in enumerate(call_graphs): if i < n_cols: ax = fig.add_subplot(gs[0, i]) @@ -1371,7 +1371,7 @@ def _normalize_sequence(seq: List[tuple]) -> List[tuple]: for pair in edges_to_draw: edge_occurrence_totals[pair] += 1 edge_occurrence_seen = defaultdict(int) - + # Draw edges for idx, (src, tgt) in enumerate(edges_to_draw): if src in node_positions and tgt in node_positions: @@ -1385,7 +1385,7 @@ def _normalize_sequence(seq: List[tuple]) -> List[tuple]: else: color = '#e74c3c' if is_core else 'gray' linewidth = 3 if is_core else 2 - + # Offset duplicate edges slightly to avoid overlapping text/arrows edge_occurrence_seen[(src, tgt)] += 1 occurrence_idx = edge_occurrence_seen[(src, tgt)] @@ -1405,14 +1405,14 @@ def _normalize_sequence(seq: List[tuple]) -> List[tuple]: y2o = y2 + perp_y * offset else: x1o, y1o, x2o, y2o = x1, y1, x2, y2 - + ax.plot([x1o, x2o], [y1o, y2o], color=color, linewidth=linewidth, alpha=0.8, zorder=1) # Draw arrow dx_line, dy_line = x2o - x1o, y2o - y1o length_line = math.sqrt(dx_line**2 + dy_line**2) if length_line > 0: dx_arrow, dy_arrow = dx_line / length_line * 0.15, dy_line / length_line * 0.15 - ax.arrow(x2o - dx_arrow, y2o - dy_arrow, dx_arrow, dy_arrow, head_width=0.05, head_length=0.05, + ax.arrow(x2o - dx_arrow, y2o - dy_arrow, dx_arrow, dy_arrow, head_width=0.05, head_length=0.05, fc=color, ec=color, zorder=2) # Annotate order number at midpoint when we have sequence data if sequence: @@ -1420,7 +1420,7 @@ def _normalize_sequence(seq: List[tuple]) -> List[tuple]: ax.text(mx, my, f'{idx + 1}', ha='center', va='center', fontsize=8, fontweight='bold', color='black', zorder=5, bbox=dict(boxstyle='round,pad=0.2', facecolor='white', alpha=0.7)) - + # Draw nodes - show all agents if we have them, otherwise show agents from edges if all_agents: # Show all agents (even if no edges, they're still part of the system) @@ -1436,12 +1436,12 @@ def _normalize_sequence(seq: List[tuple]) -> List[tuple]: if in_graph: ax.scatter([x], [y], s=500, c='#3498db', edgecolors='black', linewidths=2, zorder=3) ax.text(x, y - 0.2, agent, ha='center', va='top', fontsize=8, fontweight='bold', zorder=4) - + ax.set_xlim(-1.5, 1.5) ax.set_ylim(-1.5, 1.5) call_count_label = f"{len(sequence)} calls" if sequence else f"{len(edges)} edges" ax.set_title(f'Run {i+1}\n({call_count_label})', fontsize=12, fontweight='bold', pad=10) - + # Add a single colorbar showing call order (shared scale across runs) if sequence and order_cmap and order_norm and not order_colorbar_added: sm = plt.cm.ScalarMappable(cmap=order_cmap, norm=order_norm) @@ -1457,45 +1457,45 @@ def _normalize_sequence(seq: List[tuple]) -> List[tuple]: cbar.set_ticks(ticks) cbar.set_ticklabels([str(t + 1) for t in ticks]) order_colorbar_added = True - + # Row 2: Similarity Matrices (Jaccard and LCS side by side) - only if we have similarity stats and multiple runs if has_similarity and 'similarity_matrix' in cg_stats and len(cg_stats['similarity_matrix']) > 1: # Calculate column positions: matrices start after call graph columns # Each matrix spans 2 columns for proper spacing matrix_start = n_runs matrix_mid = matrix_start + 2 - + # Jaccard Similarity Matrix (Edge Set, No Order) - Left side ax_jaccard = fig.add_subplot(gs[1, matrix_start:matrix_mid]) ax_jaccard.grid(False) # Explicitly disable grid jaccard_matrix = np.array(cg_stats['similarity_matrix']) # Directly plot the matrix without extent - let imshow handle it naturally # Then adjust ticks and labels to match the data - im1 = ax_jaccard.imshow(jaccard_matrix, cmap='RdYlGn', vmin=0, vmax=1, + im1 = ax_jaccard.imshow(jaccard_matrix, cmap='RdYlGn', vmin=0, vmax=1, aspect='equal', interpolation='nearest', origin='upper') - + # Set ticks to match data dimensions (3x3) - only show ticks at data points ax_jaccard.set_xticks(range(n_runs)) ax_jaccard.set_yticks(range(n_runs)) ax_jaccard.set_xticklabels([f'Run {i+1}' for i in range(n_runs)], fontsize=9) ax_jaccard.set_yticklabels([f'Run {i+1}' for i in range(n_runs)], fontsize=9) - + # Add text annotations at correct data positions for i in range(n_runs): for j in range(n_runs): ax_jaccard.text(j, i, f'{jaccard_matrix[i, j]:.2f}', ha="center", va="center", color="black", fontweight='bold', fontsize=9) - + ax_jaccard.set_xlabel('Run', fontsize=11, fontweight='bold') ax_jaccard.set_ylabel('Run', fontsize=11, fontweight='bold') - ax_jaccard.set_title('Jaccard Similarity (Edge Set, No Order)', + ax_jaccard.set_title('Jaccard Similarity (Edge Set, No Order)', fontsize=12, fontweight='bold', pad=10) - + # Add colorbar cbar1 = plt.colorbar(im1, ax=ax_jaccard, fraction=0.046, pad=0.04) cbar1.set_label('Similarity', fontsize=10, fontweight='bold') cbar1.ax.tick_params(labelsize=8) - + # LCS Similarity Matrix (Sequence Order) - Right side if 'call_sequence_similarity' in stats and 'lcs_matrix' in stats['call_sequence_similarity']: ax_lcs = fig.add_subplot(gs[1, matrix_mid:]) @@ -1504,43 +1504,43 @@ def _normalize_sequence(seq: List[tuple]) -> List[tuple]: # Directly plot the matrix without extent - let imshow handle it naturally im2 = ax_lcs.imshow(lcs_matrix, cmap='RdYlGn', vmin=0, vmax=1, aspect='equal', interpolation='nearest', origin='upper') - + # Set ticks to match data dimensions (3x3) - only show ticks at data points ax_lcs.set_xticks(range(n_runs)) ax_lcs.set_yticks(range(n_runs)) ax_lcs.set_xticklabels([f'Run {i+1}' for i in range(n_runs)], fontsize=9) ax_lcs.set_yticklabels([f'Run {i+1}' for i in range(n_runs)], fontsize=9) - + # Add text annotations at correct data positions for i in range(n_runs): for j in range(n_runs): ax_lcs.text(j, i, f'{lcs_matrix[i, j]:.2f}', ha="center", va="center", color="black", fontweight='bold', fontsize=9) - + ax_lcs.set_xlabel('Run', fontsize=11, fontweight='bold') ax_lcs.set_ylabel('Run', fontsize=11, fontweight='bold') - ax_lcs.set_title('LCS Similarity (Sequence Order)', + ax_lcs.set_title('LCS Similarity (Sequence Order)', fontsize=12, fontweight='bold', pad=10) - + # Add colorbar cbar2 = plt.colorbar(im2, ax=ax_lcs, fraction=0.046, pad=0.04) cbar2.set_label('Similarity', fontsize=10, fontweight='bold') cbar2.ax.tick_params(labelsize=8) - + # Row 3: Multi-Graph Metrics Summary (only if we have similarity stats) if has_similarity and n_runs > 1: ax_metrics = fig.add_subplot(gs[2, :]) ax_metrics.axis('off') - + # Create metrics summary table metrics_data = [ ['Metric', 'Value', 'Interpretation'], - ['Overall Consistency (Jaccard)', f"{cg_stats.get('overall_consistency', 0):.4f}", + ['Overall Consistency (Jaccard)', f"{cg_stats.get('overall_consistency', 0):.4f}", f"Core edges / Total unique edges ({len(cg_stats.get('core_edges', []))} / {cg_stats.get('total_unique_edges', 0)})"], ['Mean Pairwise Jaccard', f"{cg_stats.get('mean_pairwise_similarity', 0):.4f}", f"Average edge set similarity (no order)"], ] - + # Add LCS similarity if available if 'call_sequence_similarity' in stats: lcs_stats = stats['call_sequence_similarity'] @@ -1548,33 +1548,33 @@ def _normalize_sequence(seq: List[tuple]) -> List[tuple]: f"Average sequence similarity (with order)"]) metrics_data.append(['LCS Stability Index', f"{lcs_stats.get('lcs_stability_index', 0):.4f}", f"Std dev of LCS similarities (lower = more stable)"]) - + metrics_data.extend([ ['Jaccard Stability Index', f"{cg_stats.get('stability_index', 0):.4f}", f"Std dev of Jaccard similarities (lower = more stable)"], - ['Core Edges', f"{len(cg_stats.get('core_edges', []))}", + ['Core Edges', f"{len(cg_stats.get('core_edges', []))}", f"Edges appearing in all {n_runs} runs"], ]) - + # Add core edges list (truncated if too long) core_edges = cg_stats.get('core_edges', []) core_edges_str = str(core_edges) if len(core_edges_str) > 80: core_edges_str = core_edges_str[:77] + "..." metrics_data.append(['Core Edge List', core_edges_str, 'Stable call patterns']) - + table = ax_metrics.table(cellText=metrics_data[1:], colLabels=metrics_data[0], cellLoc='left', loc='center', bbox=[0, 0, 1, 1]) table.auto_set_font_size(False) table.set_fontsize(10) table.scale(1, 2.5) - + # Style header row for i in range(3): table[(0, i)].set_facecolor('#3498db') table[(0, i)].set_text_props(weight='bold', color='white') - - ax_metrics.set_title('Multi-Graph Similarity Metrics Summary', + + ax_metrics.set_title('Multi-Graph Similarity Metrics Summary', fontsize=14, fontweight='bold', pad=20) elif n_runs == 1: # Single run: add a note in an empty subplot if available @@ -1582,12 +1582,12 @@ def _normalize_sequence(seq: List[tuple]) -> List[tuple]: try: ax_note = fig.add_subplot(gs[0, n_runs]) ax_note.axis('off') - ax_note.text(0.5, 0.5, f'Note: Similarity analysis requires\nat least 2 runs. Current: {n_runs} run(s).', + ax_note.text(0.5, 0.5, f'Note: Similarity analysis requires\nat least 2 runs. Current: {n_runs} run(s).', ha='center', va='center', fontsize=10, style='italic', bbox=dict(boxstyle='round', facecolor='wheat', alpha=0.5)) except: pass # If subplot creation fails, just skip the note - + plt.tight_layout() plt.savefig(str(Path(output_dir) / "5_call_graph_similarity.pdf"), dpi=300, bbox_inches='tight') print(f"✓ Call graph similarity chart saved to {Path(output_dir) / '5_call_graph_similarity.pdf'}") diff --git a/plot/lib/visualizers.py b/plot/lib/visualizers.py index 9055be5..3fe7991 100644 --- a/plot/lib/visualizers.py +++ b/plot/lib/visualizers.py @@ -14,37 +14,37 @@ def plot_token_consumption(total_tokens: Dict[str, int], per_agent_tokens: Dict[str, Dict[str, int]], output_file: str): """Plot token consumption charts.""" fig = plt.figure(figsize=(16, 10)) - + # Chart 1: Total token consumption (stacked bar) ax1 = plt.subplot(2, 2, 1) categories = ['Prompt Tokens', 'Completion Tokens'] values = [total_tokens['prompt'], total_tokens['completion']] colors = ['#3498db', '#e74c3c'] - + ax1.bar(categories, values, color=colors, alpha=0.8, edgecolor='black', linewidth=1.5) ax1.set_ylabel('Token Count', fontsize=12, fontweight='bold') ax1.set_title('Total Token Consumption (MAS)', fontsize=14, fontweight='bold') ax1.grid(axis='y', alpha=0.3) - + # Add value labels for i, v in enumerate(values): if v > 0: ax1.text(i, v + max(values) * 0.01, f'{v:,}', ha='center', va='bottom', fontweight='bold') - + # Chart 2: Per-agent token consumption (stacked) ax2 = plt.subplot(2, 2, 2) if per_agent_tokens: agents = list(per_agent_tokens.keys()) prompt_vals = [per_agent_tokens[agent]['prompt'] for agent in agents] completion_vals = [per_agent_tokens[agent]['completion'] for agent in agents] - + x = np.arange(len(agents)) width = 0.6 - + ax2.bar(x, prompt_vals, width, label='Prompt Tokens', color='#3498db', alpha=0.8, edgecolor='black') - ax2.bar(x, completion_vals, width, bottom=prompt_vals, label='Completion Tokens', + ax2.bar(x, completion_vals, width, bottom=prompt_vals, label='Completion Tokens', color='#e74c3c', alpha=0.8, edgecolor='black') - + ax2.set_xlabel('Agent', fontsize=12, fontweight='bold') ax2.set_ylabel('Token Count', fontsize=12, fontweight='bold') ax2.set_title('Token Consumption per Agent (Stacked)', fontsize=14, fontweight='bold') @@ -52,27 +52,27 @@ def plot_token_consumption(total_tokens: Dict[str, int], per_agent_tokens: Dict[ ax2.set_xticklabels(agents, rotation=45, ha='right') ax2.legend() ax2.grid(axis='y', alpha=0.3) - + # Add total labels for i, agent in enumerate(agents): total = prompt_vals[i] + completion_vals[i] if total > 0: ax2.text(i, total + max(prompt_vals[i] + completion_vals[i] for i in range(len(agents))) * 0.01, f'{total:,}', ha='center', va='bottom', fontweight='bold', fontsize=9) - + # Chart 3: Per-agent token consumption (side-by-side) ax3 = plt.subplot(2, 2, 3) if per_agent_tokens: agents = list(per_agent_tokens.keys()) prompt_vals = [per_agent_tokens[agent]['prompt'] for agent in agents] completion_vals = [per_agent_tokens[agent]['completion'] for agent in agents] - + x = np.arange(len(agents)) width = 0.35 - + ax3.bar(x - width/2, prompt_vals, width, label='Prompt Tokens', color='#3498db', alpha=0.8, edgecolor='black') ax3.bar(x + width/2, completion_vals, width, label='Completion Tokens', color='#e74c3c', alpha=0.8, edgecolor='black') - + ax3.set_xlabel('Agent', fontsize=12, fontweight='bold') ax3.set_ylabel('Token Count', fontsize=12, fontweight='bold') ax3.set_title('Token Consumption per Agent (Side-by-Side)', fontsize=14, fontweight='bold') @@ -80,121 +80,121 @@ def plot_token_consumption(total_tokens: Dict[str, int], per_agent_tokens: Dict[ ax3.set_xticklabels(agents, rotation=45, ha='right') ax3.legend() ax3.grid(axis='y', alpha=0.3) - + # Chart 4: Token statistics table ax4 = plt.subplot(2, 2, 4) ax4.axis('off') - + stats_data = [ ['Metric', 'Value'], ['Total Prompt Tokens', f'{total_tokens["prompt"]:,}'], ['Total Completion Tokens', f'{total_tokens["completion"]:,}'], ['Total Tokens', f'{total_tokens["total"]:,}'], ] - + if per_agent_tokens: for agent in sorted(per_agent_tokens.keys()): tokens = per_agent_tokens[agent] stats_data.append([f'{agent} Prompt', f'{tokens["prompt"]:,}']) stats_data.append([f'{agent} Completion', f'{tokens["completion"]:,}']) stats_data.append([f'{agent} Total', f'{tokens["total"]:,}']) - + table = ax4.table(cellText=stats_data[1:], colLabels=stats_data[0], cellLoc='center', loc='center', bbox=[0, 0, 1, 1]) table.auto_set_font_size(False) table.set_fontsize(9) table.scale(1, 1.5) ax4.set_title('Token Consumption Statistics', fontsize=14, fontweight='bold', pad=20) - + plt.tight_layout() plt.savefig(output_file, dpi=300, bbox_inches='tight') print(f"✓ Token consumption chart saved to {output_file}") plt.close() -def plot_delay_breakdown(delays: Dict[str, List[float]], per_agent_delays: Dict[str, List[float]], - inter_agent_delays: Dict[Tuple[str, str], List[float]], +def plot_delay_breakdown(delays: Dict[str, List[float]], per_agent_delays: Dict[str, List[float]], + inter_agent_delays: Dict[Tuple[str, str], List[float]], agent_llm_delays: Dict[str, List[float]], output_file: str): """Plot delay breakdown and distribution charts with per-agent details.""" fig = plt.figure(figsize=(20, 14)) - + # Chart 1: Delay breakdown (stacked bar) ax1 = plt.subplot(3, 3, 1) - + if delays['e2e_delay']: # E2E delay is now the total sum of all requests e2e_total = delays['e2e_delay'][0] if delays['e2e_delay'] else 0 - + # Calculate totals for components (sum of all individual delays) llm_total = sum(delays['agent_llm_delay']) if delays['agent_llm_delay'] else 0 inter_total = sum(delays['inter_agent_delay']) if delays['inter_agent_delay'] else 0 proc_total = sum(delays['agent_processing_delay']) if delays['agent_processing_delay'] else 0 - + categories = ['Total E2E Delay\n(All Requests)'] llm_values = [llm_total] inter_values = [inter_total] proc_values = [proc_total] other_values = [max(0, e2e_total - llm_total - inter_total - proc_total)] - + x = np.arange(len(categories)) width = 0.6 - + p1 = ax1.bar(x, llm_values, width, label='Agent-LLM Delay', color='#e74c3c', alpha=0.8, edgecolor='black') p2 = ax1.bar(x, inter_values, width, bottom=llm_values, label='Inter-Agent Delay', color='#3498db', alpha=0.8, edgecolor='black') - p3 = ax1.bar(x, proc_values, width, bottom=np.array(llm_values) + np.array(inter_values), + p3 = ax1.bar(x, proc_values, width, bottom=np.array(llm_values) + np.array(inter_values), label='Processing Delay', color='#2ecc71', alpha=0.8, edgecolor='black') - p4 = ax1.bar(x, other_values, width, + p4 = ax1.bar(x, other_values, width, bottom=np.array(llm_values) + np.array(inter_values) + np.array(proc_values), label='Other', color='#95a5a6', alpha=0.8, edgecolor='black') - + ax1.set_ylabel('Delay (ms)', fontsize=12, fontweight='bold') ax1.set_title('E2E Delay Breakdown', fontsize=14, fontweight='bold') ax1.set_xticks(x) ax1.set_xticklabels(categories) ax1.legend() ax1.grid(axis='y', alpha=0.3) - + # Add total value label - ax1.text(x[0], e2e_total + max(e2e_total * 0.05, 10), f'Total: {e2e_total/1000:.2f}s', + ax1.text(x[0], e2e_total + max(e2e_total * 0.05, 10), f'Total: {e2e_total/1000:.2f}s', ha='center', va='bottom', fontweight='bold') - + # Chart 2: Component delay statistics (box plot or bar chart) ax2 = plt.subplot(3, 3, 2) - + # Show component delays as box plots to avoid overlap has_data = False components = [] delay_data = [] colors_list = [] - + if delays['agent_llm_delay']: components.append('Agent-LLM') delay_data.append(delays['agent_llm_delay']) colors_list.append('#e74c3c') has_data = True - + if delays['inter_agent_delay']: components.append('Inter-Agent') delay_data.append(delays['inter_agent_delay']) colors_list.append('#3498db') has_data = True - + if delays['agent_processing_delay']: components.append('Processing') delay_data.append(delays['agent_processing_delay']) colors_list.append('#2ecc71') has_data = True - + if has_data: # Use box plot to show distribution without overlap - bp = ax2.boxplot(delay_data, labels=components, patch_artist=True, + bp = ax2.boxplot(delay_data, labels=components, patch_artist=True, widths=0.6, showmeans=True, meanline=True) - + # Color the boxes for patch, color in zip(bp['boxes'], colors_list): patch.set_facecolor(color) patch.set_alpha(0.7) - + # Style the median and mean lines for median in bp['medians']: median.set_color('black') @@ -203,7 +203,7 @@ def plot_delay_breakdown(delays: Dict[str, List[float]], per_agent_delays: Dict[ mean.set_color('red') mean.set_linewidth(2) mean.set_linestyle('--') - + ax2.set_ylabel('Delay (ms)', fontsize=12, fontweight='bold') ax2.set_title('Component Delay Distribution', fontsize=14, fontweight='bold') ax2.grid(axis='y', alpha=0.3) @@ -211,32 +211,32 @@ def plot_delay_breakdown(delays: Dict[str, List[float]], per_agent_delays: Dict[ else: ax2.text(0.5, 0.5, 'No component delay data', ha='center', va='center', fontsize=12) ax2.set_title('Component Delay Distribution', fontsize=14, fontweight='bold') - + # Chart 3: Per-agent processing delays ax3 = plt.subplot(3, 3, 3) - + if per_agent_delays: agents = list(per_agent_delays.keys()) avg_delays = [np.mean(per_agent_delays[agent]) for agent in agents] - + colors = plt.cm.Set3(np.linspace(0, 1, len(agents))) ax3.bar(agents, avg_delays, color=colors, alpha=0.8, edgecolor='black', linewidth=1.5) ax3.set_ylabel('Average Processing Delay (ms)', fontsize=10, fontweight='bold') ax3.set_title('Processing Delay per Agent', fontsize=12, fontweight='bold') ax3.set_xticklabels(agents, rotation=45, ha='right', fontsize=8) ax3.grid(axis='y', alpha=0.3) - + for i, v in enumerate(avg_delays): if v > 0: ax3.text(i, v + max(avg_delays) * 0.01, f'{v:.1f}ms', ha='center', va='bottom', fontweight='bold', fontsize=8) - + # Chart 4: Inter-agent delays ax4 = plt.subplot(3, 3, 4) - + if inter_agent_delays: pairs = [f"{src}→{tgt}" for (src, tgt) in inter_agent_delays.keys()] avg_delays = [np.mean(inter_agent_delays[pair]) for pair in inter_agent_delays.keys()] - + if pairs: colors = plt.cm.Pastel1(np.linspace(0, 1, len(pairs))) ax4.barh(pairs, avg_delays, color=colors, alpha=0.8, edgecolor='black', linewidth=1.5) @@ -245,20 +245,20 @@ def plot_delay_breakdown(delays: Dict[str, List[float]], per_agent_delays: Dict[ ax4.set_yticks(range(len(pairs))) ax4.set_yticklabels(pairs, fontsize=8) ax4.grid(axis='x', alpha=0.3) - + for i, v in enumerate(avg_delays): if v > 0: ax4.text(v + max(avg_delays) * 0.01, i, f'{v:.1f}ms', ha='left', va='center', fontweight='bold', fontsize=8) else: ax4.axis('off') ax4.text(0.5, 0.5, 'No inter-agent delay data', ha='center', va='center', fontsize=10) - + # Chart 5: Detailed delay breakdown by agent and communication ax5 = plt.subplot(3, 3, 5) - + # Collect all detailed delay information breakdown_data = [] - + # Agent-LLM delays per agent for agent_name, llm_delays in agent_llm_delays.items(): if llm_delays: @@ -272,7 +272,7 @@ def plot_delay_breakdown(delays: Dict[str, List[float]], per_agent_delays: Dict[ 'type': 'agent_llm', 'color': '#e74c3c' }) - + # Inter-agent communication delays for (source, target), comm_delays in inter_agent_delays.items(): if comm_delays: @@ -286,7 +286,7 @@ def plot_delay_breakdown(delays: Dict[str, List[float]], per_agent_delays: Dict[ 'type': 'inter_agent', 'color': '#3498db' }) - + # Agent processing delays for agent_name, proc_delays in per_agent_delays.items(): if proc_delays: @@ -300,16 +300,16 @@ def plot_delay_breakdown(delays: Dict[str, List[float]], per_agent_delays: Dict[ 'type': 'processing', 'color': '#2ecc71' }) - + if breakdown_data: # Sort by total delay (descending) breakdown_data.sort(key=lambda x: x['total'], reverse=True) - + # Show all items labels = [d['label'] for d in breakdown_data] totals = [d['total'] for d in breakdown_data] colors = [d['color'] for d in breakdown_data] - + # Create horizontal bar chart y_pos = np.arange(len(labels)) ax5.barh(y_pos, totals, color=colors, alpha=0.8, edgecolor='black', linewidth=1) @@ -318,23 +318,23 @@ def plot_delay_breakdown(delays: Dict[str, List[float]], per_agent_delays: Dict[ ax5.set_xlabel('Total Delay (ms)', fontsize=10, fontweight='bold') ax5.set_title('Detailed Delay Breakdown\n(All Components)', fontsize=12, fontweight='bold') ax5.grid(axis='x', alpha=0.3) - + # Add value labels max_total = max(totals) if totals else 0 for i, (total, avg, count) in enumerate(zip(totals, [d['avg'] for d in breakdown_data], [d['count'] for d in breakdown_data])): - ax5.text(total + max_total * 0.01, i, f'{total:.0f}ms\n(avg: {avg:.0f}ms, n={count})', + ax5.text(total + max_total * 0.01, i, f'{total:.0f}ms\n(avg: {avg:.0f}ms, n={count})', ha='left', va='center', fontsize=7, fontweight='bold') else: ax5.axis('off') ax5.text(0.5, 0.5, 'No detailed delay data', ha='center', va='center', fontsize=10) - + # Chart 6: Delay statistics table ax6 = plt.subplot(3, 3, 6) ax6.axis('off') - + # Chart 5: Component delays comparison ax5 = plt.subplot(3, 3, 5) - + if delays['e2e_delay']: e2e_total = delays['e2e_delay'][0] stats_data = [ @@ -342,7 +342,7 @@ def plot_delay_breakdown(delays: Dict[str, List[float]], per_agent_delays: Dict[ ['Total E2E Delay', f'{e2e_total/1000:.2f} s ({e2e_total:.2f} ms)'], ['Total E2E Delay', f'{e2e_total/60000:.2f} minutes'], ] - + if delays['agent_llm_delay']: llm_total = sum(delays['agent_llm_delay']) stats_data.append(['Total LLM Delay', f'{llm_total/1000:.2f} s']) @@ -358,14 +358,14 @@ def plot_delay_breakdown(delays: Dict[str, List[float]], per_agent_delays: Dict[ stats_data.append(['Total Processing Delay', f'{proc_total/1000:.2f} s']) stats_data.append(['Processing Count', f'{len(delays["agent_processing_delay"])}']) stats_data.append(['Avg Processing Delay', f'{np.mean(delays["agent_processing_delay"]):.2f} ms']) - + table = ax6.table(cellText=stats_data[1:], colLabels=stats_data[0], cellLoc='center', loc='center', bbox=[0, 0, 1, 1]) table.auto_set_font_size(False) table.set_fontsize(8) table.scale(1, 1.5) ax6.set_title('Delay Statistics', fontsize=12, fontweight='bold', pad=10) - + # Chart 7-9: Per-agent delay details ax7 = plt.subplot(3, 3, 7) ax7.axis('off') @@ -374,28 +374,28 @@ def plot_delay_breakdown(delays: Dict[str, List[float]], per_agent_delays: Dict[ for agent in sorted(per_agent_delays.keys()): delays_list = per_agent_delays[agent] agent_stats.append([agent, f'{np.mean(delays_list):.2f}', f'{np.max(delays_list):.2f}']) - + table = ax7.table(cellText=agent_stats, colLabels=['Agent', 'Mean (ms)', 'Max (ms)'], cellLoc='center', loc='center', bbox=[0, 0, 1, 1]) table.auto_set_font_size(False) table.set_fontsize(8) table.scale(1, 1.5) ax7.set_title('Per-Agent Processing Delays', fontsize=12, fontweight='bold', pad=10) - + ax8 = plt.subplot(3, 3, 8) ax8.axis('off') if inter_agent_delays: comm_stats = [] for (src, tgt), delays_list in sorted(inter_agent_delays.items()): comm_stats.append([f'{src}→{tgt}', f'{np.mean(delays_list):.2f}', f'{np.max(delays_list):.2f}']) - + table = ax8.table(cellText=comm_stats[:10], colLabels=['Communication', 'Mean (ms)', 'Max (ms)'], cellLoc='center', loc='center', bbox=[0, 0, 1, 1]) table.auto_set_font_size(False) table.set_fontsize(7) table.scale(1, 1.2) ax8.set_title('Inter-Agent Delays (Top 10)', fontsize=12, fontweight='bold', pad=10) - + ax9 = plt.subplot(3, 3, 9) ax9.axis('off') note_text = "Note: Processing delay is calculated as invoke_agent duration minus child spans.\n" @@ -406,7 +406,7 @@ def plot_delay_breakdown(delays: Dict[str, List[float]], per_agent_delays: Dict[ ax9.text(0.1, 0.5, note_text, fontsize=9, verticalalignment='center', bbox=dict(boxstyle='round', facecolor='wheat', alpha=0.5)) ax9.set_title('Processing Delay Notes', fontsize=12, fontweight='bold', pad=10) - + plt.tight_layout() plt.savefig(output_file, dpi=300, bbox_inches='tight') print(f"✓ Delay breakdown chart saved to {output_file}") @@ -416,11 +416,11 @@ def plot_delay_breakdown(delays: Dict[str, List[float]], per_agent_delays: Dict[ def plot_cpu_memory_usage(usage: Dict, output_file: str): """Plot CPU and memory usage charts (process-level only).""" fig, ((ax1, ax2), (ax3, ax4)) = plt.subplots(2, 2, figsize=(16, 10)) - + # Chart 1: Process CPU usage over time # Check if we have per-agent data (distributed system) or combined data (monolithic) has_per_agent = 'per_agent' in usage.get('cpu', {}) and usage['cpu'].get('per_agent') - + if has_per_agent and usage['cpu']['per_agent']: # Distributed system: plot per-agent CPU usage colors = plt.cm.Set3(range(len(usage['cpu']['per_agent']))) @@ -431,7 +431,7 @@ def plot_cpu_memory_usage(usage: Dict, output_file: str): start_time = times[0] if times else 0 times = [t - start_time for t in times] ax1.plot(times, values, label=f'{agent_name} CPU (%)', color=colors[idx], linewidth=2, alpha=0.8) - + ax1.set_xlabel('Time (seconds)', fontsize=12, fontweight='bold') ax1.set_ylabel('CPU Usage (%)', fontsize=12, fontweight='bold') ax1.set_title('Process CPU Usage Over Time (Per-Agent)', fontsize=14, fontweight='bold') @@ -440,31 +440,31 @@ def plot_cpu_memory_usage(usage: Dict, output_file: str): elif usage['cpu']['process']: # Monolithic system: plot combined CPU usage times_proc, values_proc = zip(*usage['cpu']['process']) - + # Normalize times to start from 0 start_time = times_proc[0] if times_proc else 0 times_proc = [t - start_time for t in times_proc] - + ax1.plot(times_proc, values_proc, label='Process CPU (%)', color='#3498db', linewidth=2, alpha=0.8) ax1.set_xlabel('Time (seconds)', fontsize=12, fontweight='bold') ax1.set_ylabel('CPU Usage (%)', fontsize=12, fontweight='bold') ax1.set_title('Process CPU Usage Over Time', fontsize=14, fontweight='bold') ax1.legend() ax1.grid(alpha=0.3) - + # Check if all values are 0 if all(v == 0 for v in values_proc): - ax1.text(0.5, 0.95, 'Note: Process CPU is 0% - agents are lightweight\n(most time spent waiting for LLM I/O)', + ax1.text(0.5, 0.95, 'Note: Process CPU is 0% - agents are lightweight\n(most time spent waiting for LLM I/O)', transform=ax1.transAxes, ha='center', va='top', fontsize=9, bbox=dict(boxstyle='round', facecolor='yellow', alpha=0.3)) else: ax1.text(0.5, 0.5, 'No process CPU data available', ha='center', va='center', fontsize=12) ax1.set_title('Process CPU Usage Over Time', fontsize=14, fontweight='bold') - + # Chart 2: Process memory usage over time # Check if we have per-agent data (distributed system) or combined data (monolithic) has_per_agent_mem = 'per_agent' in usage.get('memory', {}) and usage['memory'].get('per_agent') - + if has_per_agent_mem and usage['memory']['per_agent']: # Distributed system: plot per-agent memory usage colors = plt.cm.Set3(range(len(usage['memory']['per_agent']))) @@ -475,7 +475,7 @@ def plot_cpu_memory_usage(usage: Dict, output_file: str): start_time = times[0] if times else 0 times = [t - start_time for t in times] ax2.plot(times, values, label=f'{agent_name} Memory (MB)', color=colors[idx], linewidth=2, alpha=0.8) - + ax2.set_xlabel('Time (seconds)', fontsize=12, fontweight='bold') ax2.set_ylabel('Memory Usage (MB)', fontsize=12, fontweight='bold') ax2.set_title('Process Memory Usage Over Time (Per-Agent)', fontsize=14, fontweight='bold') @@ -484,11 +484,11 @@ def plot_cpu_memory_usage(usage: Dict, output_file: str): elif usage['memory']['process']: # Monolithic system: plot combined memory usage times_proc, values_proc = zip(*usage['memory']['process']) - + # Normalize times to start from 0 start_time = times_proc[0] if times_proc else 0 times_proc = [t - start_time for t in times_proc] - + ax2.plot(times_proc, values_proc, label='Process Memory (MB)', color='#3498db', linewidth=2, alpha=0.8) ax2.set_xlabel('Time (seconds)', fontsize=12, fontweight='bold') ax2.set_ylabel('Memory Usage (MB)', fontsize=12, fontweight='bold') @@ -498,7 +498,7 @@ def plot_cpu_memory_usage(usage: Dict, output_file: str): else: ax2.text(0.5, 0.5, 'No process memory data available', ha='center', va='center', fontsize=12) ax2.set_title('Process Memory Usage Over Time', fontsize=14, fontweight='bold') - + # Chart 3: Process CPU usage statistics ax3.axis('off') cpu_stats = [] @@ -510,7 +510,7 @@ def plot_cpu_memory_usage(usage: Dict, output_file: str): cpu_stats.append(['Process CPU Min', f'{np.min(proc_values):.2f}%']) cpu_stats.append(['Process CPU Std', f'{np.std(proc_values):.2f}%']) cpu_stats.append(['Non-zero Count', f'{sum(1 for v in proc_values if v > 0)}/{len(proc_values)}']) - + if cpu_stats: table = ax3.table(cellText=cpu_stats, colLabels=['Metric', 'Value'], cellLoc='center', loc='center', bbox=[0, 0, 1, 1]) @@ -521,7 +521,7 @@ def plot_cpu_memory_usage(usage: Dict, output_file: str): else: ax3.text(0.5, 0.5, 'No process CPU statistics available', ha='center', va='center', fontsize=12) ax3.set_title('Process CPU Usage Statistics', fontsize=14, fontweight='bold', pad=20) - + # Chart 4: Process memory usage statistics ax4.axis('off') mem_stats = [] @@ -532,7 +532,7 @@ def plot_cpu_memory_usage(usage: Dict, output_file: str): mem_stats.append(['Process Memory Max', f'{np.max(proc_values):.2f} MB']) mem_stats.append(['Process Memory Min', f'{np.min(proc_values):.2f} MB']) mem_stats.append(['Process Memory Std', f'{np.std(proc_values):.2f} MB']) - + if mem_stats: table = ax4.table(cellText=mem_stats, colLabels=['Metric', 'Value'], cellLoc='center', loc='center', bbox=[0, 0, 1, 1]) @@ -543,7 +543,7 @@ def plot_cpu_memory_usage(usage: Dict, output_file: str): else: ax4.text(0.5, 0.5, 'No process memory statistics available', ha='center', va='center', fontsize=12) ax4.set_title('Process Memory Usage Statistics', fontsize=14, fontweight='bold', pad=20) - + plt.tight_layout() plt.savefig(output_file, dpi=300, bbox_inches='tight') print(f"✓ CPU/Memory usage chart saved to {output_file}") @@ -552,11 +552,11 @@ def plot_cpu_memory_usage(usage: Dict, output_file: str): def plot_message_sizes(sizes: Dict[str, List[float]], per_agent_sizes: Dict[str, Dict[str, List[float]]], - inter_agent_sizes: Dict[Tuple[str, str], Dict[str, List[float]]], + inter_agent_sizes: Dict[Tuple[str, str], Dict[str, List[float]]], agent_llm_sizes: Dict[str, Dict[str, List[float]]], output_file: str): """Plot message size charts with per-agent details.""" fig = plt.figure(figsize=(20, 14)) - + # Chart 1: Inter-agent message sizes ax1 = plt.subplot(3, 3, 1) if sizes['inter_agent_input'] or sizes['inter_agent_output']: @@ -564,24 +564,24 @@ def plot_message_sizes(sizes: Dict[str, List[float]], per_agent_sizes: Dict[str, input_avg = np.mean(sizes['inter_agent_input']) if sizes['inter_agent_input'] else 0 output_avg = np.mean(sizes['inter_agent_output']) if sizes['inter_agent_output'] else 0 values = [input_avg, output_avg] - + colors = ['#3498db', '#e74c3c'] ax1.bar(categories, values, color=colors, alpha=0.8, edgecolor='black', linewidth=1.5) ax1.set_ylabel('Average Message Size (KB)', fontsize=12, fontweight='bold') ax1.set_title('Inter-Agent Message Sizes', fontsize=14, fontweight='bold') ax1.grid(axis='y', alpha=0.3) - + # Add value labels for i, v in enumerate(values): if v > 0: - ax1.text(i, v + max(values) * 0.01 if max(values) > 0 else 0.1, f'{v:.2f} KB', + ax1.text(i, v + max(values) * 0.01 if max(values) > 0 else 0.1, f'{v:.2f} KB', ha='center', va='bottom', fontweight='bold') else: ax1.axis('off') - ax1.text(0.5, 0.5, 'No inter-agent message size data\n(In-process calls have no message size)', + ax1.text(0.5, 0.5, 'No inter-agent message size data\n(In-process calls have no message size)', ha='center', va='center', fontsize=10, style='italic') ax1.set_title('Inter-Agent Message Sizes', fontsize=14, fontweight='bold') - + # Chart 2: Agent-LLM message sizes (per agent) ax2 = plt.subplot(3, 3, 2) if agent_llm_sizes: @@ -594,7 +594,7 @@ def plot_message_sizes(sizes: Dict[str, List[float]], per_agent_sizes: Dict[str, agents.append(agent) input_avgs.append(np.mean(agent_data['input']) if agent_data['input'] else 0) output_avgs.append(np.mean(agent_data['output']) if agent_data['output'] else 0) - + if agents: x = np.arange(len(agents)) width = 0.35 @@ -613,19 +613,19 @@ def plot_message_sizes(sizes: Dict[str, List[float]], per_agent_sizes: Dict[str, input_avg = np.mean(sizes['agent_llm_input']) if sizes['agent_llm_input'] else 0 output_avg = np.mean(sizes['agent_llm_output']) if sizes['agent_llm_output'] else 0 values = [input_avg, output_avg] - + colors = ['#3498db', '#e74c3c'] ax2.bar(categories, values, color=colors, alpha=0.8, edgecolor='black', linewidth=1.5) ax2.set_ylabel('Average Message Size (KB)', fontsize=12, fontweight='bold') ax2.set_title('Agent-LLM Message Sizes', fontsize=14, fontweight='bold') ax2.grid(axis='y', alpha=0.3) - + # Add value labels for i, v in enumerate(values): if v > 0: - ax2.text(i, v + max(values) * 0.01 if max(values) > 0 else 0.1, f'{v:.2f} KB', + ax2.text(i, v + max(values) * 0.01 if max(values) > 0 else 0.1, f'{v:.2f} KB', ha='center', va='bottom', fontweight='bold') - + # Chart 3: Per-agent message sizes (input) ax3 = plt.subplot(3, 3, 3) if per_agent_sizes: @@ -638,7 +638,7 @@ def plot_message_sizes(sizes: Dict[str, List[float]], per_agent_sizes: Dict[str, agents.append(agent) input_avgs.append(np.mean(agent_data['input']) if agent_data['input'] else 0) output_avgs.append(np.mean(agent_data['output']) if agent_data['output'] else 0) - + if agents: x = np.arange(len(agents)) width = 0.35 @@ -659,7 +659,7 @@ def plot_message_sizes(sizes: Dict[str, List[float]], per_agent_sizes: Dict[str, ax3.axis('off') ax3.text(0.5, 0.5, 'No per-agent message size data', ha='center', va='center', fontsize=10, style='italic') ax3.set_title('Message Size per Agent', fontsize=12, fontweight='bold') - + # Chart 4: Inter-agent message sizes detail ax4 = plt.subplot(3, 3, 4) if inter_agent_sizes: @@ -672,7 +672,7 @@ def plot_message_sizes(sizes: Dict[str, List[float]], per_agent_sizes: Dict[str, pairs.append(f'{src}→{tgt}') input_avgs.append(np.mean(pair_data['input']) if pair_data['input'] else 0) output_avgs.append(np.mean(pair_data['output']) if pair_data['output'] else 0) - + if pairs: x = np.arange(len(pairs)) width = 0.35 @@ -687,22 +687,22 @@ def plot_message_sizes(sizes: Dict[str, List[float]], per_agent_sizes: Dict[str, ax4.grid(axis='y', alpha=0.3) else: ax4.axis('off') - ax4.text(0.5, 0.5, 'No inter-agent pair data\n(In-process calls have no message size)', + ax4.text(0.5, 0.5, 'No inter-agent pair data\n(In-process calls have no message size)', ha='center', va='center', fontsize=10, style='italic') ax4.set_title('Message Size by Communication Pair', fontsize=12, fontweight='bold') else: ax4.axis('off') - ax4.text(0.5, 0.5, 'No inter-agent pair data\n(In-process calls have no message size)', + ax4.text(0.5, 0.5, 'No inter-agent pair data\n(In-process calls have no message size)', ha='center', va='center', fontsize=10, style='italic') ax4.set_title('Message Size by Communication Pair', fontsize=12, fontweight='bold') - + # Chart 5: Message size distribution (inter-agent) ax5 = plt.subplot(3, 3, 5) if sizes['inter_agent_input'] or sizes['inter_agent_output']: all_sizes = sizes['inter_agent_input'] + sizes['inter_agent_output'] if all_sizes: ax5.hist(all_sizes, bins=20, color='#3498db', alpha=0.7, edgecolor='black') - ax5.axvline(np.mean(all_sizes), color='red', linestyle='--', linewidth=2, + ax5.axvline(np.mean(all_sizes), color='red', linestyle='--', linewidth=2, label=f'Mean: {np.mean(all_sizes):.2f} KB') ax5.set_xlabel('Message Size (KB)', fontsize=10, fontweight='bold') ax5.set_ylabel('Frequency', fontsize=10, fontweight='bold') @@ -712,12 +712,12 @@ def plot_message_sizes(sizes: Dict[str, List[float]], per_agent_sizes: Dict[str, else: ax5.axis('off') ax5.text(0.5, 0.5, 'No inter-agent message size data', ha='center', va='center', fontsize=10) - + # Chart 6: Message size statistics ax6 = plt.subplot(3, 3, 6) ax6.axis('off') stats_data = [['Metric', 'Value']] - + if sizes['inter_agent_input']: stats_data.append(['Inter-Agent Input Mean', f'{np.mean(sizes["inter_agent_input"]):.2f} KB']) stats_data.append(['Inter-Agent Input Total', f'{np.sum(sizes["inter_agent_input"]):.2f} KB']) @@ -730,7 +730,7 @@ def plot_message_sizes(sizes: Dict[str, List[float]], per_agent_sizes: Dict[str, if sizes['agent_llm_output']: stats_data.append(['Agent-LLM Output Mean', f'{np.mean(sizes["agent_llm_output"]):.2f} KB']) stats_data.append(['Agent-LLM Output Total', f'{np.sum(sizes["agent_llm_output"]):.2f} KB']) - + if len(stats_data) > 1: table = ax6.table(cellText=stats_data[1:], colLabels=stats_data[0], cellLoc='center', loc='center', bbox=[0, 0, 1, 1]) @@ -738,7 +738,7 @@ def plot_message_sizes(sizes: Dict[str, List[float]], per_agent_sizes: Dict[str, table.set_fontsize(8) table.scale(1, 1.5) ax6.set_title('Message Size Statistics', fontsize=12, fontweight='bold', pad=10) - + # Chart 7-9: Additional details ax7 = plt.subplot(3, 3, 7) ax7.axis('off') @@ -750,7 +750,7 @@ def plot_message_sizes(sizes: Dict[str, List[float]], per_agent_sizes: Dict[str, output_total = np.sum(agent_data['output']) if agent_data['output'] else 0 if input_total > 0 or output_total > 0: agent_stats.append([agent, f'{input_total:.2f}', f'{output_total:.2f}']) - + if agent_stats: table = ax7.table(cellText=agent_stats, colLabels=['Agent', 'Input Total (KB)', 'Output Total (KB)'], cellLoc='center', loc='center', bbox=[0, 0, 1, 1]) @@ -758,7 +758,7 @@ def plot_message_sizes(sizes: Dict[str, List[float]], per_agent_sizes: Dict[str, table.set_fontsize(8) table.scale(1, 1.5) ax7.set_title('Per-Agent Message Totals', fontsize=12, fontweight='bold', pad=10) - + ax8 = plt.subplot(3, 3, 8) ax8.axis('off') if inter_agent_sizes: @@ -769,7 +769,7 @@ def plot_message_sizes(sizes: Dict[str, List[float]], per_agent_sizes: Dict[str, output_total = np.sum(pair_data['output']) if pair_data['output'] else 0 if input_total > 0 or output_total > 0: comm_stats.append([f'{src}→{tgt}', f'{input_total:.2f}', f'{output_total:.2f}']) - + if comm_stats: table = ax8.table(cellText=comm_stats[:10], colLabels=['Pair', 'Input Total (KB)', 'Output Total (KB)'], cellLoc='center', loc='center', bbox=[0, 0, 1, 1]) @@ -777,7 +777,7 @@ def plot_message_sizes(sizes: Dict[str, List[float]], per_agent_sizes: Dict[str, table.set_fontsize(7) table.scale(1, 1.2) ax8.set_title('Inter-Agent Totals (Top 10)', fontsize=12, fontweight='bold', pad=10) - + ax9 = plt.subplot(3, 3, 9) ax9.axis('off') note_text = "Note: Message sizes are calculated from actual data:\n" @@ -787,7 +787,7 @@ def plot_message_sizes(sizes: Dict[str, List[float]], per_agent_sizes: Dict[str, ax9.text(0.1, 0.5, note_text, fontsize=9, verticalalignment='center', bbox=dict(boxstyle='round', facecolor='wheat', alpha=0.5)) ax9.set_title('Measurement Notes', fontsize=12, fontweight='bold', pad=10) - + plt.tight_layout() plt.savefig(output_file, dpi=300, bbox_inches='tight') print(f"✓ Message size chart saved to {output_file}") @@ -796,18 +796,18 @@ def plot_message_sizes(sizes: Dict[str, List[float]], per_agent_sizes: Dict[str, def plot_events_with_cpu_memory(traces: List[Dict], cpu_memory_usage: Dict, output_file: str): """Plot spans/events timeline (Gantt chart) with CPU/memory usage below. - + Creates a two-panel figure: - Top: Gantt chart showing spans (or events if available) and their durations - Bottom: CPU and memory usage over time - Both panels share the same time axis with vertical lines connecting spans/events to resource usage - + Args: traces: List of trace spans (used to extract events or spans for Gantt chart) cpu_memory_usage: Dict from extract_cpu_memory_usage() output_file: Path to save the figure """ - + # Extract events from spans events = [] for span in traces: @@ -816,7 +816,7 @@ def plot_events_with_cpu_memory(traces: List[Dict], cpu_memory_usage: Dict, outp span_name = span.get('name', 'unknown') span_start = span.get('start_time', 0) span_end = span.get('end_time', 0) - + # Handle both integer (nanoseconds) and ISO string timestamps if isinstance(span_start, str): try: @@ -832,10 +832,10 @@ def plot_events_with_cpu_memory(traces: List[Dict], cpu_memory_usage: Dict, outp span_end = int(dt.timestamp() * 1e9) except (ValueError, AttributeError): span_end = 0 - + attrs = span.get('attributes', {}) agent_name = attrs.get('gen_ai.agent.name') or span.get('agent_name', 'unknown') - + for event in span_events: event_timestamp = event.get('timestamp', 0) # Handle both integer (nanoseconds) and ISO string timestamps @@ -846,7 +846,7 @@ def plot_events_with_cpu_memory(traces: List[Dict], cpu_memory_usage: Dict, outp event_timestamp = int(dt.timestamp() * 1e9) except (ValueError, AttributeError): event_timestamp = 0 - + events.append({ 'name': event.get('name', 'unknown'), 'timestamp': event_timestamp, @@ -856,7 +856,7 @@ def plot_events_with_cpu_memory(traces: List[Dict], cpu_memory_usage: Dict, outp 'agent_name': agent_name, 'attributes': event.get('attributes', {}) }) - + # If no events or very few events, use spans as Gantt chart items # Use spans if we have less than 5 events (spans are more informative for visualization) use_spans_as_events = len(events) < 5 @@ -864,14 +864,14 @@ def plot_events_with_cpu_memory(traces: List[Dict], cpu_memory_usage: Dict, outp print(f" Using spans mode: {len(events)} events found, using {len(traces)} spans instead") else: print(f" Using events mode: {len(events)} events found") - + if use_spans_as_events: # Create span-based timeline spans_for_gantt = [] for span in traces: span_start = span.get('start_time', 0) span_end = span.get('end_time', 0) - + # Handle both integer (nanoseconds) and ISO string timestamps if isinstance(span_start, str): # ISO format string, convert to nanoseconds @@ -889,7 +889,7 @@ def plot_events_with_cpu_memory(traces: List[Dict], cpu_memory_usage: Dict, outp span_end = int(dt.timestamp() * 1e9) except (ValueError, AttributeError): span_end = 0 - + if span_start > 0 and span_end > 0: attrs = span.get('attributes', {}) agent_name = attrs.get('gen_ai.agent.name') or span.get('agent_name', 'unknown') @@ -900,7 +900,7 @@ def plot_events_with_cpu_memory(traces: List[Dict], cpu_memory_usage: Dict, outp 'agent_name': agent_name, 'span_id': span.get('span_id', '') }) - + if not spans_for_gantt: print("⚠️ No spans or events data to plot") return @@ -910,7 +910,7 @@ def plot_events_with_cpu_memory(traces: List[Dict], cpu_memory_usage: Dict, outp if spans_for_gantt: first_span = spans_for_gantt[0] print(f" First span: {first_span['name']}, start={first_span['start']}, end={first_span['end']}") - + # Check if we have events data (for events mode) # Even without CPU/memory data, we should still plot events/spans timeline if not use_spans_as_events: @@ -920,17 +920,17 @@ def plot_events_with_cpu_memory(traces: List[Dict], cpu_memory_usage: Dict, outp # Warn if no CPU/memory data, but still proceed to plot events timeline if not cpu_memory_usage.get('cpu', {}).get('process'): print(" ⚠️ Warning: No CPU/memory data available, will plot events timeline only") - + fig = plt.figure(figsize=(20, 12)) - + # Determine time range and normalize to relative time (start from 0) cpu_data = cpu_memory_usage.get('cpu', {}).get('process', []) mem_data = cpu_memory_usage.get('memory', {}).get('process', []) - + # Get base time from the earliest data base_time = None all_times = [] - + if use_spans_as_events: span_start_times = [s['start'] / 1e9 for s in spans_for_gantt] span_end_times = [s['end'] / 1e9 for s in spans_for_gantt] @@ -946,21 +946,21 @@ def plot_events_with_cpu_memory(traces: List[Dict], cpu_memory_usage: Dict, outp span_starts = [e['span_start'] / 1e9 for e in events if e['span_start'] > 0] span_ends = [e['span_end'] / 1e9 for e in events if e['span_end'] > 0] all_times.extend(event_times + trace_start_times + trace_end_times + span_starts + span_ends) - + if cpu_data: cpu_times = [t for t, _ in cpu_data] all_times.extend(cpu_times) if mem_data: mem_times = [t for t, _ in mem_data] all_times.extend(mem_times) - + if not all_times: print("⚠️ No time data available") return - + base_time = min(all_times) max_time = max(all_times) - base_time - + # Debug: Print time ranges for troubleshooting if traces: first_trace_start = traces[0].get('start_time', 0) @@ -981,15 +981,15 @@ def plot_events_with_cpu_memory(traces: List[Dict], cpu_memory_usage: Dict, outp cpu_times_norm = [(t - base_time) for t, _ in cpu_data[:3]] print(f" First 3 CPU times (normalized): {[f'{t:.2f}' for t in cpu_times_norm]}") print(f" Base time: {base_time:.2f}, Max time: {max_time:.2f}") - - - + + + # Create two subplots with shared x-axis # Use sharex to ensure both panels have synchronized x-axis gs = fig.add_gridspec(2, 1, height_ratios=[1, 1], hspace=0.15) ax1 = fig.add_subplot(gs[0]) # Top: Spans/Events Gantt chart ax2 = fig.add_subplot(gs[1], sharex=ax1) # Bottom: CPU/Memory (shares x-axis with ax1) - + # ========== Top Panel: Spans/Events Gantt Chart ========== if use_spans_as_events: # Group spans by agent or operation type for better visualization @@ -1000,38 +1000,38 @@ def plot_events_with_cpu_memory(traces: List[Dict], cpu_memory_usage: Dict, outp if group_key not in span_groups: span_groups[group_key] = [] span_groups[group_key].append(span_info) - + # Sort groups by first span start time sorted_groups = sorted(span_groups.items(), key=lambda x: min(s['start'] / 1e9 for s in x[1])) - + # Create Gantt chart y_positions = {} y_pos = 0 colors = plt.cm.Set3(np.linspace(0, 1, len(sorted_groups))) - + drawn_spans = 0 skipped_spans = 0 - + for i, (group_name, group_spans) in enumerate(sorted_groups): # Sort spans in group by start time group_spans.sort(key=lambda x: x['start']) - + for span_info in group_spans: if group_name not in y_positions: y_positions[group_name] = y_pos y_pos += 1 - + y = y_positions[group_name] start = (span_info['start'] / 1e9) - base_time end = (span_info['end'] / 1e9) - base_time duration = end - start - + if duration > 0: # Draw span bar - ax1.barh(y, duration, left=start, height=0.6, color=colors[i], + ax1.barh(y, duration, left=start, height=0.6, color=colors[i], alpha=0.7, edgecolor='black', linewidth=1) drawn_spans += 1 - + # Add span name label (truncate if too long) span_name = span_info['name'] if len(span_name) > 50: @@ -1042,16 +1042,16 @@ def plot_events_with_cpu_memory(traces: List[Dict], cpu_memory_usage: Dict, outp color='black', bbox=dict(boxstyle='round,pad=0.3', facecolor='white', alpha=0.7)) else: skipped_spans += 1 - + if drawn_spans == 0: print(f" ⚠️ Warning: No spans were drawn! (skipped {skipped_spans} spans with duration <= 0)") else: print(f" ✓ Drew {drawn_spans} spans in {len(sorted_groups)} groups (skipped {skipped_spans} spans)") - + ax1.set_ylabel('Span/Agent', fontsize=12, fontweight='bold') ax1.set_title(f'Spans Timeline (Gantt Chart) - {len(spans_for_gantt)} spans', fontsize=14, fontweight='bold') ax1.set_yticks(list(range(len(y_positions)))) - ax1.set_yticklabels([name[:40] + '...' if len(name) > 40 else name + ax1.set_yticklabels([name[:40] + '...' if len(name) > 40 else name for name in y_positions.keys()], fontsize=8) ax1.grid(axis='x', alpha=0.3) # Don't set xlim here - set it after both panels are drawn to ensure synchronization @@ -1065,7 +1065,7 @@ def plot_events_with_cpu_memory(traces: List[Dict], cpu_memory_usage: Dict, outp span_start = event['span_start'] span_end = event['span_end'] event_timestamp = event['timestamp'] - + if isinstance(span_start, str): try: from datetime import datetime @@ -1087,7 +1087,7 @@ def plot_events_with_cpu_memory(traces: List[Dict], cpu_memory_usage: Dict, outp event_timestamp = int(dt.timestamp() * 1e9) except (ValueError, AttributeError): event_timestamp = 0 - + if span_name not in span_data: span_data[span_name] = { 'start': (span_start / 1e9) - base_time, @@ -1100,79 +1100,79 @@ def plot_events_with_cpu_memory(traces: List[Dict], cpu_memory_usage: Dict, outp 'timestamp': (event_timestamp / 1e9) - base_time, 'attributes': event['attributes'] }) - + sorted_spans = sorted(span_data.items(), key=lambda x: x[1]['start']) y_positions = {} y_pos = 0 colors = plt.cm.Set3(np.linspace(0, 1, len(sorted_spans))) - + drawn_spans = 0 drawn_events = 0 - + for i, (span_name, span_info) in enumerate(sorted_spans): if span_name not in y_positions: y_positions[span_name] = y_pos y_pos += 1 - + y = y_positions[span_name] start = span_info['start'] end = span_info['end'] duration = end - start - + if duration > 0: - ax1.barh(y, duration, left=start, height=0.6, color=colors[i], + ax1.barh(y, duration, left=start, height=0.6, color=colors[i], alpha=0.7, edgecolor='black', linewidth=1) drawn_spans += 1 - + for event in span_info['events']: event_time = event['timestamp'] - ax1.axvline(x=event_time, color='gray', linestyle='--', + ax1.axvline(x=event_time, color='gray', linestyle='--', linewidth=1.5, alpha=0.6, zorder=1) ax1.plot(event_time, y, 'ro', markersize=8, alpha=0.9, zorder=2) drawn_events += 1 - + print(f" ✓ Drew {drawn_spans} span bars and {drawn_events} event markers") if drawn_spans == 0 and drawn_events == 0: print(f" ⚠️ Warning: No spans or events were drawn! Check time ranges.") if span_data: first_span = list(span_data.values())[0] print(f" First span start: {first_span['start']}, end: {first_span['end']}, base_time: {base_time}") - + ax1.set_ylabel('Span/Event', fontsize=12, fontweight='bold') ax1.set_title(f'Events Timeline (Gantt Chart) - {len(events)} events', fontsize=14, fontweight='bold') ax1.set_yticks(list(range(len(y_positions)))) - ax1.set_yticklabels([name[:40] + '...' if len(name) > 40 else name + ax1.set_yticklabels([name[:40] + '...' if len(name) > 40 else name for name in y_positions.keys()], fontsize=8) ax1.grid(axis='x', alpha=0.3) # Don't set xlim here - set it after both panels are drawn to ensure synchronization ax1.invert_yaxis() - + # ========== Bottom Panel: CPU/Memory Usage ========== # Initialize variables for xlim calculation cpu_times_list = [] mem_times_list = [] - + # Check if we have CPU/memory data has_cpu_memory_data = cpu_data or mem_data - + if has_cpu_memory_data: # Normalize times cpu_times_norm = [(t - base_time, v) for t, v in cpu_data] mem_times_norm = [(t - base_time, v) for t, v in mem_data] - + # Plot CPU usage if cpu_times_norm: cpu_times, cpu_values = zip(*cpu_times_norm) cpu_times_list = list(cpu_times) # Store for xlim calculation ax2.plot(cpu_times, cpu_values, 'b-', linewidth=2, label='CPU Usage (%)', alpha=0.8) - + # Plot Memory usage (on secondary y-axis) ax2_mem = ax2.twinx() if mem_times_norm: mem_times, mem_values = zip(*mem_times_norm) mem_times_list = list(mem_times) # Store for xlim calculation ax2_mem.plot(mem_times, mem_values, 'r-', linewidth=2, label='Memory Usage (MB)', alpha=0.8) - + # Draw vertical lines from spans/events to resource usage # These lines connect the top panel (spans/events) to the bottom panel (CPU/memory) if use_spans_as_events: @@ -1180,23 +1180,23 @@ def plot_events_with_cpu_memory(traces: List[Dict], cpu_memory_usage: Dict, outp start = (span_info['start'] / 1e9) - base_time end = (span_info['end'] / 1e9) - base_time # Draw lines at span start and end to connect to CPU/memory plot - ax2.axvline(x=start, color='gray', linestyle='--', + ax2.axvline(x=start, color='gray', linestyle='--', linewidth=1, alpha=0.4, zorder=0) - ax2.axvline(x=end, color='gray', linestyle='--', + ax2.axvline(x=end, color='gray', linestyle='--', linewidth=1, alpha=0.4, zorder=0) # Also draw on top panel for visual connection - ax1.axvline(x=start, color='gray', linestyle='--', + ax1.axvline(x=start, color='gray', linestyle='--', linewidth=1, alpha=0.4, zorder=0) - ax1.axvline(x=end, color='gray', linestyle='--', + ax1.axvline(x=end, color='gray', linestyle='--', linewidth=1, alpha=0.4, zorder=0) elif events: for event in events: event_time = (event['timestamp'] / 1e9) - base_time - ax2.axvline(x=event_time, color='gray', linestyle='--', + ax2.axvline(x=event_time, color='gray', linestyle='--', linewidth=1.5, alpha=0.6, zorder=0) - ax1.axvline(x=event_time, color='gray', linestyle='--', + ax1.axvline(x=event_time, color='gray', linestyle='--', linewidth=1.5, alpha=0.6, zorder=0) - + ax2.set_xlabel('Time (seconds, relative)', fontsize=12, fontweight='bold') ax2.set_ylabel('CPU Usage (%)', fontsize=12, fontweight='bold', color='blue') ax2.tick_params(axis='y', labelcolor='blue') @@ -1207,35 +1207,35 @@ def plot_events_with_cpu_memory(traces: List[Dict], cpu_memory_usage: Dict, outp ax2.grid(axis='x', alpha=0.3) else: # No CPU/memory data, but still show the panel with a message - ax2.text(0.5, 0.5, 'No CPU/Memory data available', - ha='center', va='center', fontsize=14, + ax2.text(0.5, 0.5, 'No CPU/Memory data available', + ha='center', va='center', fontsize=14, transform=ax2.transAxes, style='italic', color='gray') ax2.set_xlabel('Time (seconds, relative)', fontsize=12, fontweight='bold') ax2.set_title('CPU and Memory Usage Over Time (No Data)', fontsize=14, fontweight='bold') ax2.grid(axis='x', alpha=0.3) - + # Still draw vertical lines from spans/events even without CPU/memory data # This helps visualize the timeline alignment if use_spans_as_events: for span_info in spans_for_gantt: start = (span_info['start'] / 1e9) - base_time end = (span_info['end'] / 1e9) - base_time - ax2.axvline(x=start, color='gray', linestyle='--', + ax2.axvline(x=start, color='gray', linestyle='--', linewidth=1, alpha=0.4, zorder=0) - ax2.axvline(x=end, color='gray', linestyle='--', + ax2.axvline(x=end, color='gray', linestyle='--', linewidth=1, alpha=0.4, zorder=0) - ax1.axvline(x=start, color='gray', linestyle='--', + ax1.axvline(x=start, color='gray', linestyle='--', linewidth=1, alpha=0.4, zorder=0) - ax1.axvline(x=end, color='gray', linestyle='--', + ax1.axvline(x=end, color='gray', linestyle='--', linewidth=1, alpha=0.4, zorder=0) elif events: for event in events: event_time = (event['timestamp'] / 1e9) - base_time - ax2.axvline(x=event_time, color='gray', linestyle='--', + ax2.axvline(x=event_time, color='gray', linestyle='--', linewidth=1.5, alpha=0.6, zorder=0) - ax1.axvline(x=event_time, color='gray', linestyle='--', + ax1.axvline(x=event_time, color='gray', linestyle='--', linewidth=1.5, alpha=0.6, zorder=0) - + # Set x-axis limits for both panels AFTER all drawing is complete # This ensures perfect synchronization when using sharex # Calculate the actual max time from all drawn elements @@ -1254,19 +1254,19 @@ def plot_events_with_cpu_memory(traces: List[Dict], cpu_memory_usage: Dict, outp if has_cpu_memory_data: all_drawn_times.extend(cpu_times_list) all_drawn_times.extend(mem_times_list) - + if all_drawn_times: actual_max_time = max(all_drawn_times) # Add small padding (1%) actual_max_time = actual_max_time * 1.01 else: actual_max_time = max_time - + # Set limits on both axes to ensure perfect alignment # This must be done after all drawing to ensure sharex works correctly ax1.set_xlim(0, actual_max_time) ax2.set_xlim(0, actual_max_time) - + # Combine legends (only if we have memory data) if has_cpu_memory_data and mem_times_norm: lines1, labels1 = ax2.get_legend_handles_labels() @@ -1275,7 +1275,7 @@ def plot_events_with_cpu_memory(traces: List[Dict], cpu_memory_usage: Dict, outp elif has_cpu_memory_data: ax2.legend(loc='upper left', fontsize=10) ax2.set_xlim(0, max_time) - + plt.subplots_adjust(hspace=0.15) plt.savefig(output_file, dpi=300, bbox_inches='tight') print(f"✓ Events/spans with CPU/memory chart saved to {output_file}") @@ -1284,7 +1284,7 @@ def plot_events_with_cpu_memory(traces: List[Dict], cpu_memory_usage: Dict, outp def plot_latency_flame(traces: List[Dict], output_file: str, title: str = "Request Latency Flame Graph"): """Render a span-based aggregated (latency) flame graph. - + This is a request/latency-style flame graph (not CPU). Width = aggregated span duration across all traces; depth = parent/child span nesting. Good for seeing where time goes. """ diff --git a/plot/plot_cpu_mem_by_example_summary.py b/plot/plot_cpu_mem_by_example_summary.py index 0e79def..00b770d 100644 --- a/plot/plot_cpu_mem_by_example_summary.py +++ b/plot/plot_cpu_mem_by_example_summary.py @@ -329,7 +329,7 @@ def main() -> int: return 0 plot_width = max(7, len(labels) * 0.32) - plot_height = 7 + plot_height = 7 output_path.parent.mkdir(parents=True, exist_ok=True) fig, (ax_cpu, ax_mem) = plt.subplots( diff --git a/plot/plot_example_metrics.py b/plot/plot_example_metrics.py index f6ddc43..41dab26 100644 --- a/plot/plot_example_metrics.py +++ b/plot/plot_example_metrics.py @@ -38,7 +38,7 @@ def load_config(config_file: Path) -> dict: """Load configuration from JSON file.""" if not config_file.exists(): return {} - + try: with open(config_file, 'r', encoding='utf-8') as f: config = json.load(f) @@ -61,7 +61,7 @@ def save_config(config: dict, config_file: Path): def parse_examples_config(config_file: Path) -> Dict[str, Dict[str, str]]: """Parse examples configuration from JSON file. - + Expected format: { "example_name": { @@ -74,7 +74,7 @@ def parse_examples_config(config_file: Path) -> Dict[str, Dict[str, str]]: """ if not config_file.exists(): raise FileNotFoundError(f"Examples config file not found: {config_file}") - + try: with open(config_file, 'r', encoding='utf-8') as f: config = json.load(f) @@ -137,7 +137,7 @@ def plot_tree_of_thoughts_model_similarity() -> None: str(tot_root / "metrics" / "gpt-5o-nano"), ), ] - + collector = MultiExampleCollector() for model_name, traces_dir, metrics_dir in model_dirs: print(f"Loading model traces: {model_name}...") @@ -154,14 +154,14 @@ def plot_tree_of_thoughts_model_similarity() -> None: except Exception as e: print(f" ✗ Error loading {model_name}: {e}") continue - + if not collector.examples: print("Error: No model data loaded for tree-of-thoughts") return - + output_dir = _script_dir / 'figures' / 'comparison' output_dir.mkdir(parents=True, exist_ok=True) - + visualizer = CrossExampleVisualizer(collector) loaded_models = [name for name, _, _ in model_dirs if name in collector.examples] output_file = output_dir / 'call_graph_similarity_by_model.pdf' @@ -180,7 +180,7 @@ def process_example( project_root: Path ) -> None: """Process a single example and generate visualizations. - + Args: example_name: Name of the example example_config: Configuration dict with traces_dir, metrics_dir, base_dir @@ -193,21 +193,21 @@ def process_example( base_dir = project_root / base_dir if not Path(base_dir).is_absolute() else Path(base_dir) else: base_dir = project_root - + traces_dir = example_config.get('traces_dir', 'traces') metrics_dir = example_config.get('metrics_dir', 'metrics') - + # Resolve traces_dir and metrics_dir if not Path(traces_dir).is_absolute(): traces_dir = str(base_dir / traces_dir) if not Path(metrics_dir).is_absolute(): metrics_dir = str(base_dir / metrics_dir) - + # Output directory: plot/figures/ output_dir = _script_dir / 'figures' / example_name # Ensure output directory and all parent directories exist output_dir.mkdir(parents=True, exist_ok=True) - + print("=" * 80) print(f"Processing example: {example_name}") print("=" * 80) @@ -217,7 +217,7 @@ def process_example( print(f"Metrics directory: {metrics_dir}") print(f"Output directory: {output_dir}") print() - + dataset_example_name = example_config.get("dataset_example_name") tags = example_config.get("tags") if isinstance(tags, str): @@ -232,7 +232,7 @@ def process_example( example_name=dataset_example_name or example_name, tags=tags, ) - + # Run analysis analyzer.analyze(output_dir=str(output_dir)) print(f"\n✓ Completed visualization for {example_name}") @@ -253,13 +253,13 @@ def main(): python plot_example_metrics.py --example brand-search-optimization \\ --example-traces-dir examples/adk/brand-search-optimization/traces \\ --example-metrics-dir examples/adk/brand-search-optimization/metrics - + # Process multiple examples from config file python plot_example_metrics.py --examples-config examples_config.json --mode per_run - + # Run latest + per_run in one command python plot_example_metrics.py --examples-config examples_config.json --mode all - + # Process single example with base directory python plot_example_metrics.py --example marketing-agency \\ --example-base-dir examples/adk/marketing-agency \\ @@ -269,18 +269,18 @@ def main(): Legacy Mode (Backward Compatible): # Analyze latest run (quick analysis) python plot_example_metrics.py --mode latest - + # Analyze all runs with statistics (recommended for academic papers) python plot_example_metrics.py --mode per_run - + # Use custom directories python plot_example_metrics.py --traces-dir custom_traces --metrics-dir custom_metrics - + # Use configuration file python plot_example_metrics.py --config config.json """ ) - + # Unified mode arguments (new) parser.add_argument( '--example', @@ -317,11 +317,11 @@ def main(): action='store_true', help='Generate tree-of-thoughts call graph similarity violins across models (hardcoded paths).' ) - + # Legacy mode arguments (backward compatible) parser.add_argument( - '--mode', - choices=['latest', 'per_run', 'all'], + '--mode', + choices=['latest', 'per_run', 'all'], default=None, help='Analysis mode: latest (quick analysis), per_run (per-run stats), or all (run latest + per_run). Default: per_run.' ) @@ -361,7 +361,7 @@ def main(): default=None, help='Save current command-line arguments to a configuration file for future use. (Legacy mode)' ) - + args = parser.parse_args() if args.tree_of_thoughts_model_similarity: @@ -376,15 +376,15 @@ def main(): raise SystemExit( f"Default examples config not found: {DEFAULT_EXAMPLES_CONFIG}" ) - + # Determine mode: unified or legacy is_unified_mode = not legacy_mode - + if is_unified_mode: # Unified mode: process examples mode = args.mode or 'all' modes_to_run = ['latest', 'per_run'] if mode == 'all' else [mode] - + if args.examples_config: # Process multiple examples from config file config_file = Path(args.examples_config) @@ -392,11 +392,11 @@ def main(): config_file = Path.cwd() / config_file if not config_file.exists(): config_file = _project_root / args.examples_config - + examples_config = parse_examples_config(config_file) print(f"Loaded examples configuration from {config_file}") print(f"Found {len(examples_config)} example(s) to process\n") - + # Process each example separately (supports running multiple modes) for example_name, example_config in examples_config.items(): for run_mode in modes_to_run: @@ -409,23 +409,23 @@ def main(): print("=" * 80) print("All examples processed!") print("=" * 80) - + elif args.example: # Process single example if not args.example_traces_dir or not args.example_metrics_dir: parser.error("--example requires --example-traces-dir and --example-metrics-dir, or use --examples-config") - + example_config = { 'traces_dir': args.example_traces_dir, 'metrics_dir': args.example_metrics_dir, 'base_dir': args.example_base_dir, } - + for run_mode in modes_to_run: process_example(args.example, example_config, run_mode, _project_root) else: parser.error("Must specify either --example or --examples-config for unified mode") - + else: # Legacy mode: backward compatible behavior # Load configuration from file if specified @@ -440,7 +440,7 @@ def main(): config = load_config(config_file) if config: print(f"Loaded configuration from {config_file}") - + # Merge config with command-line arguments (CLI args take precedence) mode = args.mode or config.get('mode', 'all') modes_to_run = ['latest', 'per_run'] if mode == 'all' else [mode] @@ -448,13 +448,13 @@ def main(): metrics_dir = args.metrics_dir or config.get('metrics_dir') base_dir = args.base_dir or config.get('base_dir') output_dir = args.output_dir or config.get('output_dir', 'visualizations') - + # Save configuration if requested if args.save_config: save_config_file = Path(args.save_config) if not save_config_file.is_absolute(): save_config_file = Path.cwd() / save_config_file - + config_to_save = { 'mode': mode, 'traces_dir': traces_dir, @@ -465,7 +465,7 @@ def main(): # Remove None values config_to_save = {k: v for k, v in config_to_save.items() if v is not None} save_config(config_to_save, save_config_file) - + print("=" * 80) print("OpenTelemetry Metrics Visualization (Legacy Mode)") print("=" * 80) @@ -478,7 +478,7 @@ def main(): print(f"Base directory: {base_dir}") print(f"Output directory: {output_dir}") print() - + # Create analyzer with specified parameters for run_mode in modes_to_run: analyzer = MetricsAnalyzer( @@ -487,7 +487,7 @@ def main(): base_dir=base_dir, analysis_mode=run_mode ) - + # Run analysis analyzer.analyze(output_dir=output_dir) diff --git a/pyproject.toml b/pyproject.toml new file mode 100644 index 0000000..e3d778f --- /dev/null +++ b/pyproject.toml @@ -0,0 +1,13 @@ +[project] +name = "maestro" +version = "0.1.0" +description = "LLM-based Multi-Agent System (MAS) evaluation suite" +readme = "README.md" +requires-python = ">=3.10" +dependencies = [] + +[dependency-groups] +dev = ["pre-commit>=4.5.1"] + +[tool.uv] +required-version = ">=0.9.0" diff --git a/src/ADK_otel/README.md b/src/ADK_otel/README.md index 0f843c6..e065261 100644 --- a/src/ADK_otel/README.md +++ b/src/ADK_otel/README.md @@ -1,3 +1,3 @@ # ADK OTEL Helpers -This package centralizes the telemetry logic shared across every ADK benchmark. See examples in `examples/adk` for usage patterns. \ No newline at end of file +This package centralizes the telemetry logic shared across every ADK benchmark. See examples in `examples/adk` for usage patterns. diff --git a/src/ADK_otel/telemetry_setup.py b/src/ADK_otel/telemetry_setup.py index c02645b..f83c04a 100644 --- a/src/ADK_otel/telemetry_setup.py +++ b/src/ADK_otel/telemetry_setup.py @@ -32,10 +32,10 @@ class JsonFileMetricExporter(MetricExporter): """Exporter that writes metrics to local JSON files.""" - + def __init__(self, file_path: str): """Initialize JSON file exporter. - + Args: file_path: Path to JSON file """ @@ -45,54 +45,54 @@ def __init__(self, file_path: str): # Use None to accept defaults (same as ConsoleMetricExporter) self._preferred_temporality = None self._preferred_aggregation = None - + def export(self, metrics_data, timeout_millis: float = 10000, **kwargs): """Export metrics to JSON file. - + Args: metrics_data: Metrics data to export timeout_millis: Maximum time to wait for export (not used for file export) **kwargs: Additional arguments - + Returns: MetricExportResult.SUCCESS """ try: # Convert metrics to serializable format metric_records = [] - + # Extract metric data from the metrics_data object if hasattr(metrics_data, 'resource_metrics'): for resource_metric in metrics_data.resource_metrics: resource_attrs = dict(resource_metric.resource.attributes) if resource_metric.resource and resource_metric.resource.attributes else {} - + for scope_metric in resource_metric.scope_metrics: scope_name = scope_metric.scope.name if scope_metric.scope else "unknown" - + for metric in scope_metric.metrics: metric_name = metric.name metric_description = metric.description if hasattr(metric, 'description') else None metric_unit = metric.unit if hasattr(metric, 'unit') else None - + # Extract data points data_points = [] if hasattr(metric, 'data') and hasattr(metric.data, 'data_points'): for data_point in metric.data.data_points: point_attrs = dict(data_point.attributes) if hasattr(data_point, 'attributes') and data_point.attributes else {} - + # Add service name as agent identifier if not present if "agent.name" not in point_attrs: # Try to infer from resource attributes or use service name agent_name = resource_attrs.get("service.name", "brand-search-optimization") point_attrs["agent.name"] = agent_name - + point_dict = { "value": data_point.value if hasattr(data_point, 'value') else None, "timestamp": data_point.time_unix_nano if hasattr(data_point, 'time_unix_nano') else None, "attributes": point_attrs, } data_points.append(point_dict) - + metric_record = { "timestamp": datetime.now().isoformat(), "metric_name": metric_name, @@ -105,31 +105,31 @@ def export(self, metrics_data, timeout_millis: float = 10000, **kwargs): "scope": scope_name, } metric_records.append(metric_record) - + # Append to file (supports incremental writes) if metric_records: with open(self.file_path, "a", encoding="utf-8") as f: for record in metric_records: f.write(json.dumps(record, default=str) + "\n") - + return MetricExportResult.SUCCESS except Exception as e: logger.error(f"Failed to export metrics to {self.file_path}: {e}", exc_info=True) return MetricExportResult.FAILURE - + def force_flush(self, timeout_millis: int = 30000, **kwargs): """Force flush any pending metrics. - + Args: timeout_millis: Maximum time to wait for flush. **kwargs: Additional arguments. """ # For file exporter, data is written immediately, so no explicit flush needed pass - + def shutdown(self, timeout_millis: int = 30000, **kwargs): """Shutdown the exporter. - + Args: timeout_millis: Maximum time to wait for shutdown. **kwargs: Additional arguments. @@ -139,23 +139,23 @@ def shutdown(self, timeout_millis: int = 30000, **kwargs): class JsonFileSpanExporter(SpanExporter): """Exporter that writes spans to local JSON files.""" - + def __init__(self, file_path: str): """Initialize JSON file exporter. - + Args: file_path: Path to JSON file """ self.file_path = Path(file_path) self.file_path.parent.mkdir(parents=True, exist_ok=True) self.spans = [] - + def export(self, spans): """Export spans to JSON file. - + Args: spans: List of spans to export - + Returns: SpanExportResult.SUCCESS """ @@ -173,7 +173,7 @@ def export(self, spans): parent_span_id = format(span.parent.span_id, "016x") elif hasattr(span, "parent_context") and span.parent_context: parent_span_id = format(span.parent_context.span_id, "016x") - + # Extract agent name from span attributes or name for better identification attributes = dict(span.attributes) if span.attributes else {} agent_name = ( @@ -182,11 +182,11 @@ def export(self, spans): (span.name.split()[-1] if "invoke_agent" in span.name and len(span.name.split()) > 1 else None) or "unknown" ) - + # Add agent_name to attributes if not already present (for backward compatibility) if "agent.name" not in attributes and agent_name != "unknown": attributes["agent.name"] = agent_name - + # Ensure gen_ai.agent.name is present (required by template) # For call_llm spans, try to extract from gcp.vertex.agent.llm_request labels if "gen_ai.agent.name" not in attributes: @@ -205,7 +205,7 @@ def export(self, spans): agent_name = adk_agent_name except (json.JSONDecodeError, KeyError, TypeError): pass - + # If still not found, use extracted agent_name or agent.name if "gen_ai.agent.name" not in attributes: if agent_name != "unknown": @@ -215,7 +215,7 @@ def export(self, spans): else: # Field must exist per template, set to null if not available attributes["gen_ai.agent.name"] = None - + # Ensure LLM usage fields are present (default to 0 if missing) if "gen_ai.usage.input_tokens" not in attributes: attributes["gen_ai.usage.input_tokens"] = 0 @@ -226,14 +226,14 @@ def export(self, spans): input_tokens = attributes.get("gen_ai.usage.input_tokens", 0) or 0 output_tokens = attributes.get("gen_ai.usage.output_tokens", 0) or 0 attributes["gen_ai.usage.total_tokens"] = input_tokens + output_tokens - + # Add LLM/MCP call counts (default to 0 if missing) if "gen_ai.llm.call.count" not in attributes: # Set to 1 if this is a call_llm span, otherwise 0 attributes["gen_ai.llm.call.count"] = 1 if "call_llm" in span.name else 0 if "gen_ai.mcp.call.count" not in attributes: attributes["gen_ai.mcp.call.count"] = 0 - + # Ensure gen_ai.operation.name is present (required by template) if "gen_ai.operation.name" not in attributes: if "call_llm" in span.name: @@ -245,7 +245,7 @@ def export(self, spans): elif span.name.lower() == "invocation": # invocation spans are top-level entry points attributes["gen_ai.operation.name"] = "invocation" - + # Ensure gen_ai.conversation.id is present (required by template) # Copy from session_id if available, otherwise set to null if "gen_ai.conversation.id" not in attributes: @@ -255,14 +255,14 @@ def export(self, spans): else: # Field must exist per template, set to null if not available attributes["gen_ai.conversation.id"] = None - + # Add missing optional fields with default values (per template) # gen_ai.agent.description - optional, only in invoke_agent spans if "gen_ai.agent.description" not in attributes: # Try to extract from system_instruction or other sources # For now, set to empty string if not available attributes["gen_ai.agent.description"] = "" - + # gen_ai.tool.* fields - only for execute_tool spans if "gen_ai.tool.name" not in attributes: attributes["gen_ai.tool.name"] = "" @@ -272,31 +272,31 @@ def export(self, spans): attributes["gen_ai.tool.call.id"] = "" if "gen_ai.tool.description" not in attributes: attributes["gen_ai.tool.description"] = "" - + # gcp.vertex.agent.tool_* fields - optional if "gcp.vertex.agent.tool_call_args" not in attributes: attributes["gcp.vertex.agent.tool_call_args"] = "" if "gcp.vertex.agent.tool_response" not in attributes: attributes["gcp.vertex.agent.tool_response"] = "" - + # mcp.* fields - optional, not applicable for Vertex AI if "mcp.server" not in attributes: attributes["mcp.server"] = "" if "mcp.tool" not in attributes: attributes["mcp.tool"] = "" - + # agent.log - optional if "agent.log" not in attributes: attributes["agent.log"] = "" - + # Note: CPU/memory metrics are collected separately in metrics JSONL file, # so we don't add them to span attributes (per template guidelines) - + # Calculate communication metrics for agent-to-agent calls communication_metrics = {} input_size = 0 output_size = 0 - + # Check if this is agent-to-agent communication # Only AgentTool calls are agent communication, not FunctionTool calls is_agent_comm = False @@ -313,53 +313,53 @@ def export(self, spans): elif "invoke_agent" in span.name: # invoke_agent spans are entry points, not agent-to-agent communication is_agent_comm = False - + if is_agent_comm: communication_metrics["is_agent_communication"] = True - + # For execute_tool spans (AgentTool): measure actual tool call arguments and responses if "execute_tool" in span.name: # Input: tool call arguments (actual data passed to the agent) tool_call_args = attributes.get("gcp.vertex.agent.tool_call_args", "") if tool_call_args and isinstance(tool_call_args, str) and tool_call_args != "{}": input_size = len(tool_call_args.encode('utf-8')) - + # Output: tool response (actual data returned from the agent) tool_response = attributes.get("gcp.vertex.agent.tool_response", "") if tool_response and isinstance(tool_response, str) and tool_response != "{}": output_size = len(tool_response.encode('utf-8')) - + # For call_llm spans: measure actual LLM request and response elif "call_llm" in span.name: # Input: LLM request (actual data sent to LLM) llm_request = attributes.get("gcp.vertex.agent.llm_request", "") if llm_request and isinstance(llm_request, str) and llm_request != "{}": input_size = len(llm_request.encode('utf-8')) - + # Output: LLM response (actual data received from LLM) llm_response = attributes.get("gcp.vertex.agent.llm_response", "") if llm_response and isinstance(llm_response, str) and llm_response != "{}": output_size = len(llm_response.encode('utf-8')) - + # Store communication metrics and add to attributes for template compliance # Template shows these fields, but per README: "If a field doesn't apply, omit it" # So we only add them when there's actual communication data # Ensure is_in_process_call is always present (default false) if "is_in_process_call" not in communication_metrics: communication_metrics["is_in_process_call"] = False - + if input_size > 0: communication_metrics["input_message_size_bytes"] = input_size attributes["communication.input_message_size_bytes"] = input_size - + if output_size > 0: communication_metrics["output_message_size_bytes"] = output_size attributes["communication.output_message_size_bytes"] = output_size - + if input_size > 0 or output_size > 0: communication_metrics["total_message_size_bytes"] = input_size + output_size attributes["communication.total_message_size_bytes"] = input_size + output_size - + # Get span kind span_kind = "INTERNAL" # Default if hasattr(span, "kind"): @@ -368,7 +368,7 @@ def export(self, spans): span_kind = SpanKind(kind_value).name if kind_value is not None else "INTERNAL" except (ValueError, AttributeError): span_kind = "INTERNAL" - + # Get resource attributes and add host.name if available resource_attrs = dict(span.resource.attributes) if span.resource and span.resource.attributes else {} if "host.name" not in resource_attrs: @@ -378,7 +378,7 @@ def export(self, spans): resource_attrs["host.name"] = hostname except Exception: pass # host.name is optional - + span_dict = { "trace_id": format(span.context.trace_id, "032x"), "span_id": format(span.context.span_id, "016x"), @@ -408,7 +408,7 @@ def export(self, spans): }, } span_data.append(span_dict) - + # Append to file (supports incremental writes) if span_data: with open(self.file_path, "a", encoding="utf-8") as f: @@ -417,12 +417,12 @@ def export(self, spans): logger.info(f"Exported {len(span_data)} spans to {self.file_path}") else: logger.warning(f"No spans to export (received {len(spans)} spans, but span_data is empty)") - + return SpanExportResult.SUCCESS except Exception as e: logger.error(f"Failed to export spans to {self.file_path}: {e}", exc_info=True) return SpanExportResult.FAILURE - + def shutdown(self): """Shutdown the exporter.""" pass @@ -430,17 +430,17 @@ def shutdown(self): def setup_tracing(service_name: Optional[str] = None, trace_file: Optional[str] = None): """Setup OpenTelemetry tracing with local JSON file export. - + This function configures OpenTelemetry to automatically collect telemetry data from ADK agents. The instrumentation is automatic - ADK and OpenTelemetry instrumentation libraries will automatically create spans for: - + - Agent invocations (invoke_agent) - LLM calls (call_llm) with model, tokens, and request/response data - Tool/function executions (execute_tool) - A2A server operations (request handling, event queue operations) - HTTP requests between agents - + The collected data includes: - Trace IDs and Span IDs for distributed tracing - Timing information (start time, end time, duration) @@ -448,7 +448,7 @@ def setup_tracing(service_name: Optional[str] = None, trace_file: Optional[str] - Attributes (gen_ai.system, gen_ai.request.model, gen_ai.usage.*, etc.) - Events (exceptions, custom events) - Resource attributes (service.name, service.version, etc.) - + Args: service_name: Service name, defaults to OTEL_SERVICE_NAME env var or "adk-agent" trace_file: Trace file path, defaults to OTEL_TRACE_FILE env var or auto-generated @@ -457,28 +457,28 @@ def setup_tracing(service_name: Optional[str] = None, trace_file: Optional[str] # Get configuration from environment variables service_name = service_name or os.getenv("OTEL_SERVICE_NAME", "adk-agent") trace_file = trace_file or os.getenv("OTEL_TRACE_FILE") - + if not trace_file: # Default file path: traces directory in current working directory trace_dir = Path.cwd() / "traces" trace_dir.mkdir(exist_ok=True) timestamp = os.getenv("OTEL_RUN_TIMESTAMP") or datetime.now().strftime("%Y%m%d_%H%M%S") trace_file = str(trace_dir / f"{service_name}_{timestamp}.jsonl") - + logger.info(f"Setting up OpenTelemetry tracing for {service_name}") logger.info(f"Trace file: {trace_file}") - + # Create resource resource = Resource.create({ "service.name": service_name, "service.version": os.getenv("OTEL_SERVICE_VERSION", "1.0.0"), "deployment.environment": os.getenv("OTEL_DEPLOYMENT_ENVIRONMENT", "local"), }) - + # Check if tracer provider already exists current_provider = trace.get_tracer_provider() provider_already_set = isinstance(current_provider, TracerProvider) - + # If current provider is not a TracerProvider instance, create a new one if not provider_already_set: tracer_provider = TracerProvider(resource=resource) @@ -490,7 +490,7 @@ def setup_tracing(service_name: Optional[str] = None, trace_file: Optional[str] # For now, we'll just use the existing provider as-is tracer_provider = current_provider logger.debug(f"Using existing TracerProvider (resource attributes cannot be updated)") - + # Create and add JSON file exporter json_exporter = JsonFileSpanExporter(trace_file) # Use SimpleSpanProcessor to export spans immediately when they end @@ -498,10 +498,10 @@ def setup_tracing(service_name: Optional[str] = None, trace_file: Optional[str] span_processor = SimpleSpanProcessor(json_exporter) tracer_provider.add_span_processor(span_processor) logger.info(f"Using SimpleSpanProcessor for trace export to {trace_file}") - + # Ensure tracer provider is set trace.set_tracer_provider(tracer_provider) - + # Set environment variables to enable instrumentation libraries os.environ.setdefault("OTEL_PYTHON_LOGGING_AUTO_INSTRUMENTATION_ENABLED", "true") # Enable full message content capture for GenAI instrumentation @@ -512,7 +512,7 @@ def setup_tracing(service_name: Optional[str] = None, trace_file: Optional[str] os.environ.setdefault("OTEL_INSTRUMENTATION_HTTP_CAPTURE_HEADERS_SERVER_RESPONSE", ".*") # Increase max attribute length to capture full messages (default is 250) os.environ.setdefault("OTEL_ATTRIBUTE_VALUE_LENGTH_LIMIT", "10000") - + # Try to enable ADK's telemetry (only succeeds if OTLP endpoint is set) # Even if it fails, instrumentation libraries will use our tracer provider otel_endpoint = os.getenv("OTEL_EXPORTER_OTLP_TRACES_ENDPOINT") or os.getenv("OTEL_EXPORTER_OTLP_ENDPOINT") @@ -523,11 +523,11 @@ def setup_tracing(service_name: Optional[str] = None, trace_file: Optional[str] logger.info(f"ADK OTLP exporter enabled for {otel_endpoint}") except Exception as e: logger.warning(f"Failed to enable ADK OTLP exporter: {e}") - + logger.info("Tracer provider configured with JSON file exporter") - + logger.info("OpenTelemetry tracing setup completed") - + except Exception as e: logger.error(f"Failed to setup tracing: {e}", exc_info=True) # Don't raise exception, allow application to continue (tracing is optional) @@ -535,67 +535,67 @@ def setup_tracing(service_name: Optional[str] = None, trace_file: Optional[str] def setup_metrics(service_name: Optional[str] = None, metrics_file: Optional[str] = None, enable_system_metrics: bool = True): """Setup OpenTelemetry metrics with system resource monitoring and local JSON file export. - + This function configures OpenTelemetry Metrics to collect: - CPU usage (percentage) - Memory usage (percentage and absolute values) - Process-specific metrics - + Metrics are exported to local JSON files, similar to trace export. - + Note: In containerized environments, psutil will report container-level metrics, not host-level metrics. For host-level metrics, use cadvisor. - + Args: service_name: Service name for resource attributes metrics_file: Metrics file path, defaults to OTEL_METRICS_FILE env var or auto-generated enable_system_metrics: Whether to enable CPU/memory monitoring (default: True) """ global _metrics_initialized, _metrics_thread - + if _metrics_initialized: logger.warning("Metrics already initialized, skipping") return - + try: service_name = service_name or os.getenv("OTEL_SERVICE_NAME", "adk-agent") metrics_file = metrics_file or os.getenv("OTEL_METRICS_FILE") - + if not metrics_file: # Default file path: metrics directory in current working directory metrics_dir = Path.cwd() / "metrics" metrics_dir.mkdir(exist_ok=True) timestamp = os.getenv("OTEL_RUN_TIMESTAMP") or datetime.now().strftime("%Y%m%d_%H%M%S") metrics_file = str(metrics_dir / f"{service_name}_{timestamp}.jsonl") - + logger.info(f"Setting up OpenTelemetry metrics for {service_name}") logger.info(f"Metrics file: {metrics_file}") - + # Create resource resource = Resource.create({ "service.name": service_name, "service.version": os.getenv("OTEL_SERVICE_VERSION", "1.0.0"), "deployment.environment": os.getenv("OTEL_DEPLOYMENT_ENVIRONMENT", "local"), }) - + # Create JSON file exporter json_exporter = JsonFileMetricExporter(metrics_file) - + # Create metric reader with periodic export (every 1 seconds) reader = PeriodicExportingMetricReader( exporter=json_exporter, export_interval_millis=1000, # Export every 1 seconds ) - + # Create meter provider meter_provider = MeterProvider( resource=resource, metric_readers=[reader], ) metrics.set_meter_provider(meter_provider) - + meter = metrics.get_meter(__name__) - + # Create custom metrics for agent communication # These will be updated via span processor hooks communication_counter = meter.create_counter( @@ -603,25 +603,25 @@ def setup_metrics(service_name: Optional[str] = None, metrics_file: Optional[str description="Number of agent-to-agent communications", unit="1", ) - + communication_duration = meter.create_histogram( name="agent.communication.duration", description="Duration of agent-to-agent communications", unit="ms", ) - + communication_message_size = meter.create_histogram( name="agent.communication.message_size", description="Size of messages passed between agents", unit="bytes", ) - + logger.info("Custom communication metrics created") - + if enable_system_metrics: try: import psutil - + # Process-specific metrics only (system metrics removed per user request) process_cpu_gauge = meter.create_observable_gauge( name="process.cpu.usage", @@ -629,26 +629,26 @@ def setup_metrics(service_name: Optional[str] = None, metrics_file: Optional[str unit="%", callbacks=[_get_process_cpu_usage], ) - + process_memory_gauge = meter.create_observable_gauge( name="process.memory.usage_bytes", description="Process memory usage in bytes", unit="bytes", callbacks=[_get_process_memory_usage], ) - + logger.info("Process metrics (CPU/memory) enabled using psutil") logger.info("Note: System-level metrics removed per user request") - + except ImportError: logger.warning("psutil not installed, skipping system metrics. Install with: pip install psutil") except Exception as e: logger.warning(f"Failed to setup system metrics: {e}") - + _metrics_initialized = True logger.info("Meter provider configured with JSON file exporter") logger.info("OpenTelemetry metrics setup completed") - + except Exception as e: logger.error(f"Failed to setup metrics: {e}", exc_info=True) # Don't raise exception, allow application to continue (metrics are optional) @@ -677,7 +677,7 @@ def _get_memory_usage_bytes(callback_options): def _get_process_cpu_usage(callback_options): """Callback to get process CPU usage percentage. - + Note: psutil.Process().cpu_percent(interval=None) requires two calls to calculate. First call returns 0.0, subsequent calls return the percentage since last call. We use interval=0.1 to get immediate accurate reading. @@ -686,13 +686,13 @@ def _get_process_cpu_usage(callback_options): try: import psutil from opentelemetry.metrics import Observation - + if _process_obj is None: _process_obj = psutil.Process() # First call to initialize, returns 0.0 _process_obj.cpu_percent(interval=0.5) return [Observation(0.0)] - + # Subsequent calls return actual CPU usage # Use interval=0.1 for more accurate reading cpu_percent = _process_obj.cpu_percent(interval=0.5) @@ -712,4 +712,3 @@ def _get_process_memory_usage(callback_options): return [Observation(memory_info.rss)] # Resident Set Size except Exception: return [] - diff --git a/src/template/otel_metrics_template_for_human_reading.json b/src/template/otel_metrics_template_for_human_reading.json index d2b96d8..f0339d4 100644 --- a/src/template/otel_metrics_template_for_human_reading.json +++ b/src/template/otel_metrics_template_for_human_reading.json @@ -51,4 +51,4 @@ }, "scope": "" } -] \ No newline at end of file +] diff --git a/tools/.gitignore b/tools/.gitignore index 6baa653..4ddaa38 100644 --- a/tools/.gitignore +++ b/tools/.gitignore @@ -21,4 +21,4 @@ telemetry_failure_plots.py telemetry_judge_heatmap_facet.py telemetry_model_boxplots.py telemetry_tavily_diff.py -verify_parquet_conversion.py \ No newline at end of file +verify_parquet_conversion.py diff --git a/tools/fake_collector/storage/traces_2025-12-16T22-01-24.370201.json b/tools/fake_collector/storage/traces_2025-12-16T22-01-24.370201.json index 44e1cab..5b142de 100644 --- a/tools/fake_collector/storage/traces_2025-12-16T22-01-24.370201.json +++ b/tools/fake_collector/storage/traces_2025-12-16T22-01-24.370201.json @@ -1767,4 +1767,4 @@ "parsed_successfully": true } ] -} \ No newline at end of file +} diff --git a/tools/fake_collector/template.json b/tools/fake_collector/template.json index eac5e56..0ab5419 100644 --- a/tools/fake_collector/template.json +++ b/tools/fake_collector/template.json @@ -72,4 +72,4 @@ } } } -] \ No newline at end of file +] diff --git a/tools/pricing.json b/tools/pricing.json index b182b55..f28b34d 100644 --- a/tools/pricing.json +++ b/tools/pricing.json @@ -10,4 +10,4 @@ "vertex_ai/gemini-2.5-flash-lite": { "input_per_1m": 0.10, "output_per_1m": 0.40 }, "vertex_ai/gemini-2.5-flash": { "input_per_1m": 0.30, "output_per_1m": 2.50 }, "vertex_ai/gemini-embedding-001": { "input_per_1m": 0.15, "output_per_1m": 0.00 } -} \ No newline at end of file +} diff --git a/uv.lock b/uv.lock new file mode 100644 index 0000000..fd1ad1c --- /dev/null +++ b/uv.lock @@ -0,0 +1,176 @@ +version = 1 +revision = 3 +requires-python = ">=3.10" + +[[package]] +name = "cfgv" +version = "3.5.0" +source = { registry = "https://pypi.org/simple" } +sdist = { url = "https://files.pythonhosted.org/packages/4e/b5/721b8799b04bf9afe054a3899c6cf4e880fcf8563cc71c15610242490a0c/cfgv-3.5.0.tar.gz", hash = "sha256:d5b1034354820651caa73ede66a6294d6e95c1b00acc5e9b098e917404669132", size = 7334, upload-time = "2025-11-19T20:55:51.612Z" } +wheels = [ + { url = "https://files.pythonhosted.org/packages/db/3c/33bac158f8ab7f89b2e59426d5fe2e4f63f7ed25df84c036890172b412b5/cfgv-3.5.0-py2.py3-none-any.whl", hash = "sha256:a8dc6b26ad22ff227d2634a65cb388215ce6cc96bbcc5cfde7641ae87e8dacc0", size = 7445, upload-time = "2025-11-19T20:55:50.744Z" }, +] + +[[package]] +name = "distlib" +version = "0.4.0" +source = { registry = "https://pypi.org/simple" } +sdist = { url = "https://files.pythonhosted.org/packages/96/8e/709914eb2b5749865801041647dc7f4e6d00b549cfe88b65ca192995f07c/distlib-0.4.0.tar.gz", hash = "sha256:feec40075be03a04501a973d81f633735b4b69f98b05450592310c0f401a4e0d", size = 614605, upload-time = "2025-07-17T16:52:00.465Z" } +wheels = [ + { url = "https://files.pythonhosted.org/packages/33/6b/e0547afaf41bf2c42e52430072fa5658766e3d65bd4b03a563d1b6336f57/distlib-0.4.0-py2.py3-none-any.whl", hash = "sha256:9659f7d87e46584a30b5780e43ac7a2143098441670ff0a49d5f9034c54a6c16", size = 469047, upload-time = "2025-07-17T16:51:58.613Z" }, +] + +[[package]] +name = "filelock" +version = "3.20.3" +source = { registry = "https://pypi.org/simple" } +sdist = { url = "https://files.pythonhosted.org/packages/1d/65/ce7f1b70157833bf3cb851b556a37d4547ceafc158aa9b34b36782f23696/filelock-3.20.3.tar.gz", hash = "sha256:18c57ee915c7ec61cff0ecf7f0f869936c7c30191bb0cf406f1341778d0834e1", size = 19485, upload-time = "2026-01-09T17:55:05.421Z" } +wheels = [ + { url = "https://files.pythonhosted.org/packages/b5/36/7fb70f04bf00bc646cd5bb45aa9eddb15e19437a28b8fb2b4a5249fac770/filelock-3.20.3-py3-none-any.whl", hash = "sha256:4b0dda527ee31078689fc205ec4f1c1bf7d56cf88b6dc9426c4f230e46c2dce1", size = 16701, upload-time = "2026-01-09T17:55:04.334Z" }, +] + +[[package]] +name = "identify" +version = "2.6.16" +source = { registry = "https://pypi.org/simple" } +sdist = { url = "https://files.pythonhosted.org/packages/5b/8d/e8b97e6bd3fb6fb271346f7981362f1e04d6a7463abd0de79e1fda17c067/identify-2.6.16.tar.gz", hash = "sha256:846857203b5511bbe94d5a352a48ef2359532bc8f6727b5544077a0dcfb24980", size = 99360, upload-time = "2026-01-12T18:58:58.201Z" } +wheels = [ + { url = "https://files.pythonhosted.org/packages/b8/58/40fbbcefeda82364720eba5cf2270f98496bdfa19ea75b4cccae79c698e6/identify-2.6.16-py2.py3-none-any.whl", hash = "sha256:391ee4d77741d994189522896270b787aed8670389bfd60f326d677d64a6dfb0", size = 99202, upload-time = "2026-01-12T18:58:56.627Z" }, +] + +[[package]] +name = "maestro" +version = "0.1.0" +source = { virtual = "." } + +[package.dev-dependencies] +dev = [ + { name = "pre-commit" }, +] + +[package.metadata] + +[package.metadata.requires-dev] +dev = [{ name = "pre-commit", specifier = ">=4.5.1" }] + +[[package]] +name = "nodeenv" +version = "1.10.0" +source = { registry = "https://pypi.org/simple" } +sdist = { url = "https://files.pythonhosted.org/packages/24/bf/d1bda4f6168e0b2e9e5958945e01910052158313224ada5ce1fb2e1113b8/nodeenv-1.10.0.tar.gz", hash = "sha256:996c191ad80897d076bdfba80a41994c2b47c68e224c542b48feba42ba00f8bb", size = 55611, upload-time = "2025-12-20T14:08:54.006Z" } +wheels = [ + { url = "https://files.pythonhosted.org/packages/88/b2/d0896bdcdc8d28a7fc5717c305f1a861c26e18c05047949fb371034d98bd/nodeenv-1.10.0-py2.py3-none-any.whl", hash = "sha256:5bb13e3eed2923615535339b3c620e76779af4cb4c6a90deccc9e36b274d3827", size = 23438, upload-time = "2025-12-20T14:08:52.782Z" }, +] + +[[package]] +name = "platformdirs" +version = "4.5.1" +source = { registry = "https://pypi.org/simple" } +sdist = { url = "https://files.pythonhosted.org/packages/cf/86/0248f086a84f01b37aaec0fa567b397df1a119f73c16f6c7a9aac73ea309/platformdirs-4.5.1.tar.gz", hash = "sha256:61d5cdcc6065745cdd94f0f878977f8de9437be93de97c1c12f853c9c0cdcbda", size = 21715, upload-time = "2025-12-05T13:52:58.638Z" } +wheels = [ + { url = "https://files.pythonhosted.org/packages/cb/28/3bfe2fa5a7b9c46fe7e13c97bda14c895fb10fa2ebf1d0abb90e0cea7ee1/platformdirs-4.5.1-py3-none-any.whl", hash = "sha256:d03afa3963c806a9bed9d5125c8f4cb2fdaf74a55ab60e5d59b3fde758104d31", size = 18731, upload-time = "2025-12-05T13:52:56.823Z" }, +] + +[[package]] +name = "pre-commit" +version = "4.5.1" +source = { registry = "https://pypi.org/simple" } +dependencies = [ + { name = "cfgv" }, + { name = "identify" }, + { name = "nodeenv" }, + { name = "pyyaml" }, + { name = "virtualenv" }, +] +sdist = { url = "https://files.pythonhosted.org/packages/40/f1/6d86a29246dfd2e9b6237f0b5823717f60cad94d47ddc26afa916d21f525/pre_commit-4.5.1.tar.gz", hash = "sha256:eb545fcff725875197837263e977ea257a402056661f09dae08e4b149b030a61", size = 198232, upload-time = "2025-12-16T21:14:33.552Z" } +wheels = [ + { url = "https://files.pythonhosted.org/packages/5d/19/fd3ef348460c80af7bb4669ea7926651d1f95c23ff2df18b9d24bab4f3fa/pre_commit-4.5.1-py2.py3-none-any.whl", hash = "sha256:3b3afd891e97337708c1674210f8eba659b52a38ea5f822ff142d10786221f77", size = 226437, upload-time = "2025-12-16T21:14:32.409Z" }, +] + +[[package]] +name = "pyyaml" +version = "6.0.3" +source = { registry = "https://pypi.org/simple" } +sdist = { url = "https://files.pythonhosted.org/packages/05/8e/961c0007c59b8dd7729d542c61a4d537767a59645b82a0b521206e1e25c2/pyyaml-6.0.3.tar.gz", hash = "sha256:d76623373421df22fb4cf8817020cbb7ef15c725b9d5e45f17e189bfc384190f", size = 130960, upload-time = "2025-09-25T21:33:16.546Z" } +wheels = [ + { url = "https://files.pythonhosted.org/packages/f4/a0/39350dd17dd6d6c6507025c0e53aef67a9293a6d37d3511f23ea510d5800/pyyaml-6.0.3-cp310-cp310-macosx_10_13_x86_64.whl", hash = "sha256:214ed4befebe12df36bcc8bc2b64b396ca31be9304b8f59e25c11cf94a4c033b", size = 184227, upload-time = "2025-09-25T21:31:46.04Z" }, + { url = "https://files.pythonhosted.org/packages/05/14/52d505b5c59ce73244f59c7a50ecf47093ce4765f116cdb98286a71eeca2/pyyaml-6.0.3-cp310-cp310-macosx_11_0_arm64.whl", hash = "sha256:02ea2dfa234451bbb8772601d7b8e426c2bfa197136796224e50e35a78777956", size = 174019, upload-time = "2025-09-25T21:31:47.706Z" }, + { url = "https://files.pythonhosted.org/packages/43/f7/0e6a5ae5599c838c696adb4e6330a59f463265bfa1e116cfd1fbb0abaaae/pyyaml-6.0.3-cp310-cp310-manylinux2014_aarch64.manylinux_2_17_aarch64.manylinux_2_28_aarch64.whl", hash = "sha256:b30236e45cf30d2b8e7b3e85881719e98507abed1011bf463a8fa23e9c3e98a8", size = 740646, upload-time = "2025-09-25T21:31:49.21Z" }, + { url = "https://files.pythonhosted.org/packages/2f/3a/61b9db1d28f00f8fd0ae760459a5c4bf1b941baf714e207b6eb0657d2578/pyyaml-6.0.3-cp310-cp310-manylinux2014_s390x.manylinux_2_17_s390x.manylinux_2_28_s390x.whl", hash = "sha256:66291b10affd76d76f54fad28e22e51719ef9ba22b29e1d7d03d6777a9174198", size = 840793, upload-time = "2025-09-25T21:31:50.735Z" }, + { url = "https://files.pythonhosted.org/packages/7a/1e/7acc4f0e74c4b3d9531e24739e0ab832a5edf40e64fbae1a9c01941cabd7/pyyaml-6.0.3-cp310-cp310-manylinux2014_x86_64.manylinux_2_17_x86_64.manylinux_2_28_x86_64.whl", hash = "sha256:9c7708761fccb9397fe64bbc0395abcae8c4bf7b0eac081e12b809bf47700d0b", size = 770293, upload-time = "2025-09-25T21:31:51.828Z" }, + { url = "https://files.pythonhosted.org/packages/8b/ef/abd085f06853af0cd59fa5f913d61a8eab65d7639ff2a658d18a25d6a89d/pyyaml-6.0.3-cp310-cp310-musllinux_1_2_aarch64.whl", hash = "sha256:418cf3f2111bc80e0933b2cd8cd04f286338bb88bdc7bc8e6dd775ebde60b5e0", size = 732872, upload-time = "2025-09-25T21:31:53.282Z" }, + { url = "https://files.pythonhosted.org/packages/1f/15/2bc9c8faf6450a8b3c9fc5448ed869c599c0a74ba2669772b1f3a0040180/pyyaml-6.0.3-cp310-cp310-musllinux_1_2_x86_64.whl", hash = "sha256:5e0b74767e5f8c593e8c9b5912019159ed0533c70051e9cce3e8b6aa699fcd69", size = 758828, upload-time = "2025-09-25T21:31:54.807Z" }, + { url = "https://files.pythonhosted.org/packages/a3/00/531e92e88c00f4333ce359e50c19b8d1de9fe8d581b1534e35ccfbc5f393/pyyaml-6.0.3-cp310-cp310-win32.whl", hash = "sha256:28c8d926f98f432f88adc23edf2e6d4921ac26fb084b028c733d01868d19007e", size = 142415, upload-time = "2025-09-25T21:31:55.885Z" }, + { url = "https://files.pythonhosted.org/packages/2a/fa/926c003379b19fca39dd4634818b00dec6c62d87faf628d1394e137354d4/pyyaml-6.0.3-cp310-cp310-win_amd64.whl", hash = "sha256:bdb2c67c6c1390b63c6ff89f210c8fd09d9a1217a465701eac7316313c915e4c", size = 158561, upload-time = "2025-09-25T21:31:57.406Z" }, + { url = "https://files.pythonhosted.org/packages/6d/16/a95b6757765b7b031c9374925bb718d55e0a9ba8a1b6a12d25962ea44347/pyyaml-6.0.3-cp311-cp311-macosx_10_13_x86_64.whl", hash = "sha256:44edc647873928551a01e7a563d7452ccdebee747728c1080d881d68af7b997e", size = 185826, upload-time = "2025-09-25T21:31:58.655Z" }, + { url = "https://files.pythonhosted.org/packages/16/19/13de8e4377ed53079ee996e1ab0a9c33ec2faf808a4647b7b4c0d46dd239/pyyaml-6.0.3-cp311-cp311-macosx_11_0_arm64.whl", hash = "sha256:652cb6edd41e718550aad172851962662ff2681490a8a711af6a4d288dd96824", size = 175577, upload-time = "2025-09-25T21:32:00.088Z" }, + { url = "https://files.pythonhosted.org/packages/0c/62/d2eb46264d4b157dae1275b573017abec435397aa59cbcdab6fc978a8af4/pyyaml-6.0.3-cp311-cp311-manylinux2014_aarch64.manylinux_2_17_aarch64.manylinux_2_28_aarch64.whl", hash = "sha256:10892704fc220243f5305762e276552a0395f7beb4dbf9b14ec8fd43b57f126c", size = 775556, upload-time = "2025-09-25T21:32:01.31Z" }, + { url = "https://files.pythonhosted.org/packages/10/cb/16c3f2cf3266edd25aaa00d6c4350381c8b012ed6f5276675b9eba8d9ff4/pyyaml-6.0.3-cp311-cp311-manylinux2014_s390x.manylinux_2_17_s390x.manylinux_2_28_s390x.whl", hash = "sha256:850774a7879607d3a6f50d36d04f00ee69e7fc816450e5f7e58d7f17f1ae5c00", size = 882114, upload-time = "2025-09-25T21:32:03.376Z" }, + { url = "https://files.pythonhosted.org/packages/71/60/917329f640924b18ff085ab889a11c763e0b573da888e8404ff486657602/pyyaml-6.0.3-cp311-cp311-manylinux2014_x86_64.manylinux_2_17_x86_64.manylinux_2_28_x86_64.whl", hash = "sha256:b8bb0864c5a28024fac8a632c443c87c5aa6f215c0b126c449ae1a150412f31d", size = 806638, upload-time = "2025-09-25T21:32:04.553Z" }, + { url = "https://files.pythonhosted.org/packages/dd/6f/529b0f316a9fd167281a6c3826b5583e6192dba792dd55e3203d3f8e655a/pyyaml-6.0.3-cp311-cp311-musllinux_1_2_aarch64.whl", hash = "sha256:1d37d57ad971609cf3c53ba6a7e365e40660e3be0e5175fa9f2365a379d6095a", size = 767463, upload-time = "2025-09-25T21:32:06.152Z" }, + { url = "https://files.pythonhosted.org/packages/f2/6a/b627b4e0c1dd03718543519ffb2f1deea4a1e6d42fbab8021936a4d22589/pyyaml-6.0.3-cp311-cp311-musllinux_1_2_x86_64.whl", hash = "sha256:37503bfbfc9d2c40b344d06b2199cf0e96e97957ab1c1b546fd4f87e53e5d3e4", size = 794986, upload-time = "2025-09-25T21:32:07.367Z" }, + { url = "https://files.pythonhosted.org/packages/45/91/47a6e1c42d9ee337c4839208f30d9f09caa9f720ec7582917b264defc875/pyyaml-6.0.3-cp311-cp311-win32.whl", hash = "sha256:8098f252adfa6c80ab48096053f512f2321f0b998f98150cea9bd23d83e1467b", size = 142543, upload-time = "2025-09-25T21:32:08.95Z" }, + { url = "https://files.pythonhosted.org/packages/da/e3/ea007450a105ae919a72393cb06f122f288ef60bba2dc64b26e2646fa315/pyyaml-6.0.3-cp311-cp311-win_amd64.whl", hash = "sha256:9f3bfb4965eb874431221a3ff3fdcddc7e74e3b07799e0e84ca4a0f867d449bf", size = 158763, upload-time = "2025-09-25T21:32:09.96Z" }, + { url = "https://files.pythonhosted.org/packages/d1/33/422b98d2195232ca1826284a76852ad5a86fe23e31b009c9886b2d0fb8b2/pyyaml-6.0.3-cp312-cp312-macosx_10_13_x86_64.whl", hash = "sha256:7f047e29dcae44602496db43be01ad42fc6f1cc0d8cd6c83d342306c32270196", size = 182063, upload-time = "2025-09-25T21:32:11.445Z" }, + { url = "https://files.pythonhosted.org/packages/89/a0/6cf41a19a1f2f3feab0e9c0b74134aa2ce6849093d5517a0c550fe37a648/pyyaml-6.0.3-cp312-cp312-macosx_11_0_arm64.whl", hash = "sha256:fc09d0aa354569bc501d4e787133afc08552722d3ab34836a80547331bb5d4a0", size = 173973, upload-time = "2025-09-25T21:32:12.492Z" }, + { url = "https://files.pythonhosted.org/packages/ed/23/7a778b6bd0b9a8039df8b1b1d80e2e2ad78aa04171592c8a5c43a56a6af4/pyyaml-6.0.3-cp312-cp312-manylinux2014_aarch64.manylinux_2_17_aarch64.manylinux_2_28_aarch64.whl", hash = "sha256:9149cad251584d5fb4981be1ecde53a1ca46c891a79788c0df828d2f166bda28", size = 775116, upload-time = "2025-09-25T21:32:13.652Z" }, + { url = "https://files.pythonhosted.org/packages/65/30/d7353c338e12baef4ecc1b09e877c1970bd3382789c159b4f89d6a70dc09/pyyaml-6.0.3-cp312-cp312-manylinux2014_s390x.manylinux_2_17_s390x.manylinux_2_28_s390x.whl", hash = "sha256:5fdec68f91a0c6739b380c83b951e2c72ac0197ace422360e6d5a959d8d97b2c", size = 844011, upload-time = "2025-09-25T21:32:15.21Z" }, + { url = "https://files.pythonhosted.org/packages/8b/9d/b3589d3877982d4f2329302ef98a8026e7f4443c765c46cfecc8858c6b4b/pyyaml-6.0.3-cp312-cp312-manylinux2014_x86_64.manylinux_2_17_x86_64.manylinux_2_28_x86_64.whl", hash = "sha256:ba1cc08a7ccde2d2ec775841541641e4548226580ab850948cbfda66a1befcdc", size = 807870, upload-time = "2025-09-25T21:32:16.431Z" }, + { url = "https://files.pythonhosted.org/packages/05/c0/b3be26a015601b822b97d9149ff8cb5ead58c66f981e04fedf4e762f4bd4/pyyaml-6.0.3-cp312-cp312-musllinux_1_2_aarch64.whl", hash = "sha256:8dc52c23056b9ddd46818a57b78404882310fb473d63f17b07d5c40421e47f8e", size = 761089, upload-time = "2025-09-25T21:32:17.56Z" }, + { url = "https://files.pythonhosted.org/packages/be/8e/98435a21d1d4b46590d5459a22d88128103f8da4c2d4cb8f14f2a96504e1/pyyaml-6.0.3-cp312-cp312-musllinux_1_2_x86_64.whl", hash = "sha256:41715c910c881bc081f1e8872880d3c650acf13dfa8214bad49ed4cede7c34ea", size = 790181, upload-time = "2025-09-25T21:32:18.834Z" }, + { url = "https://files.pythonhosted.org/packages/74/93/7baea19427dcfbe1e5a372d81473250b379f04b1bd3c4c5ff825e2327202/pyyaml-6.0.3-cp312-cp312-win32.whl", hash = "sha256:96b533f0e99f6579b3d4d4995707cf36df9100d67e0c8303a0c55b27b5f99bc5", size = 137658, upload-time = "2025-09-25T21:32:20.209Z" }, + { url = "https://files.pythonhosted.org/packages/86/bf/899e81e4cce32febab4fb42bb97dcdf66bc135272882d1987881a4b519e9/pyyaml-6.0.3-cp312-cp312-win_amd64.whl", hash = "sha256:5fcd34e47f6e0b794d17de1b4ff496c00986e1c83f7ab2fb8fcfe9616ff7477b", size = 154003, upload-time = "2025-09-25T21:32:21.167Z" }, + { url = "https://files.pythonhosted.org/packages/1a/08/67bd04656199bbb51dbed1439b7f27601dfb576fb864099c7ef0c3e55531/pyyaml-6.0.3-cp312-cp312-win_arm64.whl", hash = "sha256:64386e5e707d03a7e172c0701abfb7e10f0fb753ee1d773128192742712a98fd", size = 140344, upload-time = "2025-09-25T21:32:22.617Z" }, + { url = "https://files.pythonhosted.org/packages/d1/11/0fd08f8192109f7169db964b5707a2f1e8b745d4e239b784a5a1dd80d1db/pyyaml-6.0.3-cp313-cp313-macosx_10_13_x86_64.whl", hash = "sha256:8da9669d359f02c0b91ccc01cac4a67f16afec0dac22c2ad09f46bee0697eba8", size = 181669, upload-time = "2025-09-25T21:32:23.673Z" }, + { url = "https://files.pythonhosted.org/packages/b1/16/95309993f1d3748cd644e02e38b75d50cbc0d9561d21f390a76242ce073f/pyyaml-6.0.3-cp313-cp313-macosx_11_0_arm64.whl", hash = "sha256:2283a07e2c21a2aa78d9c4442724ec1eb15f5e42a723b99cb3d822d48f5f7ad1", size = 173252, upload-time = "2025-09-25T21:32:25.149Z" }, + { url = "https://files.pythonhosted.org/packages/50/31/b20f376d3f810b9b2371e72ef5adb33879b25edb7a6d072cb7ca0c486398/pyyaml-6.0.3-cp313-cp313-manylinux2014_aarch64.manylinux_2_17_aarch64.manylinux_2_28_aarch64.whl", hash = "sha256:ee2922902c45ae8ccada2c5b501ab86c36525b883eff4255313a253a3160861c", size = 767081, upload-time = "2025-09-25T21:32:26.575Z" }, + { url = "https://files.pythonhosted.org/packages/49/1e/a55ca81e949270d5d4432fbbd19dfea5321eda7c41a849d443dc92fd1ff7/pyyaml-6.0.3-cp313-cp313-manylinux2014_s390x.manylinux_2_17_s390x.manylinux_2_28_s390x.whl", hash = "sha256:a33284e20b78bd4a18c8c2282d549d10bc8408a2a7ff57653c0cf0b9be0afce5", size = 841159, upload-time = "2025-09-25T21:32:27.727Z" }, + { url = "https://files.pythonhosted.org/packages/74/27/e5b8f34d02d9995b80abcef563ea1f8b56d20134d8f4e5e81733b1feceb2/pyyaml-6.0.3-cp313-cp313-manylinux2014_x86_64.manylinux_2_17_x86_64.manylinux_2_28_x86_64.whl", hash = "sha256:0f29edc409a6392443abf94b9cf89ce99889a1dd5376d94316ae5145dfedd5d6", size = 801626, upload-time = "2025-09-25T21:32:28.878Z" }, + { url = "https://files.pythonhosted.org/packages/f9/11/ba845c23988798f40e52ba45f34849aa8a1f2d4af4b798588010792ebad6/pyyaml-6.0.3-cp313-cp313-musllinux_1_2_aarch64.whl", hash = "sha256:f7057c9a337546edc7973c0d3ba84ddcdf0daa14533c2065749c9075001090e6", size = 753613, upload-time = "2025-09-25T21:32:30.178Z" }, + { url = "https://files.pythonhosted.org/packages/3d/e0/7966e1a7bfc0a45bf0a7fb6b98ea03fc9b8d84fa7f2229e9659680b69ee3/pyyaml-6.0.3-cp313-cp313-musllinux_1_2_x86_64.whl", hash = "sha256:eda16858a3cab07b80edaf74336ece1f986ba330fdb8ee0d6c0d68fe82bc96be", size = 794115, upload-time = "2025-09-25T21:32:31.353Z" }, + { url = "https://files.pythonhosted.org/packages/de/94/980b50a6531b3019e45ddeada0626d45fa85cbe22300844a7983285bed3b/pyyaml-6.0.3-cp313-cp313-win32.whl", hash = "sha256:d0eae10f8159e8fdad514efdc92d74fd8d682c933a6dd088030f3834bc8e6b26", size = 137427, upload-time = "2025-09-25T21:32:32.58Z" }, + { url = "https://files.pythonhosted.org/packages/97/c9/39d5b874e8b28845e4ec2202b5da735d0199dbe5b8fb85f91398814a9a46/pyyaml-6.0.3-cp313-cp313-win_amd64.whl", hash = "sha256:79005a0d97d5ddabfeeea4cf676af11e647e41d81c9a7722a193022accdb6b7c", size = 154090, upload-time = "2025-09-25T21:32:33.659Z" }, + { url = "https://files.pythonhosted.org/packages/73/e8/2bdf3ca2090f68bb3d75b44da7bbc71843b19c9f2b9cb9b0f4ab7a5a4329/pyyaml-6.0.3-cp313-cp313-win_arm64.whl", hash = "sha256:5498cd1645aa724a7c71c8f378eb29ebe23da2fc0d7a08071d89469bf1d2defb", size = 140246, upload-time = "2025-09-25T21:32:34.663Z" }, + { url = "https://files.pythonhosted.org/packages/9d/8c/f4bd7f6465179953d3ac9bc44ac1a8a3e6122cf8ada906b4f96c60172d43/pyyaml-6.0.3-cp314-cp314-macosx_10_13_x86_64.whl", hash = "sha256:8d1fab6bb153a416f9aeb4b8763bc0f22a5586065f86f7664fc23339fc1c1fac", size = 181814, upload-time = "2025-09-25T21:32:35.712Z" }, + { url = "https://files.pythonhosted.org/packages/bd/9c/4d95bb87eb2063d20db7b60faa3840c1b18025517ae857371c4dd55a6b3a/pyyaml-6.0.3-cp314-cp314-macosx_11_0_arm64.whl", hash = "sha256:34d5fcd24b8445fadc33f9cf348c1047101756fd760b4dacb5c3e99755703310", size = 173809, upload-time = "2025-09-25T21:32:36.789Z" }, + { url = "https://files.pythonhosted.org/packages/92/b5/47e807c2623074914e29dabd16cbbdd4bf5e9b2db9f8090fa64411fc5382/pyyaml-6.0.3-cp314-cp314-manylinux2014_aarch64.manylinux_2_17_aarch64.manylinux_2_28_aarch64.whl", hash = "sha256:501a031947e3a9025ed4405a168e6ef5ae3126c59f90ce0cd6f2bfc477be31b7", size = 766454, upload-time = "2025-09-25T21:32:37.966Z" }, + { url = "https://files.pythonhosted.org/packages/02/9e/e5e9b168be58564121efb3de6859c452fccde0ab093d8438905899a3a483/pyyaml-6.0.3-cp314-cp314-manylinux2014_s390x.manylinux_2_17_s390x.manylinux_2_28_s390x.whl", hash = "sha256:b3bc83488de33889877a0f2543ade9f70c67d66d9ebb4ac959502e12de895788", size = 836355, upload-time = "2025-09-25T21:32:39.178Z" }, + { url = "https://files.pythonhosted.org/packages/88/f9/16491d7ed2a919954993e48aa941b200f38040928474c9e85ea9e64222c3/pyyaml-6.0.3-cp314-cp314-manylinux2014_x86_64.manylinux_2_17_x86_64.manylinux_2_28_x86_64.whl", hash = "sha256:c458b6d084f9b935061bc36216e8a69a7e293a2f1e68bf956dcd9e6cbcd143f5", size = 794175, upload-time = "2025-09-25T21:32:40.865Z" }, + { url = "https://files.pythonhosted.org/packages/dd/3f/5989debef34dc6397317802b527dbbafb2b4760878a53d4166579111411e/pyyaml-6.0.3-cp314-cp314-musllinux_1_2_aarch64.whl", hash = "sha256:7c6610def4f163542a622a73fb39f534f8c101d690126992300bf3207eab9764", size = 755228, upload-time = "2025-09-25T21:32:42.084Z" }, + { url = "https://files.pythonhosted.org/packages/d7/ce/af88a49043cd2e265be63d083fc75b27b6ed062f5f9fd6cdc223ad62f03e/pyyaml-6.0.3-cp314-cp314-musllinux_1_2_x86_64.whl", hash = "sha256:5190d403f121660ce8d1d2c1bb2ef1bd05b5f68533fc5c2ea899bd15f4399b35", size = 789194, upload-time = "2025-09-25T21:32:43.362Z" }, + { url = "https://files.pythonhosted.org/packages/23/20/bb6982b26a40bb43951265ba29d4c246ef0ff59c9fdcdf0ed04e0687de4d/pyyaml-6.0.3-cp314-cp314-win_amd64.whl", hash = "sha256:4a2e8cebe2ff6ab7d1050ecd59c25d4c8bd7e6f400f5f82b96557ac0abafd0ac", size = 156429, upload-time = "2025-09-25T21:32:57.844Z" }, + { url = "https://files.pythonhosted.org/packages/f4/f4/a4541072bb9422c8a883ab55255f918fa378ecf083f5b85e87fc2b4eda1b/pyyaml-6.0.3-cp314-cp314-win_arm64.whl", hash = "sha256:93dda82c9c22deb0a405ea4dc5f2d0cda384168e466364dec6255b293923b2f3", size = 143912, upload-time = "2025-09-25T21:32:59.247Z" }, + { url = "https://files.pythonhosted.org/packages/7c/f9/07dd09ae774e4616edf6cda684ee78f97777bdd15847253637a6f052a62f/pyyaml-6.0.3-cp314-cp314t-macosx_10_13_x86_64.whl", hash = "sha256:02893d100e99e03eda1c8fd5c441d8c60103fd175728e23e431db1b589cf5ab3", size = 189108, upload-time = "2025-09-25T21:32:44.377Z" }, + { url = "https://files.pythonhosted.org/packages/4e/78/8d08c9fb7ce09ad8c38ad533c1191cf27f7ae1effe5bb9400a46d9437fcf/pyyaml-6.0.3-cp314-cp314t-macosx_11_0_arm64.whl", hash = "sha256:c1ff362665ae507275af2853520967820d9124984e0f7466736aea23d8611fba", size = 183641, upload-time = "2025-09-25T21:32:45.407Z" }, + { url = "https://files.pythonhosted.org/packages/7b/5b/3babb19104a46945cf816d047db2788bcaf8c94527a805610b0289a01c6b/pyyaml-6.0.3-cp314-cp314t-manylinux2014_aarch64.manylinux_2_17_aarch64.manylinux_2_28_aarch64.whl", hash = "sha256:6adc77889b628398debc7b65c073bcb99c4a0237b248cacaf3fe8a557563ef6c", size = 831901, upload-time = "2025-09-25T21:32:48.83Z" }, + { url = "https://files.pythonhosted.org/packages/8b/cc/dff0684d8dc44da4d22a13f35f073d558c268780ce3c6ba1b87055bb0b87/pyyaml-6.0.3-cp314-cp314t-manylinux2014_s390x.manylinux_2_17_s390x.manylinux_2_28_s390x.whl", hash = "sha256:a80cb027f6b349846a3bf6d73b5e95e782175e52f22108cfa17876aaeff93702", size = 861132, upload-time = "2025-09-25T21:32:50.149Z" }, + { url = "https://files.pythonhosted.org/packages/b1/5e/f77dc6b9036943e285ba76b49e118d9ea929885becb0a29ba8a7c75e29fe/pyyaml-6.0.3-cp314-cp314t-manylinux2014_x86_64.manylinux_2_17_x86_64.manylinux_2_28_x86_64.whl", hash = "sha256:00c4bdeba853cc34e7dd471f16b4114f4162dc03e6b7afcc2128711f0eca823c", size = 839261, upload-time = "2025-09-25T21:32:51.808Z" }, + { url = "https://files.pythonhosted.org/packages/ce/88/a9db1376aa2a228197c58b37302f284b5617f56a5d959fd1763fb1675ce6/pyyaml-6.0.3-cp314-cp314t-musllinux_1_2_aarch64.whl", hash = "sha256:66e1674c3ef6f541c35191caae2d429b967b99e02040f5ba928632d9a7f0f065", size = 805272, upload-time = "2025-09-25T21:32:52.941Z" }, + { url = "https://files.pythonhosted.org/packages/da/92/1446574745d74df0c92e6aa4a7b0b3130706a4142b2d1a5869f2eaa423c6/pyyaml-6.0.3-cp314-cp314t-musllinux_1_2_x86_64.whl", hash = "sha256:16249ee61e95f858e83976573de0f5b2893b3677ba71c9dd36b9cf8be9ac6d65", size = 829923, upload-time = "2025-09-25T21:32:54.537Z" }, + { url = "https://files.pythonhosted.org/packages/f0/7a/1c7270340330e575b92f397352af856a8c06f230aa3e76f86b39d01b416a/pyyaml-6.0.3-cp314-cp314t-win_amd64.whl", hash = "sha256:4ad1906908f2f5ae4e5a8ddfce73c320c2a1429ec52eafd27138b7f1cbe341c9", size = 174062, upload-time = "2025-09-25T21:32:55.767Z" }, + { url = "https://files.pythonhosted.org/packages/f1/12/de94a39c2ef588c7e6455cfbe7343d3b2dc9d6b6b2f40c4c6565744c873d/pyyaml-6.0.3-cp314-cp314t-win_arm64.whl", hash = "sha256:ebc55a14a21cb14062aa4162f906cd962b28e2e9ea38f9b4391244cd8de4ae0b", size = 149341, upload-time = "2025-09-25T21:32:56.828Z" }, +] + +[[package]] +name = "typing-extensions" +version = "4.15.0" +source = { registry = "https://pypi.org/simple" } +sdist = { url = "https://files.pythonhosted.org/packages/72/94/1a15dd82efb362ac84269196e94cf00f187f7ed21c242792a923cdb1c61f/typing_extensions-4.15.0.tar.gz", hash = "sha256:0cea48d173cc12fa28ecabc3b837ea3cf6f38c6d1136f85cbaaf598984861466", size = 109391, upload-time = "2025-08-25T13:49:26.313Z" } +wheels = [ + { url = "https://files.pythonhosted.org/packages/18/67/36e9267722cc04a6b9f15c7f3441c2363321a3ea07da7ae0c0707beb2a9c/typing_extensions-4.15.0-py3-none-any.whl", hash = "sha256:f0fa19c6845758ab08074a0cfa8b7aecb71c999ca73d62883bc25cc018c4e548", size = 44614, upload-time = "2025-08-25T13:49:24.86Z" }, +] + +[[package]] +name = "virtualenv" +version = "20.36.1" +source = { registry = "https://pypi.org/simple" } +dependencies = [ + { name = "distlib" }, + { name = "filelock" }, + { name = "platformdirs" }, + { name = "typing-extensions", marker = "python_full_version < '3.11'" }, +] +sdist = { url = "https://files.pythonhosted.org/packages/aa/a3/4d310fa5f00863544e1d0f4de93bddec248499ccf97d4791bc3122c9d4f3/virtualenv-20.36.1.tar.gz", hash = "sha256:8befb5c81842c641f8ee658481e42641c68b5eab3521d8e092d18320902466ba", size = 6032239, upload-time = "2026-01-09T18:21:01.296Z" } +wheels = [ + { url = "https://files.pythonhosted.org/packages/6a/2a/dc2228b2888f51192c7dc766106cd475f1b768c10caaf9727659726f7391/virtualenv-20.36.1-py3-none-any.whl", hash = "sha256:575a8d6b124ef88f6f51d56d656132389f961062a9177016a50e4f507bbcc19f", size = 6008258, upload-time = "2026-01-09T18:20:59.425Z" }, +]