Enterprise AI Agents

Enterprise AI Agents is an applied research project that implements a production-ready multi-agent assistant using Google's Agent Development Kit (ADK) combined with commercial and local LLM providers. The repository contains reproducible notebooks, reference assets, and integration scaffolding that demonstrate how to orchestrate task-specific agents for enterprise service operations.

Key Capabilities

• Multi-model routing across Google ADK, OpenAI, Anthropic, Groq, and local Ollama deployments.
• Cost-optimized execution strategies with benchmarking, caching, and provider fallbacks.
• Domain agents for customer support, research, analytics, and automation workflows.
• Observability patterns covering telemetry, spend tracking, and reliability metrics.
• Deployment primitives for local development, containerized services, and CI validation.

Repository Layout

assets/                Reusable prompts, diagrams, and datasets referenced by the agents
notebooks/
  01_foundations/      Environment bootstrapping, architecture overview, and first agent build
  02_architecture/     Modular agent patterns, routing logic, and orchestration workflows
  03_integrations/     MCP connectors, knowledge retrieval, and enterprise research tooling
  04_multi_agent/      A2A coordination patterns and end-to-end service simulations
requirements.txt       Python dependency set for agents, notebooks, and tests
AGENTS.md              Contributor guide for extending the agent suite
CONTRIBUTING.md        Workflow expectations for collaborators
docker-compose.yml     JupyterLab/devcontainer configuration for local experimentation

Quick Start

1. Install Python 3.8 or later and create an isolated environment:
   python -m venv .venv && source .venv/bin/activate
2. Install dependencies: pip install -r requirements.txt
3. Populate a .env with provider credentials (GOOGLE_ADK_API_KEY, OPENAI_API_KEY, etc.).
4. Launch the interactive workspace: jupyter lab (or jupyter notebook).
5. Run notebooks in ascending order to reproduce the baseline agent fleet.

Development Workflow

• Extract reusable helpers into .py modules beside the notebooks when logic grows beyond exploratory cells.
• Keep notebooks deterministic; restart the kernel and rerun before committing changes.
• Record significant prompts, benchmarks, and design decisions inside assets/ so they can be versioned and reviewed.

Testing & Validation

• Smoke tests live under tests/ (create the directory if missing) and mirror the notebook structure (tests/test_multi_agent.py).
• Use pytest or pytest-asyncio to verify agent initialization, routing fallbacks, and regression baselines.
  Example: pytest -k concierge to run agent-specific suites.
• Track experimental metrics (latency, cost, accuracy) using fixtures so regressions are visible in CI.

Containerized Run (Optional)

• Start JupyterLab with docker-compose up to develop inside a disposable container.
• Mounts the repository at /workspace with dependencies managed via your notebooks or pip install -r requirements.txt inside the container shell.
• Adjust the service definition if you need to expose APIs or background workers in addition to the notebook environment.

Contributing

Please review AGENTS.md and CONTRIBUTING.md before proposing changes. Contributions should document the business capability being added, include reproducible notebooks, and supply smoke tests when new agents are introduced.

License

This project is released under the MIT License. See LICENSE for details.