Spike: Adding Llamaindex llm instrumentation #78
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
- Add embedding event handlers (_handle_embedding_start, _handle_embedding_end) - Extract model name, input texts, and dimension count from embedding events - Create vendor_detection.py module with VendorRule-based provider detection - Support 13+ embedding providers (OpenAI, Azure, AWS, Google, Cohere, etc.) - Add test_embedding_instrumentation.py with single and batch embedding tests - Update README with embedding documentation and provider list - Tested successfully with OpenAI embeddings API
Contributor
There was a problem hiding this comment.
should we change the description and remove/change the authors?
Contributor
Author
There was a problem hiding this comment.
I am keeping consistent with the description and authors for langchain.
| wrapper=_BaseCallbackManagerInitWrapper(llamaindexCallBackHandler), | ||
| ) | ||
|
|
||
| def _uninstrument(self, **kwargs): |
Contributor
There was a problem hiding this comment.
Should we implement this?
...-instrumentation-llamaindex/src/opentelemetry/instrumentation/llamaindex/callback_handler.py
Show resolved
Hide resolved
| response = payload.get("response") | ||
|
|
||
| # Handle both dict and object types for response | ||
| if response: |
Contributor
There was a problem hiding this comment.
nitpick which might be unrealistic, but would it make sense to group the dictionary vs non-dictionary use cases? Seems like there is a lot of isinstance("", dict) checking.
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Core Implementation:
_handle_llm_start(): Captures model name, input messages, creates LLMInvocation entity
_handle_llm_end(): Extracts response content, token usage, finalizes span
Handles LlamaIndex-specific message formats (ChatMessage objects, blocks[0].text structure)
Supports both dict and object response types
Token extraction from response.raw.usage (OpenAI format)
Testing:
test_llm_instrumentation.py: Live OpenAI API integration test
Validates span attributes, message content, token counts
Verifies metrics emission
Documentation:
Quick start guide with code example
Expected span attributes and metrics output
Key differences vs LangChain (event-based callbacks, CBEventType enum)
Span Attributes (Gen AI Semantic Conventions)
Metrics:
gen_ai.client.operation.duration (histogram)
gen_ai.client.token.usage (counter)