feat(langchain): set callback spans to active in OTEL context

[hassiebp]

Important

Enhances LangChain CallbackHandler with OTEL context management for better distributed tracing in Langfuse.

• Behavior:
  • Adds _attach_observation() and _detach_observation() methods in CallbackHandler.py to manage OTEL context.
  • Replaces direct self.runs assignments with span attachment/detachment in methods like on_chain_start, on_tool_start, on_llm_start, etc.
  • Ensures Langfuse spans are active in OTEL context for proper trace hierarchy.
• Data Structures:
  • Introduces context_tokens dictionary to track OTEL context tokens alongside runs.
• Error Handling:
  • Consistent error handling and resource cleanup in span management methods.
• Misc:
  • Minor import reordering in CallbackHandler.py.

^{This description was created by for a0ce171. You can customize this summary. It will automatically update as commits are pushed.}

---

Disclaimer: Experimental PR review

Greptile Summary

This PR enhances the LangChain CallbackHandler in Langfuse to properly integrate with OpenTelemetry (OTEL) context management. The changes implement systematic span attachment and detachment methods that ensure Langfuse spans become the "active span" in OTEL context, which is crucial for distributed tracing and frameworks like LangGraph that rely on OTEL context propagation.

The implementation adds two key methods:

• _attach_observation(): Sets a Langfuse span as active in OTEL context using trace.set_span_in_context() and context.attach(), storing the context token for later cleanup
• _detach_observation(): Properly cleans up OTEL context by calling context.detach() with the stored token and removes the span from tracking

These methods replace direct assignment to self.runs throughout all callback lifecycle methods (on_chain_start, on_tool_start, on_llm_start, etc.). The change adds a context_tokens dictionary to track OTEL context tokens alongside the existing runs dictionary, ensuring proper pairing of attach/detach operations.

This integration allows Langfuse spans to participate correctly in OTEL's context propagation mechanism, ensuring that child operations created by other OTEL-aware libraries are properly nested under Langfuse spans in the trace hierarchy. This is particularly important for complex workflows using frameworks like LangGraph where multiple tracing systems need to work together seamlessly.

Confidence score: 4/5

• This PR is safe to merge with minimal risk as it maintains backward compatibility while adding essential OTEL integration
• Score reflects well-structured implementation with consistent error handling and proper resource cleanup patterns
• Pay close attention to the _attach_observation() and _detach_observation() methods to ensure OTEL context management is working correctly

[greptile-apps]

_{1 file reviewed, no comments}

_{Edit Code Review Bot Settings | Greptile}