fix(langchain): parse tool_calls from AIMessage

[hassiebp]

Important

Enhance _convert_message_to_dict in CallbackHandler.py to capture tool_calls from AIMessage for improved traceability in LangChain workflows.

• Behavior:
  • Modify _convert_message_to_dict in CallbackHandler.py to include tool_calls from AIMessage if present.
  • Ensures tool_calls are captured in the message dictionary for better traceability.
• Misc:
  • Minor import reordering in client.py and prompt_cache.py for consistency.
  • Adjust docstring formatting in client.py for clear_prompt_cache() function.

^{This description was created by for cba05ab. You can customize this summary. It will automatically update as commits are pushed.}

---

Disclaimer: Experimental PR review

Greptile Summary

This PR adds support for capturing tool_calls from AIMessage objects in the Langchain callback handler. The change modifies the _convert_message_to_dict method in langfuse/langchain/CallbackHandler.py to check if an AIMessage has tool_calls and includes them in the message dictionary when present.

The implementation adds a conditional block that checks for the existence of the tool_calls attribute, ensures it's not None, and verifies it has content before adding it to the message dictionary. This follows the established pattern used throughout the method for other optional message attributes.

This enhancement integrates with Langfuse's observability capabilities by ensuring that tool/function calling information from LangChain agents and chat models is properly captured and tracked. Modern LLM workflows increasingly rely on tool calling for complex tasks, and this data is essential for understanding agent behavior, debugging workflows, and maintaining comprehensive trace records. The change fits naturally into the existing message conversion logic and maintains backward compatibility.

Confidence score: 5/5

• This PR is safe to merge with minimal risk
• Score reflects simple, well-tested conditional logic following established patterns
• No files require special attention

[greptile-apps]

_{1 file reviewed, no comments}

_{Edit Code Review Bot Settings | Greptile}