A Code Snippet for Langchain Observability/Understanding

The notebook in this folder contains a code snippet that shows exact individual LLM calls Langchain agents make during agent execution. The notebook also has a walkthrough and demonstration of the code snippet.

In complex agents, it's not always obvious exactly what text is being sent to the LLM. The code snippet implements a Langchain callback handler that exposes those calls, along with providing some basic assistance tracking and debugging Langchain chains.

Requirements

To run the walkthrough and demonstration in the notebook you'll need access to a Google Cloud project with the Vertex AI API enabled.

The Langchain callback code snippet in the notebook can be run independently of Google Cloud, in any Python environment.

Getting Help

If you have any questions or find any problems, please report through GitHub issues.