langchain-middleware

LangChain Middleware Example

This example demonstrates how to use GalileoMiddleware to automatically log LangChain agent executions without manual callback setup.

Overview

GalileoMiddleware provides drop-in logging for LangGraph agents. It automatically captures:

• Agent lifecycle events (start/completion)
• Model calls with prompts, responses, and metadata
• Tool calls with function names, arguments, and outputs

Setup

1. Install dependencies:

pip install -r requirements.txt

1. Configure environment variables:

cp .env.example .env
# Edit .env with your API keys

Running the Example

python main.py

The agent will:

1. Create a LangChain agent with two tools (weather and stock price)
2. Invoke the agent with a question requiring both tools. Specifically, ""What's the weather like in San Francisco and what's the current stock price of Apple?"
3. Automatically log all interactions to Galileo using middleware

What Gets Logged

The middleware creates a hierarchical trace with:

• Agent node: Tracks overall execution with input/output state
• LLM nodes: Logs each model call with prompts, responses, and timing
• Tool nodes: Records tool invocations with arguments and outputs

Key Features

• Zero configuration: Add middleware to your agent and it works automatically
• Async support: Works seamlessly with both sync and async execution
• Flexible: Can use custom loggers and control trace management
• Comprehensive: Captures all relevant execution details

Comparison with Callbacks

Feature	GalileoMiddleware	GalileoCallback
Setup	Add to middleware list	Pass to each component
Complexity	Simple	Manual setup required
Best for	LangGraph agents	Complex LangChain apps

For more details, see the middleware documentation.