LLM Path

A lightweight tool for tracing LLM API requests.

Trace demo: nanobot¹, claude code, deep research²

Features

• Transparent Proxy — Drop-in HTTP proxy that captures all LLM API traffic. Works with OpenAI, Anthropic (Claude), and Google (Gemini) APIs.
• Request Visualization — Interactive web viewer to visualize the request topology graph and show the context diff between requests.

Installation

pip install llm-path

Quick Start

1. Start the Proxy

llm-path proxy --port 8080 --target https://api.openai.com --output trace.jsonl

Replace the --target host in the command above with your LLM provider's API host.

2. Point Your Client to the Proxy

from openai import OpenAI

- client = OpenAI()
+ client = OpenAI(base_url="http://localhost:8080/v1")

All requests will be transparently forwarded to your LLM provider and recorded to the trace file.

3. Visualize the Traces

llm-path viewer trace.jsonl

Proxy for More Providers

Google Agent Development Kit (ADK)

Start the proxy:

llm-path proxy --port 8080 --target https://generativelanguage.googleapis.com --output adk.jsonl

Set the environment variable for your ADK application:

GOOGLE_GEMINI_BASE_URL=http://127.0.0.1:8080

CLI Reference

# Start proxy server
llm-path proxy [OPTIONS]
  --port      Port to listen on (default: 8080)
  --output    Output JSONL file path (required)
  --target    LLM Provider API URL (required)

# Visualize traces
llm-path viewer <input> [OPTIONS]
  --port      Port to listen on (default: 8765)
  --host      Host to bind to (default: 127.0.0.1)

Development Guide

git clone https://github.com/wang0618/llm-path.git
cd llm-path
uv sync

uv run llm-path proxy --port 8080 --target https://api.openai.com --output trace.jsonl &
uv run llm-path cook trace.jsonl -o ./viewer/public/data.json

cd viewer
npm install
npm run dev

License

MIT

Footnotes

1. nanobot is a lightweight OpenClaw implementation in Python. ↩

2. The deep research agent is from the Google ADK demo, see deepresearch ↩