Learn how to integrate Maxim observability and online evaluation with your LiteLLM Proxy in just one line of configuration.

​
Prerequisites

Install the required Python packages: 
pip install litellm[proxy]>=1.30.0 maxim-py==3.4.16 python-dotenv>=0.21.1

​
Project Layout

.
├── config.yml
├── maxim_proxy_tracer.py
├── requirements.txt
├── .env
└── (optional) Dockerfile & docker-compose.yml

​
1. Define the Tracer

Create a file maxim_proxy_tracer.py next to your proxy entrypoint:
maxim_proxy_tracer.py
from maxim.logger.litellm_proxy import MaximLiteLLMProxyTracer

# This single object wires up all LiteLLM traffic to Maxim
litellm_handler = MaximLiteLLMProxyTracer()

​
2. Update config.yml

Point LiteLLM’s callback at your tracer:
config.yml
litellm_settings:
  callbacks: maxim_proxy_tracer.litellm_handler
(Your existing model_list and general_settings remain unchanged.)

​
3. Configure Environment Variables

Add the following to a .env file or export in your shell: 
OPENAI_API_KEY=…       # API key for LiteLLM
MAXIM_API_KEY=…        # API key for Maxim ingestion
MAXIM_LOG_REPO_ID=…    # ID of your Maxim log repository

​
4. Run the Proxy Locally

You can start the proxy directly via the LiteLLM CLI: 
litellm --port 8000 --config config.yml

​
5. Run with Docker Compose

If you prefer Docker, use the provided Dockerfile and docker-compose.yml: 
docker-compose up -d
  • Port: 8000
  • Health check: GET /health
  • Logs: streamed to proxy_logs.log
That’s it—no additional code changes required. Every request through your LiteLLM Proxy will now be traced, logged, and evaluated in Maxim.