A proxy server that converts Poe and AWS model APIs to OpenAI protocol, allowing you to use these models with tools and clients built for the OpenAI API.
- Converts Poe API to OpenAI API format
- Supports AWS Bedrock for Claude-series models
- Supports function calling
- Multiple API key management
- Production-ready with proper error handling
- Python 3.10+
- Poetry for dependency management
- Clone the repository
- Install dependencies with Poetry:
poetry install- Copy the example config and customize it:
cp config.yaml.example config.yaml
# Edit config.yaml with your API keys and settingsRun the server locally:
poetry run python poe-server.pyThe repository includes a Dockerfile for easy deployment. For production use, we recommend using Docker Compose.
Example docker-compose.yml:
version: '3.8'
services:
poe-server:
image: zihaokevinzhou/poe-proxy:latest
container_name: poe-server
ports:
- "8000:8080"
environment:
- PORT=8080
- USE_HTTP=true
volumes:
- ./config.yaml:/app/config.yaml:ro
restart: unless-stopped
networks:
default:
name: poe-network
driver: bridgeDeploy with:
docker-compose up -dAccess the OpenAI-compatible API at http://localhost:8000/v1/.
port: 8080
ssl: true
ssl_dir: /etc/certificate
aws_secret_access_key: <your_aws_secret_access_key>
aws_access_key_id: <your_aws_access_key_id>
# If aws keys are set, the server will use AWS Bedrock for Claude models
api_key:
- <Poe API Key at https://poe.com/api_key>
- <You can add more keys such that if the first key is rate limited, the second key will be used>
extra_api_keys:
- <your_extra_api_key>
- <These keys can be used for authentication without leaking your Poe API key>
MIT License