Sozo

OpenAI-compatible proxy to OpenRouter, targeting Qwen 3.6 Plus.

Any client that speaks the OpenAI API (SDKs, curl, etc.) can point at Sozo and get responses from Qwen 3.6 Plus via OpenRouter — no client-side configuration changes needed. If the upstream is rate-limited, Sozo automatically retries with exponential backoff.

Quick Start

# Set your OpenRouter API key
export OPENROUTER_API_KEY=sk-or-...

# Run with Docker Compose
docker compose up -d

# Or build and run locally
make build
./bin/sozo

Usage

Sozo exposes a standard OpenAI-compatible endpoint:

curl http://localhost:8080/v1/chat/completions \
  -H "Content-Type: application/json" \
  -d '{
    "messages": [{"role": "user", "content": "Hello!"}],
    "stream": false
  }'

Streaming works too:

curl http://localhost:8080/v1/chat/completions \
  -H "Content-Type: application/json" \
  -d '{
    "messages": [{"role": "user", "content": "Hello!"}],
    "stream": true
  }'

Or use any OpenAI SDK by pointing it at Sozo:

from openai import OpenAI

client = OpenAI(base_url="http://localhost:8080/v1", api_key="unused")
response = client.chat.completions.create(
    model="anything",  # ignored — Sozo rewrites to Qwen 3.6 Plus
    messages=[{"role": "user", "content": "Hello!"}],
)

Configuration

Variable	Required	Default	Description
OPENROUTER_API_KEY	yes	—	Your OpenRouter API key
LISTEN_ADDR	no	:8080	Listen address
OPENROUTER_BASE_URL	no	https://openrouter.ai/api/v1	Upstream URL
DEFAULT_MODEL	no	qwen/qwen3.6-plus:free	Model ID injected into requests

Endpoints

Method	Path	Description
GET	/	Redirects to /docs
POST	/v1/chat/completions	Chat completions (streaming and non-streaming)
GET	/health	Health check
GET	/swagger.json	OpenAPI 3.0 spec
GET	/docs	Swagger UI

Deployment

Sozo is designed to sit behind Traefik. The included docker-compose.yml has Traefik labels pre-configured for production. For local development, docker-compose.override.yml routes via sozo.localhost on the https entrypoint.

To deploy to production, update the host rule in docker-compose.yml:

labels:
  - "traefik.http.routers.sozo.rule=Host(`your-domain.com`)"

Both compose files join the external traefik_traefik network. The Docker image is ~14MB (multi-stage build to scratch).

Development

make build       # Build binary
make test        # Run tests with race detector
make docker      # Build Docker image
make up          # docker compose up -d
make down        # docker compose down
make smoke-test  # Smoke test against running instance

License

TBD