WALD API Proxy

FastAPI proxy that exposes OpenAI-compatible endpoints for WALD.

Local Run

uv venv .venv
uv pip install -r requirements.txt
uv run python main.py

The service listens on http://127.0.0.1:8000 locally.

Docker

docker build -t wald-api .
docker run --rm -p 8000:8000 wald-api

GitHub Actions

The workflow at .github/workflows/docker-image.yml builds and pushes the image to GitHub Container Registry:

ghcr.io/<owner>/<repo>:latest
ghcr.io/<owner>/<repo>:sha-<commit>

Required repository settings:

• Enable GitHub Actions.
• After the first image is pushed, set the GHCR package visibility to public if Render should pull it without credentials.

Optional repository secret:

RENDER_DEPLOY_HOOK_URL

Set this to a Render deploy hook URL if you want GitHub Actions to trigger a Render deploy after pushing the image.

Render Free Deploy

Option A: Deploy from this GitHub repo using render.yaml.

1. Push this repository to GitHub.
2. In Render, create a new Blueprint from the repo.
3. Render will build the Dockerfile and run the web service on the free plan.

Option B: Deploy from the GHCR image.

1. Let GitHub Actions publish ghcr.io/<owner>/<repo>:latest.
2. Make the GHCR package public.
3. Create a Render web service from an existing Docker image.
4. Use ghcr.io/<owner>/<repo>:latest as the image URL.

Environment Variables

Required on Render:

None

Optional:

PROXY=

Use PROXY only if outbound requests to WALD must go through a proxy, for example:

PROXY=http://user:password@host:port

Render provides PORT automatically. The Docker command reads it through ${PORT:-8000}.

API credentials are not configured as server environment variables. Clients pass them per request with:

Authorization: Bearer <base64-json-api-key>