Context Engine (optional, opt-in)

Semantic + graph code search for SuperCoder. The desktop app stays zero-backend by default; enabling the Context engine toggle in Settings means "run this compose stack." The app then streams your repo here for indexing and the agent's codebase_search / codebase_graph tools query it.

Run the stack

cd services/context-engine
cp .env.example .env          # set SUPERCODER_OPENAI_API_KEY (server-side embedding key)
docker compose up -d --build

Brings up: postgres, redis, qdrant, falkordb, a one-shot migrate (Atlas), and the context-engine-server (HTTP on :8106) + context-engine-worker.

Check it's healthy:

docker compose ps
curl -fsS http://localhost:8106/api/health
docker compose logs context-engine-server | grep -i "streaming sync endpoints registered"

Then in the desktop app: Settings → Context engine → enable, leave the port at 8106, and open a coding session on a repo. The app streams it up; once indexing reaches done, codebase_search returns results.

Notes

• Embedding key lives here (.env → SUPERCODER_OPENAI_API_KEY), never in the app.
• Merkle trees persist to the shared merkle Docker volume (local-disk CAS, flock + sha256). No S3/MinIO. The server and worker share this volume so incremental /index/diff sees what the worker's finalizer committed.
• Port is editable in Settings (default 8106); the app builds base_url = http://127.0.0.1:<port>.
• Tear down with docker compose down (add -v to drop the indexed data volumes).