AI Designer is API-only in production. All design requests flow through the FastAPI service.
| Entry | Command | Notes |
|---|---|---|
| Docker (prod) | docker compose --profile prod up |
Builds docker/Dockerfile.production, exposes port 8000 |
| Docker (dev) | docker compose --profile dev up |
Hot-reload via docker/Dockerfile.dev |
| Local (optional) | ai-designer-api or uvicorn ai_designer.api.app:app --host 0.0.0.0 --port 8000 |
Requires Redis and OPPER_API_KEY |
| Method | Path | Purpose |
|---|---|---|
| GET | /health |
Liveness |
| POST | /api/v1/design |
Submit NL design request (multi-agent pipeline) |
| GET | /api/v1/design/{request_id} |
Poll design status |
| WS | /ws/progress/{request_id} |
Real-time pipeline progress |
| GET | /metrics |
Prometheus metrics |
See ARCHITECTURE.md for the agent pipeline and README.md for setup.
The following were removed in the production restructure:
python -m ai_designer/ai-designerCLIFreeCADCLI/cli_pkg/SystemOrchestrator(--enhanced/--legacy-enhanced)- Legacy LLM wrappers (
llm/client.py,unified_manager.py, etc.)
Use the HTTP API for all new integrations.