## Build & Development Commands
- Use `docker-compose up --build` to build and run the app locally. - Ensure `requirements.txt` is up to date with all Python dependencies. - Use `python -m app.api` to run the API locally for testing. - Swagger UI is served from `openapi.yaml` and should reflect all available endpoints.
## API Design Guidelines
- `/predict`: Accepts a test dataset (e.g., JSON or file upload) and returns model predictions. - `/train`: Accepts a dataset and optional parameters to train a new model. - `/models`: 
  - `GET`: Lists available models in the local registry.
  - `POST`: Uploads a new model to the registry.
- `/status/<job_id>`: Returns the status of a training job (e.g., pending, running, completed, failed). - All endpoints must be documented in `openapi.yaml` for DART UI integration.
## Model Registry Guidelines
- Models are stored in the `models/` directory with metadata (e.g., name, version, date). - `model_registry.py` must handle model lookup, registration, and versioning. - No cloud storage is used; all models are stored locally or on a private server.
## Inference & Training Logic
- `inference.py` should load the latest or specified model from the registry and run predictions. - `training.py` should support training from scratch or fine-tuning, saving the model to the registry. - Long-running training jobs should be handled asynchronously via `jobs.py`.
## Job Management
- Use `jobs.py` to manage background tasks (e.g., training). - Each job should have a unique ID and status tracking. - Consider using `threading`, `multiprocessing`, or a lightweight queue like `RQ`.
## Documentation Guidelines
- Keep `README.md` updated with setup, usage, and endpoint examples. - Ensure `openapi.yaml` is synchronized with actual API behavior. - Document model formats, expected input/output, and training parameters.
## Containerization Guidelines
- Use the provided `Dockerfile` and `docker-compose.yml` for reproducible builds. - Ensure all paths in `config.py` are relative or configurable via environment variables. - Avoid hardcoding file paths or secrets.
## Configuration
- Use `config.py` to load settings from a `.env` file or YAML/JSON config. - Include paths for model storage, logging, and job tracking.