Integrador_2

1) Backend setup

cd backend
python -m venv .venv
# Windows
.venv\Scripts\activate
pip install -r requirements.txt

Start backend:

uvicorn main:app --reload --host 127.0.0.1 --port 8000

Backend structure:

• app/api/routes: FastAPI route modules
  • health.py: Health check endpoints
  • auth.py: Authentication endpoints
  • model.py: Legacy model loading and segmentation
  • registry.py: Model registry management
• app/services: service logic
  • model_manager.py: Model loading and inference
  • registry.py: Model registry with Supabase persistence
  • mlflow_logger.py: MLflow tracking for model lineage
• app/core: configuration and clients
• app/schemas: Pydantic request/response schemas
• models/: Local model artifact storage
• mlruns/: MLflow tracking directory
• migrations/: Database migration files

Health check:

• GET http://127.0.0.1:8000/health

2) Frontend setup

cd frontend
npm install
npm start

Optional API override (defaults to http://127.0.0.1:8000):

set VITE_API_URL=http://127.0.0.1:8000

Model Registry Workflow

The new model registry allows uploading, managing, and activating models for segmentation.

Model Registry API

Upload a model:

POST /registry/models/upload
- Parameters: name, version, description (optional), file
- Returns: ModelRegistryResponse with model ID, metadata

List all models:

GET /registry/models
- Returns: ModelListResponse with all registered models

Get active model:

GET /registry/models/active
- Returns: Currently active ModelRegistryResponse
- Error 503: No active model

Activate a model:

POST /registry/models/{model_id}/activate
- Returns: Activated ModelRegistryResponse
- Note: Deactivates all other models atomically

Delete a model:

DELETE /registry/models/{model_id}
- Returns: Deletion confirmation
- Note: Removes both metadata and artifact files

Backend Tests

cd backend
pytest