Skip to content

DEPLOYMENT.md

Latest commit

 

History

History
33 lines (23 loc) · 859 Bytes

DEPLOYMENT.md

File metadata and controls

33 lines (23 loc) · 859 Bytes
Raw

Deployment

This repo should not be deployed as a flagship project yet. It is now structured so it can be deployed after the dataset and evaluation story are improved.

Local API

Train first:

python scripts/train.py

Run:

uvicorn app.api:app --host 0.0.0.0 --port 8000

Future Render or Railway Plan

  1. Add a Dockerfile or platform start command.
  2. Train the model during build only if the dataset is small and reproducible.
  3. Prefer storing a versioned model artifact externally for real projects.
  4. Set environment variables from .env.example.
  5. Verify /health and /predict.

Production Checklist Before Public Demo

  • Replace synthetic data with a cited public dataset.
  • Add an evaluation report committed under docs/.
  • Add model comparison and threshold tuning.
  • Add explainability.
  • Add Dockerfile.