Trash Challenge — MVP

FastAPI + Celery + Redis + Postgres + MinIO (S3) + Telegram webhook.

Quickstart (local)

1. Clone and create .env from .env.example and fill tokens.
2. Start services:

docker compose up -d

3. Create the S3 bucket:

bash scripts/create_buckets.sh

4. Expose the API (for Telegram webhook), e.g. using ngrok:

ngrok http 8000

5. Set Telegram webhook:

export $(cat .env | xargs)
bash scripts/set_webhook.sh

6. Send a photo to your bot. You should see:

• Upload to MinIO
• Celery task executing inference stub
• Score saved to Postgres
• Telegram message with points
• GET http://localhost:8000/leaderboard shows team points

Notes

• For production, replace MinIO with AWS S3 (clear S3_ENDPOINT and set AWS creds in environment).
• Replace worker/inference_stub.py with your real model.
• Add before/after logic and geofencing next.