AccessibilityPlus turns citizen photos into trustworthy accessibility data.
Every report now includes:

1. Automatic location capture – the web client requests the user’s GPS coordinates via the browser Geolocation API.
2. AI verification – images are analyzed with Amazon Rekognition and cross-checked against the description with OpenAI.
3. Persistent storage in S3 – both the raw photo and the enriched metadata live under the reports/ prefix in your bucket.
4. Map-ready feeds – the API returns the aggregated reports as JSON + GeoJSON so the Leaflet map (or any GIS tool) can render live markers.

Browser (HTML + JS)
 ├─ Captures geolocation + photo
 └─ POST /reports (FastAPI)
        │
        ├─ Upload photo → Amazon S3
        ├─ Rekognition.detect_labels(image)
        ├─ GPT compares labels vs. citizen description
        └─ Persist metadata + update reports/index.json in S3

GET /reports/geojson → Leaflet map fetches FeatureCollection for visualization

This keeps the workflow serverless-friendly: the FastAPI app can later move into AWS Lambda + API Gateway without code changes, and S3 remains the single source of truth for both media assets and map data.

Standard AWS credentials (env vars, ~/.aws/credentials, or AWS SSO) must also be configured locally.

pip install -r requirements.txt
# web server
uvicorn server:app --reload
# cli workflow for manual uploads
python main.py

• Visit http://127.0.0.1:8000/ to submit a report with location capture.
• Visit http://127.0.0.1:8000/map to see the Leaflet-based accessibility map.
• Use python main.py if you prefer the CLI helper to upload a local image directly to S3.

reports/
  {report_id}/
    photo.jpg
    metadata.json
  index.json   ← append-only manifest used by /reports + /reports/geojson

Each metadata.json bundle contains the image description, lat/lon, Rekognition labels, GPT verdict, and helpful URLs (public + pre-signed) so downstream tooling can retrieve the supporting evidence.