README.md

metadata

title: Civic ML Backend
emoji: 🏛️
colorFrom: blue
colorTo: purple
sdk: docker
sdk_version: 20.10.9
app_file: Dockerfile
pinned: false
license: mit
tags:
  - civic-services
  - machine-learning
  - fastapi
  - image-classification
  - duplicate-detection
  - clip
  - civics
  - backend
language:
  - code
  - en
library_name: fastapi
pipeline_tag: image-classification

Civic ML Backend (Full: pHash + Location)

Overview:

• FastAPI backend that accepts citizen reports and performs:
  • rule-based abuse detection
  • text duplicate detection
  • image duplicate detection using pHash (imagehash)
  • location-based duplicate detection using Haversine formula
  • image classification using CLIP (if available) with URL fallback
  • priority detection (keyword-based)

Run:

1. python3 -m venv .venv
2. source .venv/bin/activate # Windows: c
3. pip install -r requirements.txt
4. uvicorn app.main:app --reload --port 8000

Notes:

• The in-memory stores (seen_reports, seen_image_hashes, seen_locations) are ephemeral and reset on server restart.
• CLIP model download requires internet and may take time; if unavailable, the system uses URL keyword fallback for image labels.
• data/dataset.jsonl collects all incoming reports and results for later training/audit.

civic-ml