fizzlee998/nigeria-crime-intel-tracker

# Nigeria Crime Intelligence Tracker An agentic AI pipeline that automatically monitors Nigerian news sources, identifies crime-related headlines, extracts structured incident data using AI, and displays it on a live dashboard — with zero manual input after startup. ## What it does 1. **Fetches** live headlines from Nigerian news RSS feeds (Punch, Vanguard, Premium Times) 2. **Filters** them for crime relevance using keyword matching, to avoid wasting AI calls on irrelevant news 3. **Classifies** matching headlines using Google's Gemini API, extracting: - Crime type (robbery, kidnapping) - Location - Confidence level - One-line summary 4. **Deduplicates** against existing records before saving 5. **Stores** structured results in a SQLite database 6. **Displays** everything on a live, auto-updating web dashboard 7. **Repeats automatically** every hour via a background scheduler — no manual re-running required ## Why this exists Crime reporting in Nigerian news is scattered across dozens of outlets, in unstructured prose, with no central place to see patterns. This project is a prototype for turning that raw news flow into structured, searchable crime intelligence — a foundation for the kind of evidence-based crime analysis used by researchers, NGOs, journalists, and policy analysts. ## Tech stack - **Python** — core pipeline - **feedparser** — RSS ingestion - **Google Gemini API** (gemini-2.5-flash) — headline classification - **SQLite** — structured storage - **Flask** — live dashboard - **schedule** — automatic hourly re-runs ## Current scope (MVP) - Nigeria only - Two crime categories: robbery, kidnapping - Three news sources - Keyword-based pre-filtering (not yet ML-based) ## Roadmap - [ ] Expand RSS source coverage - [ ] Expand crime category taxonomy - [ ] Add maps/charts for geographic and trend visualization - [ ] Add downloadable reports (PDF/Excel) - [ ] Add source verification / cross-referencing before publication ## Author Built by Abdulafee …