habeneyasu/disaster-preparedness-mvp

A localized, multi-modal MVP Disaster Preparedness System for Ethiopia. Integrates Generative AI (Hugging Face NLP) & Predictive AI (Scikit-Learn ML) with a decoupled FastAPI backend, local SQLite database, and interactive Folium maps under a unified Gradio dashboard UI. Engineered for constraint-heavy, offline deployment. # Disaster Preparedness MVP Multi-modal disaster triage for Ethiopia: field reports are summarized (BART), classified for risk (RandomForest), mapped (Folium), and audited in SQLite. Exposed via FastAPI and a Gradio dashboard. ## Stack | Layer | Technology | |-------|------------| | API | FastAPI, Pydantic, Uvicorn | | UI | Gradio (`/ui`) | | NLP | Transformers, PyTorch CPU (`facebook/bart-base`) | | ML | Scikit-Learn, Pandas | | GIS | Folium | | Storage | SQLite (`data/query_log.db`) | | Tooling | uv, Docker Compose | ## Quick start ```bash uv sync uv run uvicorn app.main:app --reload --host 0.0.0.0 --port 8000 ``` | URL | Purpose | |-----|---------| | localhost | Gradio dashboard | | localhost | OpenAPI | | localhost | Health check | ## API ```bash curl localhost curl -X POST localhost \ -H "Content-Type: application/json" \ -d '{ "district": "Gambela Town", "hazard_type": "flood", "raw_report": "Heavy seasonal downpours caused river overflow and flooded residential lowlands." }' curl "localhost" ``` ## Pipeline ```text POST /ui or POST /api/summarize → NLP (summary) → ML (risk + confidence) → GIS (risk_map.html) → SQLite (query_log) ``` ## Docker ```bash docker compose up --build ``` ## Data | File | Role | |------|------| | `data/districts_data.csv` | District features (source, tracked in git) | | `data/disaster_model.pkl` | Trained classifier (generated) | | `data/query_log.db` | Audit log (generated) | | `data/risk_map.html` | Latest map output (generated) |