A "Command Center" view of Montgomery's public safety trends and AI-driven analysis.
SafeMGM turns raw public safety records into a single, judge-ready dashboard you can query in plain English—combining local 911 + crime datasets, an interactive safety map, trend analytics, and news context enrichment.
Built for the World Wide Vibes Hackathon (GenAI.Works Academy) by Abdullah Malik.
| AI Analyst Terminal | Geospatial Intelligence Map |
|---|---|
Cities publish important safety data, but it’s often shipped as static CSV/GeoJSON files—hard to interpret quickly, and missing the “what’s happening right now” context.
SafeMGM bridges that gap:
- Citizens can ask natural-language questions (no spreadsheets).
- Community leaders can quickly spot patterns (map + trends).
- Judges can evaluate a complete end-to-end GenAI product: retrieval/context building → streaming chat → real-time enrichment → graceful fallbacks.
- AI Safety Analyst (streaming chat): Ask questions like “Any spikes in 911 calls?” or “Safety near downtown” and get data-cited answers.
- Geospatial “Command Center” map: Leaflet map with severity-coded incidents (from local GeoJSON).
- Trend analytics: 12-month chart comparing 911 call volume vs. crime incidents.
- News enrichment: A live news panel for the active query (Bright Data / NewsAPI), with graceful degradation to sample articles so demos never break.
- Open dashboard → point out stats cards, trend chart, map, AI Terminal, and news feed.
- Ask the AI Terminal:
- “Compare crime across neighborhoods”
- “Any unusual spikes in 911 calls?”
- “Tell me about safety near downtown”
- Change the question and show the News Panel updating with the same query context.
- Explain the reliability story:
- If Gemini key is missing,
/api/chatreturns a clear 503 configuration error. - If news keys are missing, the app returns sample news (demo still works).
- If Gemini key is missing/rate-limited,
/api/statsfalls back to a static safety score so the dashboard still loads.
- If Gemini key is missing,
- Next.js 15 App Router UI in
src/app/page.tsx - Local datasets in
src/data/911-calls.csvcrime-stats.csvcrime-mapping.geojson
- Server-side data loader + context builder in
src/lib/montgomery-data.ts- Builds a bounded “AI context” summary (top crimes, trends, recent incidents) for safe prompting.
- AI streaming via Vercel AI SDK + Google Gemini in
src/app/api/chat/route.ts- Streams responses to the client using the
useChathook.
- Streams responses to the client using the
- News enrichment via
src/app/api/news/route.ts+src/lib/brightdata.ts- Primary: NewsAPI.org (no credit card)
- Fallback: Bright Data SERP API
- Fallback of last resort: sample news (keeps the demo working)
GET /api/stats: totals + top crime type + safety score (AI-powered when available)GET /api/trends: monthly trend series for chartsGET /api/data: dataset access (supports filters)?dataset=911|crime|mapping- Optional:
limit,neighborhood,dateStart,dateEnd,incidentType
GET /api/news?q=...: news articles for the News Panel (always returns 200 with{ articles: [] }on failure)POST /api/chat: streaming AI chat
- Framework: Next.js 15 (App Router), React 19, TypeScript
- UI: Tailwind CSS + shadcn/ui
- AI: Vercel AI SDK (
ai,@ai-sdk/react) + Google Gemini (@ai-sdk/google) - Mapping: Leaflet + react-leaflet (CartoDB Dark Matter tiles)
- Charts: Recharts
- Data parsing: PapaParse
- Deploy: Vercel (
maxDuration: 60on chat route)
- Node.js 20+ recommended
npm install
npm run devThen open http://localhost:3000.
Copy .env.example to .env.local and fill what you have.
GOOGLE_GENERATIVE_AI_API_KEY: required for/api/chat(Gemini streaming)
NEWS_API_KEY: enables NewsAPI.org results (recommended for easy demos)BRIGHT_DATA_API_TOKEN+BRIGHT_DATA_SERP_ZONE: enables Bright Data SERP fallback
If no news keys are provided, SafeMGM returns sample news to keep the UI functional for demos.
- Track: Public Safety & Community Resilience (GenAI for Social Good)
- Challenge: Transforming static municipal data into real-time, citizen-centric insights.
SafeMGM is designed with a city-agnostic architecture, meaning it can be deployed for any city that provides CSV or GeoJSON open data (standard for 3,000+ US municipalities).
- SaaS for Small-Mid Cities: Many cities lack the budget for custom safety dashboards. SafeMGM can be offered as a "Dashboard-as-a-Service."
- Real Estate Integration: APIs for neighborhood safety scores can be sold to real estate platforms.
- Enterprise Enrichment: Insurance companies can use the trend analytics for risk assessment.
- Uses local datasets stored in
src/data/(CSV/GeoJSON) and summarizes them for analysis. - Important: The app provides informational insights only; it does not predict individual behavior or provide emergency services. Always call 911 in an emergency.
Abdullah Malik
- GitHub:
https://github.com/AbdullahMalik17 - Repo:
https://github.com/AbdullahMalik17/WW_Hacathan