🏛️ CivicSense — AI-Driven Municipal Complaint System

An intelligent civic platform that uses NLP, Computer Vision, and Geospatial Analytics to automatically classify, prioritize, and route citizen complaints — modeled after Mumbai's BMC Ward system.

---

🔗 Live Deployment (Production)

Important

The backend is hosted on Hugging Face Spaces (16GB RAM) to handle large AI models, while the frontend is on Vercel.

• 🌍 Live Portal: Visit CivicSense
• ⚙️ Backend API: https://frank10101-civicsense.hf.space
• 🚦 API Health: Check Status

---

📌 Project Overview

Citizens submit complaints via text, voice, or image. Our AI Neural Suite instantly orchestrates the following:

1. Categorization: Automatically identifies the department (Roads, Waste, Water, etc.) using BART-MNLI.
2. Urgency Scoring: Assigns priority (Critical, High, Medium, Low) using DistilBERT sentiment & keyword analysis.
3. Visual Triage: Detects civic issues in photos (e.g., potholes, garbage) using YOLOv8.
4. Voice Recognition: Transcribes audio complaints via OpenAI Whisper.
5. Duplicate Detection: Prevents multiple reports for the same incident using MiniLM-L6-v2 semantic embeddings.
6. Geospatial Intelligence: Maps locations to Mumbai's BMC Wards using point-in-polygon geometry.

---

🛠️ System Architecture

Layer	Technology
Frontend	React 19 + TypeScript + Vite + Tailwind CSS
Backend	FastAPI (Python 3.10)
Database	Supabase (PostgreSQL + pgvector)
AI/ML	HuggingFace, YOLOv8, Whisper, SentenceTransformers
Real-time	WebSockets (Dashboard Live Feed)
Deployment	Docker (Backend) + Vercel (Frontend)

---

📖 Essential User Guide

🔑 Demo Credentials

Role	Email	Password
Citizen	citizen@example.com	password123
City Admin	admin@city.gov	adminpassword
Roads Officer	roads@city.gov	password123

---

🚶 The "Quick-Start" Flow

1. Reporting an Issue (Citizen)

1. Login as a Citizen.
2. Click "Report Issue" from the home command center.
3. Describe the issue (Type it or click the Mic to speak).
4. Location: Click the map to pin the exact spot.
5. Attachments: Upload a photo for AI visual verification.
6. Submit: Watch the AI analyze and route it in < 2 seconds.

2. Managing the City (Officer/Admin)

1. Login as an Official.
2. View the Live Triage Map to see hotspots across Mumbai.
3. Open a complaint to see AI insights (Urgency, Category, Visual Tags).
4. Resolve: Click "Resolution Proof" -> Enter your work note -> Status becomes Resolved.

3. Verifying Results (Citizen)

1. Go to "Track Status".
2. Click "View Resolution Proof" to see what the officer documented.

---

📸 Visual Walkthrough

1. Citizen Experience

Figure 1: High-fidelity municipal landing page with role-aware portals.

Figure 2: AI-driven reporting flow with geospatial map integration.

Figure 3: Personal tracking hub with real-time AI triage status.

2. Administrator Executive Suite

Figure 4: Executive Command Center with Live Heatmap and Triage Feed.

Figure 5: Verifiable resolution proof system for municipal transparency.

---

⚙️ Local Development Setup

1. Backend

cd backend
pip install -r requirements.txt
# Set .env with SUPABASE_URL, SUPABASE_KEY, JWT_SECRET
uvicorn main:app --reload

2. Frontend

cd frontend
npm install
# Set VITE_API_URL in .env
npm run dev

---

👥 Team — DevCraft

Built with ❤️ for Mumbai's Digital Transformation.