Inspiration
Every day, millions of citizens worldwide face broken roads, overflowing drains, power outages, and public safety hazards — but reporting them is frustrating, slow, and often goes nowhere.
Whether you're in India, Nigeria, Indonesia, or Brazil — civic complaint systems are broken everywhere. Paper forms, zero transparency, no follow-up, no accountability.
What We Built
VYAVASTHA is a full-stack AI-powered civic complaint platform that bridges the gap between citizens and local government.
For Citizens:
- Submit complaints with AI-verified real photos
- AI auto-classifies category, severity and routes to correct department
- Real-time popup notifications when complaint status changes
- Track complaint journey: Submitted → Assigned → In Progress → Resolved
- Upvote community complaints for collective pressure
- 25 language support for global accessibility
For Government Officials:
- Structured dashboard to assign, escalate and resolve complaints
- AI generates professional ATR (Action Taken Report) instantly
- Miro board embedded for department planning and coordination
- Analytics dashboard showing complaint trends by category and area
- Auto-escalation alerts for overdue complaints
Image Authenticity Verification (3 Layers)
No fake complaints. No AI-generated evidence. Real photos only.
- Layer 1 — EXIF Metadata Check: Real camera photos contain metadata: camera make, model, GPS coordinates. AI images have none of this.
- Layer 2 — Sightengine AI Detection: Sightengine API scans every upload and rejects AI-generated images instantly.
- Layer 3 — Sanity Check: File size between 20KB–15MB, minimum 400x400px dimensions, only JPG/PNG/WebP formats accepted.
AI Features (Powered by Featherless AI)
When a citizen types a complaint, Featherless AI instantly:
- Detects category: Road, Water, Electricity, Sanitation, Public Safety
- Routes to correct department: PWD, Water Board, BESCOM, BBMP, Police
- Assigns severity: Low / Medium / High / Urgent
- Generates a professional summary
- Writes a complete ATR report for officials
- Predicts expected resolution time
Real-Time Notifications (Socket.io)
When an official updates a complaint status, the citizen's portal instantly shows a color-coded popup notification — no refresh needed.
- 🟢 Green — Complaint Resolved
- 🔵 Blue — Complaint Assigned
- 🟠 Orange — Work In Progress
- 🔴 Red — Complaint Rejected
Runs automatically every midnight. Citizens are notified instantly when a complaint is escalated to senior authorities.
How We Built It
Frontend: React 18, Vite, Tailwind CSS, Zustand, i18next
Backend: Node.js, Express, MongoDB, Mongoose, JWT, Socket.io, node-cron, Multer
AI & Sponsor Tools:
- Featherless AI — LLM complaint classification and ATR generation
- Sightengine — AI-image and authenticity detection
- Miro — Embedded planning board in government admin dashboard
Infrastructure: Docker
Challenges We Faced
- Integrating Socket.io correctly inside Docker containers
- Building 3-layer image verification fast enough for real-time use
- Balancing AI response quality vs speed
- Implementing complete i18n for 25 languages in a short timeframe
- Building a production-grade role-based system as a student team
What We Learned
- How to integrate LLM APIs for structured real-world outputs
- Real-time architecture with Socket.io user rooms
- Image authenticity detection using metadata and ML APIs
- SLA-based automation with cron jobs
- Building for real civic use cases that matter to people
Impact
VYAVASTHA makes civic accountability visible, complaints actionable, and governance more responsive — for any city, any government, any citizen worldwide.
Log in or sign up for Devpost to join the conversation.