Turn your complaint into a formal letter and escalate to regulators if ignored. Any country. Any sector.
🔗 Live App: https://redressamd.vercel.app 🤗 HuggingFace Space: https://huggingface.co/spaces/lablab-ai-amd-developer-hackathon/redress
Most people have a legitimate complaint against a bank, telco, utility provider, landlord, or government agency but have no idea how to formally escalate it. They either give up, send an ineffective email, or don't know which regulatory body to contact.
Redress solves that. It's an AI agent that:
- Understands your complaint through a short conversation
- Drafts a professionally structured complaint letter
- Identifies the right channel and regulatory body for your country and sector
- Generates an escalation letter to regulators if your complaint is ignored
- Lets you download every letter as a PDF
📺 Demo Video: Watch here
Stage 1 — Understand The agent asks up to 3 clarifying questions to gather the key details: organization name, nature of the issue, date it occurred, and any prior contact attempts.
Stage 2 — Draft Once enough context is gathered, the agent generates a formal complaint letter with the correct recipient, recommended channel, and relevant regulatory body details.
Stage 3 — Escalate If the user indicates their complaint was ignored, the agent generates an escalation letter addressed directly to the regulatory body, including step-by-step filing instructions.
| Layer | Technology |
|---|---|
| Framework | Next.js 16 App Router |
| Language | TypeScript |
| Styling | Tailwind CSS |
| Auth | NextAuth.js + Google OAuth |
| Database | MongoDB + Mongoose |
| State Management | Zustand |
| Web Search | Tavily |
| PDF Generation | jsPDF |
| Deployment | Vercel + HuggingFace Spaces |
Note: This project was built for the AMD Developer Hackathon 2026. AMD Developer Cloud credits were applied for on May 5th but were not received before the submission deadline. The agent logic is fully compatible with AMD's OpenAI-compatible API and can be switched by updating two environment variables.
- Email + password registration and login
- Google OAuth
- Create and manage multiple complaint sessions
- Multi-turn conversational agent
- Formal complaint letter generation
- Regulatory body detection with live web search via Tavily
- Escalation letter generation
- PDF download for all letters
- Complaint history with stage tracking
- Admin dashboard
- Mobile responsive
- Node.js 18+
- MongoDB URI
- Groq or AMD Developer Cloud API key
- Tavily API key
- Google OAuth credentials (optional)
git clone https://github.com/AMDHACKATHON/Redress
cd Redress
npm installCreate a .env.local file in the root:
MONGODB_URI=
AMD_API_KEY=
AMD_API_URL=https://api.groq.com/openai/v1/chat/completions
NEXTAUTH_SECRET=
NEXTAUTH_URL=http://localhost:3000
GOOGLE_CLIENT_ID=
GOOGLE_CLIENT_SECRET=
TAVILY_API_KEY=
npm run devOpen http://localhost:3000
app/
(auth)/ — Login and register pages
(dashboard)/ — Protected dashboard, complaint chat, profile
admin/ — Admin dashboard
api/ — All API routes (auth, complaints, letters, profile)
components/
letter/ — LetterDisplay, PDFDownloadButton
lib/
models/ — Mongoose models
mongodb.ts — DB connection
api.ts — Axios instance
store.ts — Zustand store
search.ts — Tavily search utility
types/
index.ts — TypeScript interfaces
| Name | Role | GitHub |
|---|---|---|
| Ezekiel Samuel | Lead Engineer | @samkiell |
| Zabdiel Anyaogu | Frontend + Demo | @fwesh |
Built for the AMD Developer Hackathon 2026 on lablab.ai Track: AI Agents & Agentic Workflows Prize pool: $10,000 + AMD Radeon AI PRO R9700 GPU
Redress — AMD Developer Hackathon 2026