Summary
- AI-powered email triage and digesting for forwarded inboxes
- Unique per-user forwarding addresses with Cloudflare Email Routing
- Email classification, summarization, and action-item extraction using Qwen models
- Human-in-the-loop inbox for approving or rejecting AI-recommended actions
- PostgreSQL + pgvector backend for persistence and semantic indexing
- Built with Next.js 16, TypeScript, and a modern dashboard UI
Built a hackathon project called Email Digest Agent: an AI-powered inbox for forwarded email. Instead of connecting directly to users’ mailboxes, the app gives each user a unique forwarding address and processes inbound messages through Cloudflare Email Routing, which keeps setup lightweight and avoids complicated integrations to multiple email provider and impossible OAuth approval. Once an email arrives, the system classifies it, summarizes it, extracts action items, and stores the result in a dashboard for quick review.
The product is designed around human-in-the-loop decision-making. Users can browse emails in an inbox view, inspect AI-generated summaries and recommendations, and approve or reject actions before anything is executed. Deployed on Alibaba cloud, combines Next.js 16, PostgreSQL with pgvector in backend, and Qwen Cloud models for classification, summarization, rule evaluation, and embeddings, giving the app both automation and control.
For the hackathon, I focused on making the key workflow practical and end-to-end: registration, forwarding address generation, inbound email handling, AI processing, notification registering, and a clean inbox UI. The result is a productivity app with PWA that turns messy forwarded mail into an organized, actionable digest.