This is a submission for the Postmark Challenge: Inbox Innovators.
What I Built
Postmark Kanban Inbox is a simple ticket management system that turns incoming emails into categorized, actionable Kanban cards - powered by Postmark's inbound webhooks with facebook/bart-large-mnli, a powerful language model that can categorize messages without any fine-tuning.
Demo
You can run the full app locally by following the steps in the README.
The README includes everything you need to:
- Start both the frontend and backend servers
- Set up a Postmark inbound webhook (with
ngrok) - Send test emails or use
curl/JSON for local testing
Code Repository
You can explore the full project source code here:
🔗 https://github.com/iefendiev/postmark-kanban-inbox
The repository contains:
- A
server/folder with the Express + Prisma backend - A
client/folder with the React + Tailwind frontend - A comprehensive README with setup instructions and usage notes
How I Built It
This project was built with a clean separation between backend and frontend, allowing full control over both Postmark webhook handling and UI rendering.
Backend
- Node.js + Express for setting up the webhook and exposing RESTful endpoints
- Prisma + SQLite for a fast, file-based data layer — ideal for local testing
- Postmark Inbound Webhook to receive and parse raw email data
-
Hugging Face Inference API using the
facebook/bart-large-mnlimodel for zero-shot classification of ticket category and priority
Frontend
- Vite + React + TypeScript for a fast, modern development environment
- Tailwind CSS + Shadcn UI for a polished and consistent interface
- TanStack Query to manage API data and caching
- @dnd-kit for Kanban-style drag-and-drop ticket management
- Confetti animation triggers when a ticket is marked as resolved 🎉
I usually stick to frontend work, but this challenge pushed me to try new things, and I’m really glad it did.
Top comments (3)
nice projecttt, i created something very similar with support tickets but doesnt include a kanban board 😅 all the best 🔥
Love how you brought AI categorization and Postmark together with Kanban flow. How accurate did you find the message classification in real use?
Thanks! It was actually pretty accurate, it picks up context really well. For example, if the mail says something like “ASAP” or “urgent,” it reliably classifies the priority as high, even without exact keyword matches.