Support ticket triage app

This repository is the starting point for an educational project in which computer science students will build a support ticket triage application. The planned application will use a large language model (LLM) from the backend to help analyze incoming support requests and support the triage process.

The codebase is intentionally a basic skeleton. It provides a small chat interface and server endpoint as a first step: before building the full triage workflow, students can use this mock chat to validate the connection between the application and an LLM provider.

Current status

The chat feature is currently a mockup implemented as an echo service. The client sends a message to /api/chat, and the server returns it in a confirmation response. No LLM provider is called, and no LLM configuration is needed to run this version.

This simple end-to-end flow is intended to make it possible to verify the application's client-to-backend connection first, then integrate and validate an LLM provider as an intermediate step, before implementing the complete support ticket triage feature. The triage route and LLM client are placeholders for that future work.

Sample data

The data/ folder contains two support-ticket datasets:

• 500 unlabelled tickets in support_tickets_500.csv and support_tickets_500.json, for exploring and developing the triage workflow.
• 100 labelled tickets in support_tickets_100_labelled.csv and support_tickets_100_labelled.json. These include triage annotations such as category, priority, and whether a ticket needs human review, making them useful for testing and evaluating triage approaches.

Run locally

Requires Node.js 20.9 or newer.

npm install
npm run dev

Open http://localhost:3033 and send a message. Both npm run dev and npm run start use port 3033.

The Send button is disabled when the message is empty or contains only spaces, and while a request is in progress.

The client posts { "message": "your text" } to /api/chat. The App Router handler in app/api/chat/route.ts returns { "answer": "Hello, you just said: your text" }. Invalid JSON or an empty/non-string message returns HTTP 400 with an error.

Checks

npm run typecheck
npm run build