Skip to content
unlearndevPublic

About

The Fixlist app built in the Test AI Features with Evals course

Resources

Stars

0 stars

Watchers

0 watching

Forks

Repository files navigation

Fixlist

The repairs inbox for a small lettings agency. Tenants report problems by email or a web form, staff type up phone calls, and an AI triage feature reads each report and suggests a category, an urgency and a job summary for the contractor. Staff check the triage before anyone is booked.

It's the demo project for the Unlearn course "Test AI Features with Evals". Every property, tenant, phone number and report in it is made up. Phone numbers are in Ofcom's range for drama (07700 900xxx) and email addresses use example.com.

Running it

Node 22.18 or later (it runs the TypeScript files directly).

npm install
npm run dev        # the inbox, on http://localhost:5173
npm test           # unit tests
npm run typecheck

Triaging a report calls Claude, so it needs an Anthropic API key in your environment:

export ANTHROPIC_API_KEY=...   # from the Claude Console
npm run triage -- T-1022       # triage one report from the log again

What's where

  • src/triage/: the AI feature. prompt.ts has the system prompt (the agency's repairs policy) and builds the message for each report, triage.ts calls Claude with structured output, config.ts picks the model.
  • data/triage-log.jsonl: two months of usage. Every report, what the triage said, and what staff saved after checking it.
  • data/properties.json: the agency's properties and tenants.
  • server/: the small API, mounted on Vite's dev server.
  • src/views/: the inbox.

About

The Fixlist app built in the Test AI Features with Evals course

Resources

Stars

0 stars

Watchers

0 watching

Forks

Releases

Packages

Contributors

Languages