12PHY External Assessment Tutor

A small web app that gives your QCE Year 12 Physics class an AI tutor scoped to the Unit 3 & 4 external assessment (Gravity and motion, Electromagnetism, Special relativity, Quantum theory). Static frontend + a single serverless function that proxies chat requests to the Claude API, so your API key never reaches students' browsers.

Features

• Concept Q&A mode — explains Unit 3/4 concepts, checks understanding, works through examples.
• Practice exam mode — generates QCE-style multiple choice / short response / extended response questions with mark allocations, and marks student answers point-by-point in the style of a QCE Instrument-Specific Marking Guide (ISMG).
• Formula sheet — a quick-reference panel of the formulas used across Unit 3/4, also fed to the AI so its notation stays consistent.
• Optional class access code — a shared passphrase so a leaked URL doesn't let random visitors burn through your API budget.

This is a study-support tool built from general QCE Physics syllabus knowledge — it is not an official QCAA resource. Tell students to cross-check anything mark-critical against the current QCAA syllabus and the authorised formula and data booklet.

Project structure

index.html          Main page
style.css            Styling
app.js               Chat UI logic (client-side)
api/chat.js          Serverless function that calls the Claude API
data/content.json    Topic list + formula sheet (used by both the UI and the AI's system prompt)

Deploy (Vercel — recommended)

1. Push this repo to GitHub (already done if you're reading this from the repo).
2. Go to vercel.com, "Add New Project", import this repo. No build step or framework preset is needed — Vercel will serve the static files and deploy api/chat.js as a serverless function automatically.
3. In the Vercel project's Settings → Environment Variables, add:
   • ANTHROPIC_API_KEY — your Anthropic API key. Required.
   • CLAUDE_MODEL — optional, defaults to claude-sonnet-5. Set to claude-haiku-4-5-20251001 for lower cost at higher class volumes.
   • CLASS_ACCESS_CODE — optional shared passphrase for your class.
4. Deploy. Share the resulting URL with your class, plus the access code (if you set one) separately, e.g. verbally or via your LMS — not in the same channel as the link.

Local development

You need the Vercel CLI to run the serverless function locally (a plain static server won't execute api/chat.js):

npm install -g vercel
cp .env.example .env   # fill in ANTHROPIC_API_KEY
vercel dev

Then open the printed local URL.

Cost & usage notes

• Every message round-trip is one Claude API call billed to your key. There's no built-in per-student rate limiting — the access code deters casual misuse, but a determined user with the code could still send many requests. For a single class this is normally a non-issue, but keep an eye on your Anthropic usage dashboard, especially in the lead-up to the exam when usage will spike.
• Conversation history lives only in the browser tab (not persisted server-side or shared between students).

Customising the content

Edit data/content.json to adjust the topic list or formula sheet — both the sidebar UI and the AI's system prompt (built in api/chat.js) read from this one file, so there's no need to edit the prompt separately for content changes.