Hackathon-ready MVP for Mnemosyne: The Architecture of Memory.

• Upload image/text/audio files
• Analyze files with OpenAI (or deterministic fallback if no API key)
• Cluster memories by semantic/emotion similarity
• Procedurally generate room/exhibit layout
• Render a first-person explorable 3D museum in browser
• Show contextual plaque overlays + narration via browser speech synthesis

npm install
npm run dev

Open http://localhost:3000.

Create .env.local:

OPENAI_API_KEY=your_key_here
OPENAI_MODEL=gpt-4.1-mini

If no key is set, the app still works using fallback analysis.

• / landing
• /upload upload + pipeline trigger
• /museum/[sessionId] generated museum

• POST /api/upload
• POST /api/analyze
• POST /api/cluster
• POST /api/generate-layout
• POST /api/build-scene
• GET /api/build-scene?sessionId=...
• POST /api/pipeline (end-to-end)

• W/A/S/D move
• Shift sprint
• Hold mouse button + drag to look

npm run test
npm run test:e2e