Live demo: neuron-map-six.vercel.app
NeuronMap turns study notes into a living knowledge universe. Upload Markdown notes, AI extracts the concepts and how they relate, and everything renders as an animated cosmic graph — stars (core topics) orbited by planets (sub-topics) orbited by asteroids (details). Upload more notes and shared concepts are linked across documents with golden threads.
- AI knowledge extraction — concepts are tiered into star / planet / asteroid by heading depth and connected with typed, weighted relationship edges (contains / depends / related / contrast).
- My Universe — every document is a galaxy on one canvas. Zoom out for a Spore-style overview with golden lanes showing how many concepts two documents share; zoom in and planets, asteroids and their labels appear progressively.
- System drill-down — click any star to warp into its system: the orbiting bodies on the left, the knowledge panel on the right, with breadcrumbs all the way back up (planet → star → universe).
- Cross-document linking — the same concept appearing in two documents is detected at extraction time and linked (never merged), marked with a golden ring.
- Curate your universe — delete a document (with all its knowledge) or any single node; deleting a star removes exactly its contains-subtree, nothing else.
- Staged extraction progress — live progress bar driven by a status polling API.
- Private by design — per-user storage with row-level security; files are read via the SDK, never by URL; all AI calls stay server-side.
Markdown notes → AI extraction (nodes + edges) → PostgreSQL → Canvas universe
Markdown is the Phase 1 input on purpose: headings encode hierarchy deterministically, extraction is cheap and verifiable. PDF can arrive later as a PDF-to-Markdown adapter feeding the same pipeline.
| Layer | Choice |
|---|---|
| Framework | Next.js 16 App Router |
| UI | React 19 + Tailwind CSS v4 |
| Auth | Supabase Auth |
| Database | Supabase PostgreSQL |
| ORM | Prisma 7 + @prisma/adapter-pg |
| Storage | Supabase Storage (private + RLS) |
| AI | Anthropic Claude API |
| Rendering | Custom Canvas + requestAnimationFrame |
| Deploy | Vercel (Sydney) |
npm install
npx prisma generate
npx prisma db push
npm run devCreate .env.local with your own keys (never commit it):
NEXT_PUBLIC_SUPABASE_URL=
NEXT_PUBLIC_SUPABASE_ANON_KEY=
DATABASE_URL=
DIRECT_URL=
ANTHROPIC_API_KEY=npm test # unit tests (extraction transform, subtree deletion, cross-doc matching)
npm run build # type check + production build- Skill-tree learning view (visual direction settled) as a switchable second view.
- Async extraction jobs for very large documents.
- Semantic cross-document merging (embeddings + pgvector) behind the existing link table.
- AI chat over the knowledge base.
- Bring-your-own-API-key.
中文说明见 README.zh.md