recall-local

Search your own notes by meaning with local embeddings. Nothing leaves the browser.

• Search your notes by meaning with local embeddings
• A memory map of your notes with zoom, pan and note labels
• Notes stored on the device in IndexedDB
• Works on phones
• No server and no account

Try it

Live demo: https://recall-local.vercel.app

npm install
npm run dev

Open http://localhost:3000. Requires Node 22 or newer.

1. Press Try with sample notes, or add .md and .txt files.
2. Ask in your own words. Matches glow on the map.
3. Scroll to zoom, drag to pan, double click to reset.

Configuration

No environment variables. API keys are entered in the app and stay in the browser.

How it works

flowchart LR
  N[Notes] --> K[Chunker]
  K --> E[Embedding worker]
  E --> S[(IndexedDB)]
  Q[Query] --> E
  E --> R[Cosine ranking]
  S --> R
  S --> P[PCA 2D]
  P --> M[Memory map]
  R --> M

Loading

Added files are chunked, embedded in a worker and stored. A query is embedded and ranked against stored vectors. The map projects all vectors to two dimensions. Full diagrams and the module map are in docs/architecture.md.

Key concepts

Term	Meaning
Passage	A chunk of a note of roughly 70 words.
Embedding	A vector for the meaning of text, computed locally.
Semantic search	Finding by meaning: 'what to wear in rain' matches 'pack a rain jacket'.
PCA	Projection of high-dimensional vectors to two dimensions for the map.
IndexedDB	Browser storage where notes and vectors persist on the device.

Design system

Typography: Display and text, Manrope, weight 800 for titles; Code and labels, JetBrains Mono.

Token	Value	Use
night	#08110e	Page background
panel	#0f1b17	Surfaces
text	#e8f4ee	Primary text
muted	#7f9a8e	Secondary text
mint	#5cf2b0	Primary accent
amber	#ffc857	Note color

• The map is the hero: the notes become a place.
• Six note colors keep documents distinguishable.

Motion, components and rationale: docs/design-system.md.

Data flow and privacy

Data	Where it goes	Stored
Notes and vectors	Computed and stored in the browser	IndexedDB
Embedding model	Downloaded once from the model host, then cached	Browser cache
Queries	Stay in the browser	Not stored

Limits

• Only .md and .txt files.
• A small model is less precise on very short passages.
• IndexedDB can be blocked in private browsing.

Deployment

The app is fully client-side, so it can be hosted as static files.

• GitHub Pages: npm run deploy:pages builds a static export and publishes it to the gh-pages branch. Enable Pages from that branch; on a free plan the repository must be public.
• Vercel or any Node host: use the Deploy button above. No configuration is needed.

Documentation

Document	What it answers
docs/index.md	Map of all documentation
docs/architecture.md	Diagrams and modules
docs/spec/spec.md	Requirements and acceptance criteria
docs/spec/traceability.md	Requirement to code, test and evidence
docs/design-system.md	Tokens, motion, components
docs/glossary.md	Definitions
docs/evaluation.md	Self-assessment against a review rubric
docs/adr	Decision records

For AI agents and tools

• AGENTS.md defines the workflow and quality gates for agents and people.
• llms.txt is served at /llms.txt when deployed and points to the key documents.
• docs/spec/requirements.json is the machine-readable requirement list with status, files and tests.
• npm run verify is the single deterministic gate: typecheck, lint, traceability check, tests and build.

LLM integration: No language model generates text here. The embedding model runs locally and its output is numbers only, so there is no prompt injection surface.

Scripts

Script	Purpose
npm run dev	Development server
npm run build	Production build
npm run typecheck	TypeScript check
npm run lint	ESLint
npm test	Unit tests
npm run spec:check	Traceability gate
npm run verify	All of the above

Contributing

See CONTRIBUTING.md. Security reports: SECURITY.md.

License

MIT. Uses Transformers.js (Apache-2.0).