Local-first browser extension
VectorMark
VectorMark is a local-first bookmark manager using TypeScript, IndexedDB, embeddings, and vector search to organize and retrieve saved websites.
Overview
VectorMark is a privacy-first bookmark manager that makes large bookmark collections easier to organize and retrieve. Instead of relying only on folders or exact keyword matches, it uses local embeddings and vector similarity to find conceptually related pages.
Bookmark data and the search index stay in the browser. This local-first architecture reduces external dependencies while keeping semantic retrieval responsive and private.
My contribution
Designed the local-first architecture and implemented bookmark ingestion, automatic organization, local embedding workflows, IndexedDB persistence, and HNSW-based vector retrieval in TypeScript.
Engineering highlights
- Semantic bookmark search using local embeddings
- Privacy-first storage and retrieval inside the browser
- IndexedDB persistence for bookmark and vector data
- HNSW indexing for efficient vector similarity search
Technology