Vector Bookmark

Semantic recall for everything you've read. 100% local hybrid BM25 + vector search across your browsing history.

Vector Bookmark turns your browsing history into a searchable second brain — without sending a single byte to anyone else's server. While you read, a tiny local daemon (vbmd) running on your own machine indexes the pages you actually spend time on. When you need to find something later — a half-remembered article, a tutorial you skimmed last month, a Stack Overflow answer — you search by meaning, not just keywords. KEY FEATURES • Hybrid search: BM25 full-text + vector embeddings combined with Reciprocal Rank Fusion. Finds the page even when you forget the exact words. • 100% local by default: SQLite database on your disk. No accounts, no cloud, no telemetry. • Passive capture: pages you spend at least 10 seconds on (configurable) are auto-indexed. No clicks needed. • Manual indexing on demand: open the popup, pick a mode (full text, LLM summary, manual notes, or meta-only), tag, and confirm. • Tag-aware: assign tags from the popup, get LLM-suggested tags from the existing taxonomy, browse a tag cloud, filter searches by tag. • YouTube-aware: extract the transcript or top comments of a video as the indexed content. • Timeline view: see your reading volume per day, drill into any day's pages. • Omnibox: type @recall <query> in the address bar to search. • Privacy-first: incognito tabs are never touched, password fields are never read, your blacklist is yours. • Bring your own LLM (optional): point the daemon at any OpenAI-compatible endpoint for embeddings and summaries — local (Ollama, llama.cpp) or cloud, your choice. REQUIREMENTS You need to install and run the companion daemon `vbmd` on your machine. Setup instructions and the full source code are at https://github.com/drecchia/chrome-plugin-vector-bookmark. WHY <all_urls>? The single purpose of this extension is to remember pages across your entire browsing — so it must observe all sites you visit. You stay in control via the user-managed blacklist and the dwell threshold. PRIVACY Read the full privacy policy at https://raw.githubusercontent.com/drecchia/chrome-plugin-vector-bookmark/main/PRIVACY.md.