goniszewski/grimoire

What it solves

Grimoire is a local-first bookmark manager designed for technical resources. It solves the problem of organizing and retrieving large volumes of saved links by extracting readable content from web pages and providing advanced search capabilities, all while keeping data private and stored on the user's own machine.

How it works

The system consists of a React-based frontend and a Bun/Hono daemon that manages a local SQLite database. When a URL is saved, the daemon executes a progressive ingestion pipeline: it fetches the content, extracts readable text and metadata (from sources like GitHub, YouTube, and PDFs), and optionally uses AI providers to generate summaries, tags, and semantic embeddings for better searchability.

Who it’s for

It is built for people who save a high volume of technical resources and need a private, local way to organize and search through them using both keywords and meaning.

Highlights

• Local-First Privacy: Data is stored in a local SQLite database and files, with no requirement for cloud storage.
• Hybrid Search: Combines SQLite FTS5 keyword search with semantic embedding search.
• AI-Powered Organization: Optional integration with providers like OpenAI, Ollama, and Anthropic for automated summarization and tagging.
• Broad Extraction: Specifically handles technical content from GitHub repositories, issues, StackOverflow, and YouTube transcripts.
• Integration Ready: Includes a browser bookmarklet for quick capture and an MCP (Model Context Protocol) interface for local AI clients.