Show HN: I build a local archive for AI chats that are hard to export openrouter

Category: ai-ml

Tags: local-first, semantic-search, ai-chat, vector-database, mcp

Score: 6.8/10 (Innovation: 5, Technical: 7, Documentation: 8, Utility: 7)

ThreadShelf is a local-first archive and semantic search tool for AI conversations across six providers, including hard-to-export sources like OpenRouter and Google AI Studio. It normalizes exports into a common format, runs local embeddings and LanceDB vector storage, and offers a React UI, HTTP API, and MCP server for retrieval and optional continuation via local llama.cpp or OpenRouter.

Target audience: AI power users, developers, and researchers who want local control and cross-provider search over their conversation history

Repository: https://github.com/ChrystianSchutz/ThreadShelf · TypeScript · MIT · 14 stars

View on Hacker News