What is weloom?

WeLoom is a Python tool that turns your personal chat history into a local-first memory system and AI companion. You export conversations from messaging platforms, feed them into WeLoom, and can then chat with an AI that responds in the style of the person whose messages were in those exports, a digital clone built from real conversation data. The pipeline runs in five stages: import (load chat JSON files into a local SQLite database), index (build a full-text search index using SQLite's FTS5 engine), distill (extract reply samples and build a personality profile), analyze (generate long-form reports summarizing the relationship and communication patterns), and chat (run a conversation session where the AI uses the indexed memories and personality profile to reply authentically). All data, the database, profile, reports, and cache, is stored locally in a storage folder. The AI model calls out to an OpenAI-compatible API only when you explicitly run commands that need it. Search is done with SQLite FTS5 (a built-in full-text search engine), with fallbacks to keyword matching and LIKE queries if FTS5 is not available. The system retrieves relevant past messages as evidence before generating each reply, and you can pass a flag to see the retrieval trace for any response. It currently supports chat exports in WeChat (via WeFlow) and QQ (via QQChatExporter V5) JSON formats. The only hard dependencies are Python 3.10 and SQLite, both standard. An AI API key is required only for report generation and chat. The full README is longer than what was shown.