I built Recall, a background daemon that indexes your files, Gmail, Google Calendar, Google Drive, Canvas/Schoology, Notion, and cal.ai into a single local vector database — then lets you search everything with one query.

Nothing leaves your machine. Embeddings run locally via Ollama (or optionally Gemini/NIM). Sub-200ms semantic search over 1,000+ docs on M-series Mac.

Front-ends: Raycast extension, `recall` CLI, and an MCP server for AI assistants.

Would love feedback on the architecture and any ideas for new connectors.