Bao

Bao is a Python personal AI assistant focused on persistent memory across sessions, a closed-loop experience engine, background subagents, and MCP tools. The goal is to let the assistant remember preferences, project context, and past failures instead of starting over every time.

I built the core system independently with help from Codex and Claude Code. The desktop app, CLI, Docker deployment, and messaging channels share the same assistant core, avoiding separate implementations for each entry point.

Technically, Bao uses PySide6 / QML, LanceDB, provider abstraction, MCP, skills, Docker Compose, and pytest. The provider layer covers OpenAI-compatible APIs, Anthropic, Gemini, Codex OAuth, and related modes.

The repository and showcase site are public, and the visual system is one of the presentation strengths. The product is runnable and still being refined, so the public page does not include adoption metrics.