hahhforest/pi-textbook

This open-source project is a structured, checkpoint-driven guide to building a Pi-style conversational agent from zero. The 15 checkpoints are the real differentiator: instead of a monolithic tutorial you abandon halfway, each checkpoint is a working milestone you can inspect, run, and break intentionally. That kind of incremental, verifiable structure is rare in agent-building education, which tends to be either 'clone this repo and run it' or a 400-page textbook with no runnable examples. The content is in Chinese, which narrows the audience but also means it is filling a genuine gap for Mandarin-speaking developers who are tired of translating English LLM tutorials. Reservation: the repo is early-stage, so some later checkpoints may be sparse or incomplete depending on when you look at it. If you are building agents in Python and want to understand the internals rather than just wrap an SDK, this is worth cloning and working through. -> Best for: AI engineer or ML researcher building agent systems from first principles