Analysis updated 2026-05-18
Give a coding agent a structured reading path before it writes code against the pi framework.
Look up which pi document covers the SDK, RPC layer, extensions, or sessions.
Use the included example questions to start a conversation with an AI assistant about building on pi.
| enderzcx/pi-docs-playbook | husu/loom | jackson-video-resources/markov-hedge-fund-method | |
|---|---|---|---|
| Stars | 120 | 120 | 120 |
| Language | — | — | Python |
| Setup difficulty | — | moderate | easy |
| Complexity | 1/5 | 2/5 | 3/5 |
| Audience | developer | developer | developer |
Figures from each repo's GitHub metadata at analysis time.
pi-docs-playbook is a documentation navigator and reading guide for pi, an AI agent framework built by earendil-works. It is not a tutorial for pi and does not modify pi in any way. Instead, it mirrors pi's official Markdown documentation, reorganizes it by topic, and tells AI coding agents like Claude Code, Codex, or Cursor which documents to read depending on what a developer is trying to build. The problem it addresses is that pi's documentation is dense and technically detailed. When an AI coding agent is handed the pi codebase directly, it often guesses at behavior from memory, misses important documents, confuses different subsystems, or cannot tell what pi already handles from what the developer's application must handle separately. This repository adds a navigable reading path on top of the official docs. The repository contains an AGENTS.md file with rules for coding agents, a PROMPT.md with ready-to-paste instructions for starting a session, a source folder containing mirrored pi documentation at its original file paths, a catalog folder organizing those docs by topic, usage instructions, and example questions you can ask an agent. For a human developer, the workflow is to open PROMPT.md, hand the repository to a coding agent, and pick an example question to start from. For an AI agent, the workflow is to read AGENTS.md first, then consult the task reading matrix to identify which source files are relevant to the current task. The README is written in Chinese and includes a reminder that pi handles the agent loop, tool calling, sessions, extensions, and related infrastructure. Business logic, approval flows, idempotency, error recovery, and domain rules remain the responsibility of the application built on top of pi. The repository was last mirrored from the upstream pi project in June 2026.
This repo across BitVibe Labs
Verify against the repo before relying on details.