PocketFlow-Tutorial-Codebase-Knowledge

This project is a Python tool that reads a GitHub repository and generates a written tutorial explaining how the code works, aimed at beginners. You give it a repository URL (or a local folder), and it crawls the code, identifies the key concepts and components, then writes a structured guide that walks through the codebase in plain language.

The output includes diagrams and explanations of how the main pieces of the code relate to each other. The project includes a gallery of pre-generated tutorials for well-known open-source projects, so you can see the style of output before running it yourself.

Under the hood the tool is built on PocketFlow, a small 100-line framework for running AI agent workflows. You configure which AI model to use by setting environment variables. The default is Google Gemini Pro 2.5, but you can point it at other providers or a local model running through Ollama. A caching layer stores AI responses so repeated runs do not re-query the model for the same code.

To run it, you clone the repo, install Python dependencies, set your API key, and then call the main script with a repository URL. Optional flags let you filter which file types to include, exclude directories like tests, set a file size limit, and choose an output language other than English. The generated tutorial lands in an output folder as markdown files.

A hosted version is also available online where you can paste a GitHub link without installing anything locally.