ai-data-extraction

Why we included this project

Most people never look at what their coding assistant has saved about them, but that history is exactly the kind of data you need to build a fine-tuning corpus. This toolkit digs through the local storage of Cursor, Codex, Claude Code, Windsurf, Trae, Continue, Gemini CLI, and OpenCode and normalizes the chat messages, code context, diffs, tool calls, and timestamps into clean JSONL. The records keep the full conversation structure rather than just the text, which matters if you want to study how models behave on real developer workflows or assemble training data from your own machine. It runs on the Python standard library alone, so there is no dependency setup to fight through. If you want to recover or archive your own assistant history, this is a practical starting point.