What is git-log-humanizer?

Git Log Humanizer is an automation workflow built on a tool called n8n, which lets people build automated pipelines by connecting visual blocks instead of writing a full application. This project watches a GitHub repository for push events, meaning every time someone commits new code, and takes the raw commit messages developers actually write, which are often short and informal, like fixed bug or added some stuff to script. It rewrites those messages into clear, professional changelog entries and updates project documentation automatically. The workflow has four parts. First, a data intake step listens for GitHub push events and pulls out the individual commit messages. Second, those messages go to a small, fast AI language model, run through a service called Groq using the Llama 3.1 model, instructed to return its answer in a strict, predictable data format rather than free flowing text. Third, a documentation step takes that structured output and writes it into the repository's CHANGELOG.md file by committing the change through GitHub's API. Fourth, a logging step records each change into a database table for later tracking. To handle testing without cluttering the real project's commit history, the workflow includes a fallback: if it runs without any real commit data present, it substitutes a small set of example developer style messages so a developer can test the pipeline safely. It also uses a structured output parser to stop the AI model from wrapping its answer in extra formatting that would break the parsing, and it updates its destinations, such as the changelog, a database, and chat notifications, independently, so a slow notification service does not hold up the rest of the update. Setting it up requires an active n8n installation, either self hosted or through n8n's cloud service, plus API access to Groq for the AI processing, a GitHub personal access token with write permission, and optionally a Discord or Gmail webhook for notifications. You install it by importing a provided workflow file into your n8n workspace, filling in your credentials for each connected service, and turning the workflow on. The output changelog data is organized into three categories: features, bug fixes, and performance tweaks, each defaulting to the word None if nothing of that type was found. The project is released under the MIT License, and its author invites others to fork it for their own changelog pipelines.