AI-Git-Bot

AI-Git-Bot automates repeatable engineering chores inside Gitea, GitHub, GitLab, and Bitbucket. It is self-hostable, AI-agnostic, and is described as a gateway rather than a script. The system is aimed at teams that want to handle routine software-development work through Git events they already generate.

Its workflows include writing structured issues from vague bug reports, reviewing pull requests with inline and summary comments, and keeping follow-up questions in the PR thread with session memory. It can also generate and run end-to-end tests per pull request using Playwright, deploy a per-PR preview environment, post the report, and tear the preview down when the PR closes. Another workflow reads an assigned issue, drafts the change in a workspace, validates it with the project’s own build tooling, and opens a PR for review. A separate workflow generates white-box unit tests for a PR diff, runs them with the project’s runner, commits them to the PR branch, and posts pass/fail and coverage reports. The bot also supports slash commands in PR comments such as rerun-tests, regenerate-tests, generate-tests, and rerun-unit-tests.

AI-Git-Bot ships with integrations for Gitea, GitHub, GitLab, and Bitbucket, and the page says it supports multiple AI providers at once. Named providers include Anthropic, OpenAI-compatible providers, Google Gemini, Ollama, and llama.cpp. Local models such as Ollama and llama.cpp run on the user’s own hardware. The project is delivered as a self-hosted deployment using Docker Compose, with one PostgreSQL database and a web-based admin UI at localhost:8080. The setup instructions show a git clone, docker compose up -d, and then creating an administrator account and wiring a bot through a webhook URL. It also states that API keys and other credentials are encrypted at rest with AES-256-GCM.

The page presents AI-Git-Bot as production-oriented and says every workflow is opt-in per bot.