Hey Gitea community! ![]()
I’m excited to share a project I’ve been working on: AI Gitea Bot – a fully self-hosted, open-source Spring Boot application that brings AI-powered code reviews to your Gitea instance.
What does it do?
When you open or update a Pull Request, the bot automatically reviews the diff and posts detailed feedback as a review comment. But that’s just the beginning:
-
Automatic PR Code Reviews – Get instant AI feedback on every PR -
Interactive Bot Commands – Mention @ai_botin any comment to ask questions or request deeper analysis -
Inline Review Responses – The bot can reply directly to inline review comments with file-aware context -
Conversation Memory – Full session management keeps context across PR updates and follow-up discussions
Choose Your AI Provider
One of the key design goals was flexibility. You can use:
| Provider | Description |
|---|---|
| Anthropic Claude | Claude 3.5/4 via Anthropic API |
| OpenAI | GPT-4o and other models via OpenAI API (or compatible endpoints) |
| Ollama | Run local LLMs like Llama, Mistral, Qwen – no API key required! |
This means you can run the entire stack on your own infrastructure with Ollama if you want complete data privacy.
Autonomous Issue Agent 
I’m currently working on an exciting feature: an autonomous issue implementation agent. Assign the bot to a Gitea issue, and it will:
-
Analyze the issue description
-
Explore the repository structure
-
Generate an implementation plan
-
Create a feature branch with the changes
-
Open a Pull Request automatically
This agentic functionality includes AI-driven validation – the agent can run build tools (mvn compile, npm build, etc.) and iteratively fix errors before submitting. This feature is in active development and will be released soon! A screenshot will follow (only 1 embedded media-file allowed).
Links
-
Docker Hub: tmseidel/ai-gitea-bot - Docker Image
Feedback Welcome!
I’d love to hear your thoughts, feature requests, or contributions. The project is MIT licensed and open for collaboration.
Has anyone else experimented with AI-assisted code reviews in Gitea? Would love to hear about your experiences!

