Skip to content

AI-powered code review automation using Kaito as the backend. A fork of qodo-ai/pr-agent enhanced with Kaito for better contextual understanding and more accurate PR reviews.

License

Notifications You must be signed in to change notification settings

kaito-project/kaito-pr-agent

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

KAITO PR-Agent

AI-powered code review automation using KAITO as the backend

This is a fork of qodo-ai/pr-agent modified for the KAITO project. See NOTICE for attribution details.

Why We Forked PR-Agent

We forked qodo-ai/pr-agent to integrate KAITO RAGEngine as the backend, improving PR-Agent's performance by gathering additional context from your codebase. This enhanced contextual understanding allows for more accurate code reviews, better suggestions, and smarter PR descriptions.


🚀 Quick Start

Get started in minutes with GitHub integration:

Prerequisites: You'll need a KAITO inference workspace running with your preferred model.

GitHub Action (Recommended)

Easy setup, no infrastructure needed. Setup Guide →

GitHub App (Advanced)

More control, runs in your infrastructure. Setup Guide →

📚 Essential KAITO Documentation:

Live Example: The kaito-pr-agent GitHub App with Qwen2.5-Coder-32B model is actively reviewing all PRs in the KAITO repository.


Architecture Diagram

This diagram illustrates how the PR-Agent GitHub app works with KAITO.

Architecture Diagram

Key Components & Their Significance:

  • 🔄 GitHub App - Receives PR events, manages comments, and handles authentication
  • 🤖 PR-Agent Pod - Orchestrates the analysis workflow and manages tool execution (/review, /describe, /improve)
  • ⚡ KAITO Workspace - Provides scalable Kubernetes-native AI model serving with automatic GPU provisioning
  • 🧠 KAITO RAGEngine - Enriches context by retrieving relevant code snippets and documentation from your codebase

Setup & Usage

1. Run with Docker

For quick testing, use the script below:

chmod +x scripts/run_docker.sh
./scripts/run_docker.sh

2. Run in Kubernetes

Apply the job manifest:

kubectl apply -f config/pr-agent-job.yaml

3. Run in GitHub Actions

Use the provided workflow in .github/pr-review.yaml to automate PR reviews.


Configuration

In all setups, replace the following placeholders with your actual values:

  • OLLAMA_API_BASE → Your KAITO inference workspace URL
  • GITHUB_USER_TOKEN → Your GitHub API token
  • PR_URL → The pull request URL to review

Ensure these values are correctly set in your Docker environment, Kubernetes job, or GitHub Actions workflow before running.

Supported Features

  • Auto Review (/review): AI-powered code review with security and quality feedback
  • Auto Description (/describe): Generate PR titles, summaries, and labels
  • Code Suggestions (/improve): Get improvement suggestions for your code
  • Question Answering (/ask): Ask questions about the PR changes

Links

License

See LICENSE for license information.

Credits

This project is a fork of qodo-ai/pr-agent. All credit for the original work and core functionality goes to the qodo-ai/pr-agent authors and contributors. We are grateful for their excellent work that made this fork possible.

For detailed attribution information, see NOTICE.

Contact

"KAITO devs" [email protected]

About

AI-powered code review automation using Kaito as the backend. A fork of qodo-ai/pr-agent enhanced with Kaito for better contextual understanding and more accurate PR reviews.

Resources

License

Code of conduct

Contributing

Security policy

Stars

Watchers

Forks

Packages

 
 
 

Contributors 3

  •  
  •  
  •  

Languages