🤖 AI Pull Request Reviewer

An automated code review agent that leverages Google's Gemini models and LangGraph to provide intelligent, multi-perspective feedback on GitHub Pull Requests. This agent scans PR diffs for security vulnerabilities, performance bottlenecks, and adherence to Python best practices, then posts summarized reports directly back to GitHub.

🌟 Features

• Multi-Agent Analysis: Uses a specialized graph-based workflow to run three distinct analysis nodes:
  • Security Agent: Scans for SQL injection, XSS, hardcoded secrets, and PII exposure.
  • Performance Agent: Identifies N+1 queries, expensive loops, and memory leaks.
  • Style Agent: Ensures PEP8 compliance, proper naming conventions, and type hint usage.
• Interactive Web UI: Built with Streamlit for easy PR URL entry and real-time analysis visualization.
• Direct GitHub Integration: Automatically fetches PR file patches and posts structured review comments (including code suggestions) back to the PR.
• Structured Output: Utilizes Pydantic models to ensure consistent reporting of issue severity, line numbers, and actionable advice.

🏗️ Architecture

The system is built on a LangGraph state machine, which orchestrates the flow of data between different AI specialized agents:

1. Entry: Fetches files from GitHub via GitHubConnector.
2. Security Scan: Initial analysis for critical vulnerabilities.
3. Performance Scan: Evaluation of code efficiency.
4. Style Scan: Linting and best practice review.
5. Aggregator: Consolidates all agent findings into a final report.

🛠️ Tech Stack

• LLM: Google Gemini 2.5 Flash Lite
• Orchestration: LangChain & LangGraph
• Frontend: Streamlit
• API: PyGithub
• Validation: Pydantic

🚀 Getting Started

Prerequisites

• Python 3.11+
• A GitHub Personal Access Token (with repo permissions)
• A Google AI (Gemini) API Key

Installation

1. Clone the repository:

git clone https://github.com/manthan-jsharma/ai-pr-agent.git
cd ai-pr-agent

2. Install dependencies:

pip install -r requirements.txt

3. Configure Environment Variables: Create a .env file in the root directory or configure Streamlit secrets:

GITHUB_ACCESS_TOKEN=your_github_token
GOOGLE_API_KEY=your_gemini_api_key

💻 Usage

Streamlit Web Interface (Recommended)

Launch the interactive dashboard to review PRs by URL:

streamlit run app.py

• Enter a public GitHub PR URL (e.g., https://github.com/user/repo/pull/1).
• Click Run Analysis to see identified issues.
• Review suggestions and click Post These Comments to GitHub to update the PR.

CLI Version

For automated or local testing, run the agent via the terminal:

python main.py

Note: Update the REPO and PR_ID variables in main.py before running.

📁 Project Structure

• agent.py: Defines the LangGraph workflow and individual agent prompts.
• app.py: Streamlit frontend implementation.
• github_utils.py: GitHub API wrappers for fetching files and posting comments.
• models.py: Data schemas for CodeIssue and PRReviewResult.
• main.py: Entry point for CLI-based reviews.
• .devcontainer/: Configuration for VS Code Dev Containers.