AI-Powered Code Analysis System & PR Monitoring

An intelligent code analysis system that combines semantic search with AI agentic workflows for automated Pull Request review and code analysis. Clone GitHub or Azure DevOps repositories, store them in a vector database, and perform intelligent semantic searches across your codebase with AI-powered analysis.

Features

• Multi-Platform Support: Clone from GitHub and Azure DevOps repositories
• Semantic Search: Use natural language queries to find relevant code
• Vector Storage: Efficient storage and retrieval using Supabase vector database
• AI-Powered PR Analysis: Automated Pull Request review using agentic workflows
• Smart File Filtering: Automatically processes only relevant code files
• Interactive CLI: Easy-to-use command-line interface
• Batch Operations: Store and search multiple repositories

Quick Start

Prerequisites

• Python 3.10+
• OpenAI API key (for embeddings and LLM)
• Supabase account and database

Note🚩: I used Qwen-3-coder as the llm but you can use any llm of your choice, check the workflow\nodes.py to LLM assignment.

Installation

1. Clone this repository:

git clone <your-repo-url>
cd repository-search-system

2. Install dependencies:

pip install -r requirements.txt

3. Set up environment variables:

cp .env.example .env
# Edit .env with your API keys

4. Set up Supabase database:
   • Run the SQL commands in supabase_schema.sql in your Supabase SQL editor
   • This creates the necessary tables and functions

Usage

Store a Repository

Store a GitHub repository:

python store_repos.py

Store a custom repository interactively:

python store_repos.py --custom

Search Repositories

Interactive search:

python search_repos.py --interactive

Search all repositories:

python search_repos.py --all "authentication logic"

Run example searches:

python search_repos.py

Run AI-Powered PR Analysis

Execute the agentic workflow for PR analysis:

python examples/pr_analysis_workflow_demo.py

Test the agentic workflow components:

python examples/test_agentic_workflow.py

Configuration

Environment Variables

Create a .env file with the following variables:

# OpenAI API Key for embeddings
OPENAI_API_KEY=your_openai_api_key_here

# Supabase Database URL
SUPABASE_DB_URL=postgresql://user:password@host:port/database

# Optional: Azure DevOps Personal Access Token
AZURE_API_KEY=your_azure_pat_here

Supported File Types

The system automatically processes these file types:

• Code: .py, .js, .ts, .java, .cpp, .cs, .php, .rb, .go, .rs, .swift
• Web: .html, .css, .scss, .jsx, .tsx
• Config: .json, .yaml, .yml, .toml, .sql
• Documentation: .md, .txt, .rst
• Special files: README, LICENSE, Dockerfile, requirements.txt, etc.

API Reference

Core Functions

store_repos.py

# Store GitHub repository
store_github_repo(github_url, branch=None, refresh=False)

# Store Azure DevOps repository  
store_azure_repo(organization, project, repository, branch=None, pat=None, refresh=False)

search_repos.py

# Search specific repository
search_repository(repo_name, query, limit=5, show_content=False)

# Search all repositories
search_all_repos(query, limit=3)

# Interactive search interface
interactive_search()

Utility Functions

utils/git_utils.py

# Get repository information from GitHub
get_github_repo_info(github_url, branch=None)

# Get repository information from Azure DevOps
get_azure_repo_info(organization, project, repository, branch=None, pat=None)

utils/vector_utils.py

# Store repository in vector database
store_repo_in_own_collection(repo_info, refresh=False)

# Search within repository collection
search_repo_collection(repo_name, query, limit=5)

# List all repository collections
list_repo_collections()

Examples

Store Multiple Repositories

# GitHub repository
store_github_repo("https://github.com/user/repo", branch="main")

# Azure DevOps repository
store_azure_repo("org", "project", "repo", branch="develop")

Search Examples

# Find authentication code
search_repository("my-repo", "user authentication and login")

# Find database queries
search_repository("my-repo", "SQL queries and database connections")

# Find API endpoints
search_repository("my-repo", "REST API endpoints and routes")

AI-Powered PR Analysis Examples

# Run the complete agentic workflow for PR analysis
from workflow.graph import create_default_workflow
from workflow.state import WorkflowState

# Create workflow
workflow = create_default_workflow()

# Create initial state with PR data
state = WorkflowState(collection_name="repo-name", pr_data=json_pr_data)

# Execute workflow
result = workflow.invoke(state)

Architecture

┌─────────────────┐    ┌──────────────────┐    ┌─────────────────┐
│   Git Repos     │───▶│   Processing     │───▶│   Supabase      │
│ (GitHub/Azure)  │    │   & Embedding    │    │  Vector Store   │
└─────────────────┘    └──────────────────┘    └─────────────────┘
                              │                          │
                              ▼                          ▼
                       ┌──────────────────┐    ┌─────────────────┐
                       │   File Filter    │    │   Search API    │
                       │   & Content      │    │   & Results     │
                       └──────────────────┘    └─────────────────┘

Agentic Workflow

The system implements an AI-powered agentic workflow for Pull Request (PR) analysis using LangGraph. This workflow consists of specialized agents that collaborate to analyze code changes and provide comprehensive feedback.

Workflow Components

1. Parsing Function (parse_pr_data): Extracts repository information and collection name from PR data
2. Analysis Agent (pr_analysis_agent): Reviews code changes for correctness, style, maintainability, and security risks
3. Tools: Vector database query tools that allow agents to explore the repository context

Agent Capabilities

The PR Analysis Agent has access to the following tools:

• list_directories: List all file paths in the repository
• get_metadata_by_id: Inspect metadata for a file without reading its full content
• get_content_by_id: Fetch the actual code of a file for deeper analysis
• search_vector_database: Search semantically for related files or concepts in the repo

Workflow Execution

The agentic workflow follows this sequence:

1. Parse PR data to extract repository information
2. Analyze code changes using the AI agent with access to repository tools
3. Generate structured feedback including:
   • Summary of the PR
   • Detailed findings for each file
   • Security concerns
   • Final recommendation

State Management

The workflow maintains state throughout execution using the WorkflowState dataclass, which tracks:

• Collection name and PR data
• Completed workflow nodes
• Analysis results and metadata
• Execution timing information

Troubleshooting

Common Issues

1. "No collections found": Run store_repos.py first to add repositories
2. Authentication errors: Check your API keys in .env file
3. Database connection: Verify your SUPABASE_DB_URL is correct
4. Large repositories: The system skips files >100KB automatically

Performance Tips

• Use specific queries for better results
• Limit search results for faster responses
• Refresh collections only when necessary
• Monitor your OpenAI API usage

Contributing

1. Fork the repository
2. Create a feature branch
3. Make your changes
4. Add tests if applicable
5. Submit a pull request