A powerful CLI tool that combines PR comments mining and automated documentation generation for GitHub repositories. Extract coding guidelines from PR reviews and generate comprehensive documentation using AWS Bedrock LLM.

Features

🔍 PR Comments Mining

• Extract and analyze PR review comments from GitHub repositories
• Generate LLM-friendly coding guidelines from expert reviews
• Classify comments by type (code standards, discussions, general)
• Support for checkpoint-based processing with resume capability

📚 Documentation Generation

• Automatically generate comprehensive documentation for codebases
• Support for JavaScript, TypeScript, JSX, and TSX files
• Dependency graph analysis and visualization
• AWS Bedrock LLM-powered intelligent documentation
• Compressed SKF format for efficient AI parsing

🚀 Repository Management

• Automatic GitHub repository cloning
• Support for both local and remote repositories
• Intelligent file discovery and parsing
• Clean temporary file management

Installation

Prerequisites

• Python 3.8+
• Node.js and npm (for JavaScript/TypeScript parsing)
• AWS CLI configured with Bedrock access
• Git

Install from Source

git clone <repository-url>
cd LLM-AutoDoc
pip install -e .

Install Dependencies

pip install -r requirements.txt

Configuration

Environment Variables

# GitHub Access
export GITHUB_TOKEN="your_github_token"

# AWS Configuration
export AWS_PROFILE="qa"  # or your preferred profile
export AWS_REGION="us-east-1"
export BEDROCK_MODEL_ID="us.anthropic.claude-3-5-sonnet-20241022-v2:0"

# Optional: Custom settings
export BEDROCK_TEMPERATURE="0.1"
export BEDROCK_MAX_TOKENS="4000"

AWS Bedrock Setup

1. Configure AWS credentials: aws configure
2. Request access to Claude models in AWS Bedrock console
3. Ensure your region supports Bedrock (us-east-1 recommended)

Usage

The unified tool provides several commands for different use cases:

Environment Setup

Setup Dependencies

Set up all required dependencies and validate configuration:

# Complete environment setup
autodoc setup

# Run diagnostics to check current setup
autodoc setup --diagnostics

PR Comments Analysis

Generate Coding Guidelines

Extract coding guidelines from PR review comments:

# Generate guidelines from top 10 PRs
autodoc generate https://github.com/owner/repo -k 10

# Custom output file
autodoc generate https://github.com/owner/repo -k 5 --output my-guidelines.txt

# Resume interrupted processing
autodoc generate https://github.com/owner/repo -k 10 --resume

Analyze Top PRs

Get repositories with most review activity:

# Top 5 PRs by comment count
autodoc top https://github.com/owner/repo -k 5

# JSON output format
autodoc top https://github.com/owner/repo -k 10 --format json

Analyze Specific PR

Get detailed information about a specific PR:

autodoc pr https://github.com/owner/repo/pull/123

Classify PR Comments

Analyze and classify PR comments:

autodoc classify https://github.com/owner/repo -k 5 --output analysis.txt

Documentation Generation

Generate Documentation for Local Repository

# Basic documentation generation
autodoc document /path/to/local/repo

# With custom output file
autodoc document /path/to/local/repo --output my-docs.md

# Generate both full and compressed versions
autodoc document /path/to/local/repo --compress

Generate Documentation for Remote Repository

# Clone and generate documentation
autodoc document https://github.com/owner/repo

# Keep cloned repository after processing
autodoc document https://github.com/owner/repo --keep-clone

# Generate compressed documentation
autodoc document https://github.com/owner/repo --compress

Command Reference

Global Options

• --token: GitHub personal access token
• --quiet: Reduce verbose output

Commands

generate - PR Comments Guidelines

Generate LLM-friendly coding guidelines from PR comments.

Arguments:

• repo_url: GitHub repository URL
• -k: Number of top PRs to analyze (default: 5)
• --output: Output file name (auto-generated if not provided)
• --resume: Resume from checkpoint
• --checkpoint-dir: Checkpoint directory (default: .checkpoints)

document - Documentation Generation

Generate comprehensive documentation for a repository.

Arguments:

• repo_path: Repository path (local directory or GitHub URL)
• --output: Output documentation file name (default: documentation.md)
• --compress: Also generate compressed SKF format
• --keep-clone: Keep cloned repository after processing

top - Top PRs Analysis

Fetch top PRs by comment count.

Arguments:

• repo_url: GitHub repository URL
• -k: Number of top PRs to fetch (default: 5)
• --format: Output format (text/json, default: text)

pr - Single PR Analysis

Fetch detailed information about a specific PR.

Arguments:

• pr_url: GitHub PR URL
• --format: Output format (text/json, default: text)

classify - Comment Classification

Classify PR comments using Bedrock.

Arguments:

• repo_url: GitHub repository URL
• -k: Number of top PRs to analyze (default: 5)
• --output: Output file name (default: pr_analysis.txt)
• --resume: Resume from checkpoint
• --checkpoint-dir: Checkpoint directory

Examples

Complete Workflow Example

# 1. Generate coding guidelines from PR comments
autodoc generate https://github.com/facebook/react -k 15 --output react-guidelines.txt

# 2. Generate comprehensive documentation
autodoc document https://github.com/facebook/react --compress --output react-docs.md

# 3. Analyze top PRs for insights
autodoc top https://github.com/facebook/react -k 10 --format json > react-top-prs.json

Local Development Workflow

# Generate documentation for current project
autodoc document . --compress

# Analyze your team's PR patterns
autodoc generate https://github.com/yourorg/yourproject -k 20

File Formats

Documentation Output

• Markdown (.md): Human-readable comprehensive documentation
• SKF (.skf.txt): Compressed format optimized for AI/LLM consumption

PR Analysis Output

• Text (.txt): LLM-friendly coding guidelines
• JSON: Structured data for programmatic use

Supported Languages

Documentation Generation

• JavaScript (.js)
• TypeScript (.ts)
• JSX (.jsx)
• TSX (.tsx)

Parser Requirements

• JavaScript/TypeScript: Node.js and npm

Troubleshooting

Common Issues

AWS Bedrock Access

# Check AWS configuration
aws sts get-caller-identity

# Verify Bedrock access
aws bedrock list-foundation-models --region us-east-1

GitHub Rate Limits

# Use personal access token
export GITHUB_TOKEN="your_token_here"

# Check rate limit status
curl -H "Authorization: token $GITHUB_TOKEN" https://api.github.com/rate_limit

Parser Setup

# Setup Node.js dependencies
cd LLM-AutoDoc/parsers && npm install

Error Messages

• "AWS credentials not found": Configure AWS CLI or set environment variables
• "Cannot connect to AWS Bedrock": Check region and model access permissions
• "Node.js parser not set up": Run npm install in parsers directory
• "Repository not found": Verify repository URL and access permissions