Skip to content

Latest commit

 

History

History
312 lines (225 loc) · 9.73 KB

USAGE.md

File metadata and controls

312 lines (225 loc) · 9.73 KB

RepoMaster User Guide

This comprehensive guide provides everything you need to use RepoMaster effectively, from basic configuration to advanced usage patterns.

📋 Table of Contents

🚀 Getting Started

Prerequisites

  • Python 3.11+
  • Git
  • Internet connection for repository cloning

Installation

git clone https://github.com/QuantaAlpha/RepoMaster.git
cd RepoMaster
pip install -r requirements.txt

Basic Configuration

Copy the example configuration file and customize it with your API keys:

cp configs/env.example configs/.env
# Edit the configuration file with your API keys
nano configs/.env  # or use your preferred editor

Configuration Template (from configs/env.example):

# Set the default API provider (openai, claude, deepseek, azure_openai)
DEFAULT_API_PROVIDER=openai

# OpenAI Configuration
OPENAI_API_KEY=your_openai_key
OPENAI_MODEL=gpt-5

# Claude Configuration  
ANTHROPIC_API_KEY=your_claude_key
ANTHROPIC_MODEL=claude-4-sonnet

# DeepSeek Configuration
DEEPSEEK_API_KEY=your_deepseek_key
DEEPSEEK_MODEL=deepseek-v3

# Google Gemini Configuration
GEMINI_API_KEY=your_gemini_key
GEMINI_MODEL=gemini-2.5-pro

# Web Search APIs (Required for deep search functionality)
SERPER_API_KEY=your_serper_key          # For Google search results - Get API key at: https://serper.dev/login
JINA_API_KEY=your_jina_key              # For web content extraction - Get API key at: https://jina.ai/

💡 Note: The configs/env.example file contains the complete configuration template with all available options and detailed comments.

🧠 Multi-Agent Intelligence System

🤖 One Interface, All GitHub Resources

Simply describe your task in natural language. RepoMaster automatically finds the right GitHub tools and makes them work together to solve your task.

python launcher.py --mode backend --backend-mode unified

🎯 Intelligent Multi-Agent Orchestration

RepoMaster features a sophisticated Multi-Agent System where specialized AI agents work in harmony to deliver optimal solutions. Our intelligent dispatcher automatically routes tasks to the most suitable agent combination:

🔍 Deep Search Agent 💻 Programming Assistant Agent 🏗️ Repository Exploration Agent
Advanced Search & Web Analysis Code Generation & Programming Repository Understanding & Task Execution
• Advanced web research & data retrieval • Intelligent code generation • Autonomous code exploration
• Information synthesis & analysis • Algorithm implementation • Complex task orchestration
• Query optimization • Debug & code optimization • Multi-repo coordination

🚀 How Multi-Agent System Works

👤 User Task Input
     ↓
🧠 AI Intelligent Dispatcher
     ↓
🔀 Task Analysis & Agent Selection
     ↓
┌─────────────────┬─────────────────┬─────────────────┐
│🔍 Deep Search &  │💻 Programming    │🏗️ Repository     │
│ Web Research     │ Assistant        │ Exploration     │
│                 │                 │                 │
│ • Web search     │ • Code generation│ • Repo analysis  │
│ • Data synthesis │ • Algorithm impl │ • Task execution │
│ • Context build  │ • Debug support  │ • Multi-repo ops │
└─────────────────┴─────────────────┴─────────────────┘
     ↓
🎯 Intelligent Result Orchestration
     ↓
✅ Perfect Solution Delivered

✨ Key Innovation: No manual agent selection required - our AI dispatcher intelligently combines agents based on task complexity and requirements, ensuring optimal performance for every request.

💻 Multi-Agent Access Interfaces

🤖 Unified Multi-Agent Interface (Recommended)

The primary way to use RepoMaster - one command, all GitHub resources at your service:

python launcher.py --mode backend --backend-mode unified

Why Unified Multi-Agent Interface?

  • 🧠 AI-Powered Task Analysis: Automatically understands your intent
  • 🤝 Intelligent Agent Collaboration: Seamlessly coordinates multiple agents as needed
  • 🎯 Context-Aware Routing: Dynamically selects optimal agent combinations
  • ⚡ Zero Configuration: No manual agent selection required

🌐 Web Interface (Visual Multi-Agent Dashboard)

Launch the interactive web interface for visual multi-agent interaction:

python launcher.py --mode frontend
# Access: http://localhost:8501

# Configure file upload size limit (default: 200MB)
python launcher.py --mode frontend --max-upload-size 500  # Set to 500MB

Multi-Agent Dashboard Features:

  • 🌐 Interactive multi-agent chat interface
  • 📁 File upload and management across agents (configurable size limit)
  • 👥 Multi-user session support
  • 📊 Real-time agent collaboration visualization

🔧 Direct Agent Access (Advanced)

For developers who want direct access to individual agents:

Individual Agent Interfaces (Click to expand) 
# Direct access to Deep Search Agent
python launcher.py --mode backend --backend-mode deepsearch

# Direct access to Programming Assistant Agent
python launcher.py --mode backend --backend-mode general_assistant

# Direct access to Repository Exploration Agent
python launcher.py --mode backend --backend-mode repository_agent

💡 Note: These direct agent interfaces are primarily for development, testing, and specialized workflows. For optimal performance and seamless agent collaboration, the unified multi-agent interface is recommended for production use.

Shell Script Shortcuts

# Frontend
bash run.sh frontend

# Backend modes
bash run.sh backend unified
bash run.sh backend deepsearch  
bash run.sh backend general_assistant
bash run.sh backend repository_agent

🔧 Advanced Usage

Command Line Configuration Options

RepoMaster supports various configuration options to customize your experience:

# Frontend configuration options
python launcher.py --mode frontend \
  --streamlit-port 8502 \              # Custom port (default: 8501)
  --streamlit-host 0.0.0.0 \           # Custom host (default: localhost)
  --max-upload-size 1000 \             # File upload limit in MB (default: 200)
  --log-level DEBUG                    # Logging level (default: INFO)

# Backend configuration options  
python launcher.py --mode backend --backend-mode unified \
  --api-type openai \                  # API provider (default: basic)
  --temperature 0.1 \                  # Model temperature (default: 0.1)
  --work-dir /custom/path \            # Working directory (default: coding)
  --timeout 300 \                      # Request timeout in seconds (default: 120)
  --max-tokens 8000                    # Maximum token count (default: 4000)

File Upload Size Configuration:

  • Default limit: 200MB
  • Range: 1MB - 2000MB (2GB)
  • Affects web interface file uploads
  • Example: --max-upload-size 500 sets limit to 500MB

Basic Programming Interface

from core.agent_scheduler import RepoMasterAgent

# Simple task execution
task = "Transform this portrait into Van Gogh style using content.jpg and style.jpg"
result = repo_master.solve_task_with_repo(task)

For detailed programming examples, see our Documentation.

📝 Use Cases

🤖 AI/ML Tasks

"Train an image classifier on CIFAR-10 dataset using transfer learning"

  • Automatically finds relevant ML repositories and frameworks
  • Sets up complete training pipeline with best practices
  • Handles data loading, model configuration, and training execution

📄 Data Processing

"Extract tables from PDF reports and convert to structured CSV format"

  • Discovers PDF processing libraries and tools
  • Implements extraction pipeline with error handling
  • Outputs clean, structured data in the desired format

🌐 Web Development

"Create a REST API for user authentication with JWT tokens"

  • Searches for authentication frameworks and security libraries
  • Generates production-ready API with proper security practices
  • Includes documentation and testing examples

👁️ Computer Vision

"Detect and count objects in surveillance video footage"

  • Finds state-of-the-art object detection models
  • Implements video processing pipeline with optimization
  • Provides detailed analysis results and visualizations

📖 Running Tests

# Run configuration tests
python test_config.py

# Run full test suite
pytest tests/

# Run specific benchmark
python -m core.git_task --config configs/gittaskbench.yaml

🤝 Contributing

Development Environment Setup

git clone https://github.com/QuantaAlpha/RepoMaster.git
cd RepoMaster
pip install -e ".[dev]"
pre-commit install

Contribution Types

  • 🐛 Bug fixes
  • ✨ New feature development
  • 📚 Documentation improvements
  • 🧪 Test case additions
  • 🔧 Tools and utilities

📞 Support

Last updated: December 2024