Skip to main content
RepoMaster provides a comprehensive suite of specialized AI agents designed to handle different types of tasks, from web research to repository analysis and general programming assistance.

​
Available Agents

RepoMaster includes four specialized agents, each optimized for specific use cases:

Deep Search Agent

Advanced web search and information retrieval with real-time data access

Programming Assistant

General-purpose coding assistance for algorithms, debugging, and code examples

Repository Agent

Analyze and execute tasks using GitHub or local repositories

Unified Interface

Intelligent orchestration that automatically selects the best agent for your task

​
Agent Architecture

All RepoMaster agents are built on the AutoGen framework and share a common architecture:

​
Core Components

Agent Pairs
  • Assistant Agent: Plans and reasons about tasks
  • User Proxy Agent: Executes tools and code
Tool Registration
  • Agents are equipped with specialized tools via register_toolkits()
  • Tools are exposed as callable functions with type annotations
LLM Configuration
  • Configurable LLM backend (Azure OpenAI, OpenAI, etc.)
  • Supports caching for improved performance

​
Communication Pattern

# Basic agent interaction pattern
assistant.initiate_chat(
    user_proxy,
    message=user_query,
    max_turns=30,
    summary_method="reflection_with_llm"
)

​
Choosing the Right Agent

  • You need real-time information from the web
  • Your task requires current events or latest data
  • You want to research a topic across multiple sources
  • You need to answer questions beyond the model’s knowledge cutoff
  • You need help with general programming questions
  • You want to write code snippets or algorithms
  • You need debugging assistance
  • You want code explanations or tutorials
  • No specific repository is required
  • You have a specific GitHub repository to work with
  • You need to analyze local code repositories
  • Your task requires understanding existing codebases
  • You want to execute tasks using specialized libraries
  • You’re unsure which agent is best for your task
  • Your task might require multiple agents
  • You want intelligent automatic agent selection
  • You prefer a single entry point for all capabilities

​
Agent Capabilities Comparison

CapabilityDeep SearchProgramming AssistantRepositoryUnified
Web Search✅❌❌✅
Code Execution❌✅✅✅
Repository Analysis❌❌✅✅
General Coding Help❌✅❌✅
Multi-Agent Orchestration❌❌❌✅
Automatic Mode Selection❌❌❌✅

​
Configuration

All agents share common configuration options:
​
llm_config
dict
required
LLM configuration including API keys, model settings, and caching options
​
code_execution_config
dict
required
Code execution settings including working directory and Docker usage

​
Example Configuration

from configs.oai_config import get_llm_config
import uuid
import os

# Get LLM configuration
llm_config = get_llm_config()

# Setup execution environment
work_dir = os.path.join(os.getcwd(), "coding", str(uuid.uuid4()))
code_execution_config = {
    "work_dir": work_dir,
    "use_docker": False
}

​
Next Steps

Deep Search Agent

Learn about web search and information retrieval

Programming Assistant

Explore general coding assistance features

Repository Agent

Discover repository analysis capabilities

Unified Interface

Use the intelligent multi-agent orchestrator

Build docs developers (and LLMs) love

Get started for freeTalk to us