README.md

LLMxMapReduce_V3

Introduction

we introduce LLM×MapReduce-V3, an interactive, modular, and self-organized multi-agent system for academic survey generation. Building upon LLM×MapReduce-V2, our framework leverages the Model Context Protocol to enable composable modules, adaptive planning, and human-in-the-loop alignment. The system supports dynamic workflows through an agent-level planner, allowing flexible orchestration of modules such as document digestion, skeleton construction, and survey writing.

Quick Start

Environment Setup

1. Install uv

# (Optional) Upgrade pip
pip install --upgrade pip

# Install uv
pip install uv
# Or use Homebrew (macOS/Linux)
brew install uv

2. Create and Activate a Virtual Environment

cd LLMxMapReduce_V3

# Create a new environment
uv venv

# Activate environment
# Linux /macOS
source .venv/bin/activate
# Windows Powershell
.venv\Scripts\Activate.ps1

# Check
which python

3. Install Dependencies

uv pip install -r requirements.txt

Configure APIs

Edit config/unified_config.json with your API keys and endpoints. Supports OpenAI plus search engines (SerpAPI, Bing, Google):

    "api_keys": {
        "openai": {
            "api_key": "YOUR_OPENAI_KEY",
            "base_url": "YOUR_OPENAI_URL"
        },
        "search_engines": {
            "serpapi_key": "",
            "bing_subscription_key": "",
            "google_custom_search": {
                "api_key": "",
                "search_engine_id": ""
            }
        }
    },

Start Survey Writing

bash start.sh "Your Topic" "Your Description"