https://github.com/xiongQvQ/multi_agent_demo

ReadyTensor Multi-Agent Demo

A simple multi-agent system built with LangGraph, Gemini, and LangChain to demonstrate agent coordination and tool usage.

Features

• 3 Agents: Researcher, Analyst, Reporter
• 3 Tools: Search, Calculator, File processing
• Orchestration: LangGraph workflow management
• LLM: Google Gemini integration

Setup

1. Install dependencies:

pip install -r requirements.txt

2. Set up environment variables:

cp .env.example .env
# Edit .env and add your API keys:
# GOOGLE_API_KEY=your_gemini_api_key_here
# SERPER_API_KEY=your_serper_api_key_here (optional - for real search)

3. Run the demo:

python main.py

API Keys

• GOOGLE_API_KEY: Required for Gemini LLM
• SERPER_API_KEY: Optional for real Google search via Serper.dev API (falls back to simulated search if not provided)

Architecture

User Input → Researcher (Search) → Analyst (Calculate) → Reporter (Generate Report) → Output

Demo Example

Ask: "Analyze Apple's stock performance"

• Researcher: Searches for Apple stock information
• Analyst: Calculates relevant metrics
• Reporter: Generates final analysis report