🧠 STORM Research: Multi-Agent Collaborative Research System add examples

[jh941213]

Description:

🧠 Feature Overview

I've developed a multi-agent collaborative research system based
on AutoGen that implements the STORM (Synthesis of Topic Outline
through Retrieval and Multi-perspective question asking)
methodology. This could be a valuable addition to the AutoGen
ecosystem.

🚀 What it does

• Multi-Agent Research Collaboration: Multiple AI analysts
  work together to conduct comprehensive research
• Multi-Source Information Gathering: Integrates 5+ search
  tools (Tavily, DuckDuckGo, Naver News, Wikipedia, ArXiv)
• Interactive Feedback System: Users can provide feedback to
  improve research quality iteratively
• Real-time Progress Tracking: WebSocket/SSE-based live
  research monitoring
• Comprehensive Report Generation: Academic-level reports
  with multi-perspective insights

🛠️ Technical Implementation

• Built on AutoGen 0.6.4 framework
• FastAPI backend with Next.js 15 frontend
• Supports Azure OpenAI, OpenAI, and Anthropic models
• Real-time communication via WebSocket and SSE

📁 Repository

https://github.com/jh941213/STORM_Research_AG

💡 Value Proposition

This implementation demonstrates how AutoGen can be used for
complex, multi-step research workflows involving:

• Dynamic agent generation based on research topics
• Parallel and sequential interview modes
• Multi-turn conversations between specialized agents
• Integration with external knowledge sources

Would the AutoGen team be interested in this as:

1. An official example/template?
2. Integration into the core framework?
3. Documentation case study?

🤝 Contribution

I'm happy to contribute this implementation or collaborate on
integrating similar functionality into AutoGen.