storm
An LLM-powered knowledge curation system that researches a topic and generates a full-length report with citations.
Overview
STORM (Synthesis of Topic Outlines through Retrieval and Multi-perspective Question Asking) is an advanced LLM-powered knowledge curation system that automatically researches topics and generates comprehensive, Wikipedia-style articles with citations. The system combines Internet search capabilities with multi-perspective question asking to synthesize information from diverse sources into coherent, structured reports. With over 28,000 GitHub stars, STORM has gained significant traction in the research community. The enhanced Co-STORM version introduces human-AI collaborative features, allowing users to work alongside the AI system for more aligned and preferred information seeking. STORM supports multiple language models through litellm integration, various search engines including Bing Search, and can ground research on user-provided documents via VectorRM. The system features a modular architecture with customizable retrieval and search integration, making it adaptable to different use cases. While it cannot produce publication-ready articles that meet professional standards, experienced Wikipedia editors have found it valuable for pre-writing research and initial content generation. The tool includes both a research preview web interface and a streamlit-based demo for local development, providing accessible entry points for different user needs.
Pros
- + Automated multi-perspective research that synthesizes information from diverse Internet sources into structured, Wikipedia-style articles with proper citations
- + Human-AI collaborative features through Co-STORM enable interactive knowledge curation with user guidance and preferences
- + Flexible architecture supporting multiple language models, search engines, and document sources through modular components and extensive customization options
Cons
- - Cannot produce publication-ready articles and requires significant manual editing and fact-checking before professional use
- - Quality and accuracy depend heavily on the underlying language model and search results, potentially leading to inconsistencies or outdated information
- - Complex setup and configuration may be challenging for non-technical users despite simplified installation options
Use Cases
- • Pre-writing research assistance for Wikipedia editors and content creators who need comprehensive topic overviews before manual article development
- • Academic research synthesis for students and researchers who need to quickly gather and organize information from multiple sources on specific topics
- • Knowledge base generation for organizations that need to create structured reports from internal documents and external sources