STORM

STORM (Synthesis of Topic Outlines through Retrieval and Multi-perspective Question Asking) is a Stanford OVAL Lab system that writes Wikipedia-like articles from scratch based on internet search. It simulates a multi-perspective research process: the system asks questions from different angles, retrieves sources, and synthesizes an outlined, cited article. Co-STORM extends it with human-AI collaborative knowledge curation, letting people steer the information-seeking process.

The codebase ships as the knowledge-storm Python package with litellm integration so different components can use different models to balance cost and quality. The research preview at storm.genie.stanford.edu has been tried by more than 70,000 people; experienced Wikipedia editors have found it useful in the pre-writing stage, though the authors note it does not produce publication-ready articles without editing. The work is published as arXiv papers 2402.14207 and 2408.15232, MIT licensed, with over 31k GitHub stars.

1. pip install knowledge-storm
2. Configure LLM and retrieval models via litellm
3. Run the pipeline on a topic, or try the live research preview