Knowledge organization system: Automatic generation of Wikipedia style

By /

A knowledge collation system driven by a large language model that can study a topic and generate a complete report with citations.

This is an open source project developed based on the Stanford OVAL Laboratory, and its goal is to use the Large Language Model (LLM) to automatically generate Wikipedia-style language.Structured strip quotes。

Project overview

  • Project name: STORM (Synthesis of Topic Outlines through Retrieval and Multi-perspective Question Asking), which is fully known as the topic outline synthesis system, generates an article outline through retrieval and multi-perspective questions, and then outputs a complete report.
  • Latest version features:
    • Support LiteLLM interface (updated in January 2025)
    • Integrate Co‑STORM to further support human-machine collaborative knowledge collation
    • The latest stable version is v1.1.0, which can be downloaded through pip install knowledge-storm installation

STORM’s core process

1. Pre-writing phase (Pre‑writing)

  • The system first automatically retrieves network information and generates outline material for the article;
  • Ensure that all aspects of the topic are covered by “multi-perspective questions”(such as simulating expert conversations from different angles).

2. Writing stage (Writing)

  • Based on the outline, write a complete article through LLM (such as GPT‑ 4o) and automatically insert reference links to make the content more credible.

Co‑STORM: Enhancement of Human-Machine Collaboration

  • Based on STORM, Co‑STORM introduces a collaboration mechanism:
    • Collaborative Q & A with multiple LLM agents (experts, hosts);
    • Humans can also intervene, participate in questions and answers, and adjust directions;
    • The system provides a dynamic “mind map” to visualize the knowledge structure and reduce complexity.

📦Technology stack and usage

  • Use Python and use knowledge-storm package installation;
  • Support multiple retrieval interfaces (such as Bing, You.com, VectorRM, etc.);
  • Supports multiple LLMs and can be customized;
  • User code demonstration adopts STORMWikiRunner and CoStormRunner Class to perform specific tasks.

Applicable population

  • Researchers, students, content creators, etc.;
  • Suitable for use before writingTheme research and outline construction;
  • Although the generated articles require manual polishing, a large amount of first draft workload has been reduced

Summary

STORM is an automated system for the “knowledge collation → structured writing” process. It combines Internet search and LLM models to support the generation of Wikipedia-style articles with citations. Co‑STORM further supports user participation and realizes knowledge production for human-computer collaboration. The code is open source, functional modules are flexible, and can be widely customized and extended. It is very suitable for scenarios with high demands for content quality and reliability.

Github：https://github.com/stanford-oval/storm

Oil tubing:

Scroll to Top