Verified project record

stanford-oval/storm

STORM automates the research and writing process to generate long, Wikipedia-style articles with citations from scratch. It uses multi-perspective question asking and simulated conversations grounded in retrieved internet sources to curate knowledge before writing.

Project overview

The system coordinates multi-perspective question asking and simulated expert conversations grounded in internet sources to research and generate cited long-form articles.

Project type: AI Agent · RAG · Workflow · Prompt Engineering
Use cases: Documents & Office · Knowledge Q&A · Search & Research
Deployment: Python environment
License: MIT

Best for

Researchers, developers, or general users seeking to automate the breakdown, perspective-driven research, and drafting of long, cited articles.
Users looking to participate in steering a dynamic, shared knowledge curation process alongside multiple LLM experts.

Key capabilities

Breaks down generating long articles with citations into pre-writing research and writing stages to produce a full-length article.
Discovers different perspectives by surveying existing articles from similar topics and uses them to control the question-asking process to improve research depth.
Simulates a conversation between a Wikipedia writer and a topic expert grounded in Internet sources to update understanding and ask follow-up questions.
Implements a turn management policy to support smooth collaboration among Co-STORM LLM experts, a Moderator, and human users to steer the discussion.
Maintains a dynamic updated mind map to organize collected information into a hierarchical concept structure and build a shared conceptual space between the human user and the system.
Provides a modular engine interface allowing customization of different language models, retrieval integrations, and pipeline modules.
Supporting user participation in the knowledge curation process.
Developing abstractions for curated information to support presentation formats beyond the Wikipedia-style report.

Limitations and risks

The system cannot produce publication-ready articles that often require a significant number of edits.
While the system can be locally run and self-hosted, the data boundary is not documented, and local orchestration does not preclude remote inference or cloud round-trips via configured language models or search services.

Getting started

Install the knowledge-storm package via pip. Configuration is considered medium difficulty because it requires writing or running Python scripts for library usage, configuring external API keys for language models in secrets.toml, and setting up search engine access.
Install the package, configure API keys in secrets.toml, and run the example script for STORM or Co-STORM.

Alternatives and comparisons

Dynamically decomposes long-form writing tasks through human-like adaptive planning that interleaves recursive task decomposition, heterogeneous integration, and execution.
Extracts reusable principles from literature and composes them into traceable, inspectable Idea Cards and validation-ready research packs.
Provides attendee and presenter resources for a workshop demonstrating a multi-agent creative writer application using Azure OpenAI.

Project comparisons

Evidence and sources

GitHub project description: An LLM-powered knowledge curation system that researches a topic and generates a full-length report with citations.
README: We add [litellm](https://github.com/BerriAI/litellm) integration for language models and embedding models in `knowledge-storm` v1.1.0.
README: While the system cannot produce publication-ready articles that often require a significant number of edits, experienced Wikipedia editors have found it helpful in their pre-writi…
README: STORM breaks down generating long articles with citations into two steps: 1. **Pre-writing stage**: The system conducts Internet-based research to collect references and generates…
README: Co-STORM codebase is now released and integrated into `knowledge-storm` python package v1.0.0. Run `pip install knowledge-storm --upgrade` to check it out.

Reviewed at: 2026-07-17T10:03:08Z

Open original GitHub project