Skip to content
Get packs

STORM

Open-source LLM system that researches a topic online and drafts a Wikipedia-style article with citations.

GitHub Download v1.1.0
Stars
31.8k
Forks
3k
Licence
MIT
Language
Python
Last update
1.1 years ago
Type
AI model / agent

No commits in over a year — check for forks or alternatives before adopting.

What it is

STORM is a research system built on large language models. Given a topic, it searches the internet, simulates conversations to gather information, builds an outline, and writes a full-length article with citations. Co-STORM extends it so a human can take part in the discussion and steer the research. The README says the output is not publication-ready and often needs significant editing, but experienced Wikipedia editors have found it useful for pre-writing.

  • Generate a Wikipedia-style draft article on a topic with citations from internet sources
  • Run STORM's research stage only, then reuse the saved results for later writing
  • Use Co-STORM to explore a topic in a discussion with AI experts and a moderator
  • Ground the research on your own documents using the VectorRM retriever

Great for

  • Developers who want to build a customizable research or report-writing pipeline
  • Writers who want a first-draft outline and source list before writing
  • Researchers exploring a new topic who want a structured overview with references
  • Teams experimenting with different language models and search engines

Not ideal for

  • Anyone needing publication-ready text, since the README says articles often need significant edits
  • People without an API key for a language model or search provider, since the system depends on them
  • Users who need a finished no-setup app, since the README centers on a Python package and code setup
Step-by-step guide: how to use STORMInstall, first run, examples, tips and FAQs
STORM alternativesLangChain vs STORMFinGPT vs STORMOpen Deep Research vs STORM

How to get it

  1. 1

    Install with a package manager

    You'll need: Python 3.11 (per the README's conda example) · Conda or another Python environment manager · API keys for a language model provider and a search provider

    Install the library from PyPI

    pip install knowledge-storm

    Clone the repository for source install

    git clone https://github.com/stanford-oval/storm.git

    Create a Python 3.11 environment

    conda create -n storm python=3.11

    Install requirements from source

    pip install -r requirements.txt
  2. 2

    Run it

    • Install the package with pip install knowledge-storm
    • Set your language model API key (for example OPENAI_API_KEY) and search API key (for example YDC_API_KEY) as environment variables
    • Create a STORMWikiLMConfigs object and a retrieval module such as YouRM
    • Create a STORMWikiRunner and call run with a topic
    • Call post_run and summary to review the results
  3. 3

    Get the most out of it

    • Use a cheaper model for conversation simulation and a stronger model for article generation, as the README suggests
    • Try different retrievers such as BingSearch, VectorRM for your own documents, or DuckDuckGoSearchRM
    • Use the do_research flag to reuse saved research results instead of repeating searches
    • Try Co-STORM to add your own input to the discussion and build a mind map of the topic
    • Check the STORMWikiRunnerArguments class for more configuration options

Downloads

Source code (.zip)main branch · straight from GitHubRelease v1.1.0Release notes and sourceOfficial website & docsstorm.genie.stanford.edu

Strengths

  • Modular design: language models and search engines can be swapped, per the README's litellm and retriever support
  • Produces outlines and cited text rather than a single unsourced answer
  • Open source under the MIT license with an active research and release history

Trade-offs

  • The README says output is not publication-ready and often needs significant editing
  • Requires configured language model and search API keys, which may carry costs
  • Installation and customization assume Python and some comfort with code

How it compares

ProjectStarsTypeLevelUpdatedBest for
storm (this)31.8kAI model / agentintermediate1.1 years agoDevelopers who want to build a customizable research or report-writing pipeline
langchain149kFrameworkintermediate4 days agoPython developers building LLM-powered applications
FinGPT21.5kAI model / agentadvanced11 days agoDevelopers who want to experiment with finance-tuned LLMs for sentiment analysis
LightRAG40.5kLibraryintermediate7 days agoDevelopers building question-answering over their own document collections
deer-flow84.9kAI model / agentadvanced4 days agoDevelopers who want to experiment with multi-agent workflows on their own infrastructure
ai-agents-for-beginners78.7kLearning resourcebeginner4 days agoDevelopers new to AI agents who want guided lessons with code

Stats synced from GitHub. The assessment above was drafted with AI from the project's README on 2026-10-29 and may miss recent changes — check the README before relying on a command. MIT licence: free to use, modify and distribute, including commercially, as long as you keep the copyright notice.