license: cc0-1.0
dataset_info:
features:
- name: Domain
dtype: string
- name: File
dtype: string
- name: URL
dtype: string
- name: Content
dtype: string
splits:
- name: train
num_bytes: 53016572
num_examples: 172
download_size: 19041868
dataset_size: 53016572
configs:
- config_name: default
data_files:
- split: train
path: data/train-*
Context & Motivation
https://llmstxt.org/ is a project from Answer.AI which proposes to "standardise on using an /llms.txt
file to provide information to help LLMs use a website at inference time."
I've noticed many tool providers begin to offer /llms.txt
files for their websites and documentation. This includes developer tools and platforms like Perplexity, Anthropic, Hugging Face, Vercel, and others.
I've also come across https://directory.llmstxt.cloud/, a directory of websites that have /llms.txt
files which is curated by these folks: https://x.com/llmsdottxt. I thought it would be fun to use this awesome resource to collect all of the files into a single dataset. They're simply markdown files. This dataset can then be used to build cool applications.
Thank you to Answer.AI and Jeremy Howard, the providers that are adopting this standard, and the maintainers of https://directory.llmstxt.cloud/.
How this dataset was made
This is the notebook that fetches files that linked to from https://directory.llmstxt.cloud/ and uses the kagglehub
Python client library to publish the resulting output as this dataset.
Inspiration
- Give your LLM application access to this dataset to enhance its interactions with these tools, e.g., for code-generation tasks
- Search and knowledge retrieval
- Extract and summarize common developer tasks to generate novel benchmarks for LLM evaluation
- Validate the correctness of the llms.txt files
Contributing
I'd love if anyone is interested in contributing to improving the notebook that extracts the llms.txt
files. Leave a comment on this dataset or on the notebook. Feel free to also ping me with interesting demos or applications you create with this dataset.
Photo by Solen Feyissa on Unsplash