28 points ericciarla | 8 comments | | HN request time: 1.074s | source | bottom

Hey HN! It’s Eric from Firecrawl (https://firecrawl.dev).

I just launched llms.txt Generator, a tool that transforms any website into a clean, structured text file optimized for feeding to LLMs. You can learn more about the standard at https://llmstxt.org.

Here’s how it works under the hood:

1. We use Firecrawl, our open-source scraper, to fetch the full site, handling JavaScript-heavy pages and complex structures. 2. The markdown content is parsed and then the title and description are extracted using GPT-4o-mini. 3. The everything is combined and the result is a lightweight llms.txt file that you can paste into any LLM.

Let me know what you think!

1. jondwillis ◴[] No.42214547[source]
Plain HTTP and passing an API key as a URL query parameter? Yikes!
2. DrillShopper ◴[] No.42214869[source]
Thanks for facilitating even more widespread and frictionless copyright violations
replies(1): >>42220258 #
3. IndieCoder ◴[] No.42215031[source]
I like the idea but Firecrawl and GPT4o is quite heavy. I use https://github.com/unclecode/crawl4ai in some projects, it works very well without these dependencies and is modular so you can use LLMs but do not have to :)
4. throwaway314155 ◴[] No.42217413[source]
For a simple solution, you can just right click->Save Page As.. and upload the resulting `.html` file into Claude/ChatGPT as an attachment. They're both more than capable of parsing the article content from the HTML without needing any pre-processing.
5. e-clinton ◴[] No.42220258[source]
Wait till you hear about copy/paste
replies(2): >>42228406 #>>42229447 #
6. ◴[] No.42228406{3}[source]
7. DrillShopper ◴[] No.42229447{3}[source]
Not the same thing and not at the same scale