Stop Scraping RAW HTML into your LLM. Try our new feature

You're paying for your LLM to read ads. Every time you send raw HTML to a language model, it has to process boilerplate, image tags, script tags, and this just burns through tokens. So, let's fix that. Here at Zite, we have auto extract, and we've just added a new feature to it called page content. Page content will take the page and return to you structured JSON that has the head headline, the title, the main body of text, and the navigation in a neat JSON format. and it's very very simple to use. We're adding that to our existing line of auto extract templates which are product, article, jobs, etc. And I think it's a really cool feature that I think you should give a go. Now, if we look at this code here, this is basically just getting information from this returns policy. I'm asking for the browser HTML. So, we got a representation of what it is here and I'm saying give me the page content. Now, if I look at the actual HTML that the browser returns, we got all of this boiler plate. All of this stuff here that we just don't need and we don't want to pass to our LLM, it's going to waste so many tokens. There is 4,400 4,245 lines of text that needs to be processed here. Now, if I look at what comes out of our page object here, we have neat JSON, which is, you know, 200 odd lines. And here is what you get back. You get the breadcrumbs, the headline, the title, and the item main in text format. This has all of the text information that you need to give to your LLM, so you can process the information and get what you need out of it. It also comes with the navigation, so you could also use this to go through different pages if you so needed to. And it has some footer information, too. Now, we think this is a great way to get this information out and save you money through tokens. It obviously costs a little bit more to make this work through our site API. However, the cost is going to be greatly offset and it is very minimal depending on what pages you're trying to scrape. This feature is live now. So, if you have a Zite API account, you can go and try it. If you don't have a Zite API account, then I'll leave some links down below where you can find out more information. Go and get one, get some free credit, and give this a go and see if it fits into your use case.