LLM Scraper

LLM Scraper is a TypeScript library that allows you to convert any webpages into structured data using LLMs.

[!TIP] Under the hood, it uses function calling to convert pages to structured data. You can find more about this approach here

Features

• Supports Local (GGUF), OpenAI, Groq chat models
• Schemas defined with Zod
• Full type-safety with TypeScript
• Based on Playwright framework
• Streaming when crawling multiple pages
• Supports 4 input modes:
  • html for loading raw HTML
  • markdown for loading markdown
  • text for loading extracted text (using Readability.js)
  • image for loading a screenshot (multi-modal only)

Make sure to give it a star!

Getting started

1. Install the required dependencies from npm:

   npm i zod playwright llm-scraper
   

2. Initialize your LLM:

   OpenAI

   import OpenAI from 'openai'
   const model = new OpenAI()
   

   Groq

   import OpenAI from 'openai'
   const model = new OpenAI({
    baseURL: 'https://api.groq.com/openai/v1',
    apiKey: process.env.GROQ_API_KEY,
   })
   
   // Note!
   // Tool calls are used so only the models listed here are supported for Groq
   // when invoking any .run(...) call
   // https://console.groq.com/docs/tool-use
   

   Local

   import { LlamaModel } from 'node-llama-cpp'
   const model = new LlamaModel({ modelPath: 'model.gguf' })
   

3. Create a new browser instance and attach LLMScraper to it:

   import { chromium } from 'playwright'
   import LLMScraper from 'llm-scraper'
   
   const browser = await chromium.launch()
   const scraper = new LLMScraper(browser, model)
   

Example

In this example, we're extracting top stories from HackerNews:

import { chromium } from 'playwright'
import { z } from 'zod'
import OpenAI from 'openai'
import LLMScraper from 'llm-scraper'

// Launch a browser instance
const browser = await chromium.launch()

// Initialize LLM provider
const llm = new OpenAI()

// Create a new LLMScraper
const scraper = new LLMScraper(browser, llm)

// Define schema to extract contents into
const schema = z.object({
  top: z
    .array(
      z.object({
        title: z.string(),
        points: z.number(),
        by: z.string(),
        commentsURL: z.string(),
      })
    )
    .length(5)
    .describe('Top 5 stories on Hacker News'),
})

// URLs to scrape
const urls = ['https://news.ycombinator.com']

// Run the scraper
const pages = await scraper.run(urls, {
  model: 'gpt-4-turbo',
  schema,
  mode: 'html',
  closeOnFinish: true,
})

// Stream the result from LLM
for await (const page of pages) {
  console.log(page.data)
}

Contributing

As an open-source project, we welcome contributions from the community. If you are experiencing any bugs or want to add some improvements, please feel free to open an issue or pull request.