Skip to main content

FireCrawlLoader

This notebook provides a quick overview for getting started with FireCrawlLoader. For detailed documentation of all FireCrawlLoader features and configurations head to the API reference.

Overview​

Integration details​

ClassPackageLocalSerializablePY support
FireCrawlLoader@langchain/community🟠 (see details below)betaβœ…

Loader features​

SourceWeb LoaderNode Envs Only
FireCrawlLoaderβœ…βŒ

FireCrawl crawls and convert any website into LLM-ready data. It crawls all accessible sub-pages and give you clean markdown and metadata for each. No sitemap required.

FireCrawl handles complex tasks such as reverse proxies, caching, rate limits, and content blocked by JavaScript. Built by the mendable.ai team.

This guide shows how to scrap and crawl entire websites and load them using the FireCrawlLoader in LangChain.

Setup​

To access FireCrawlLoader document loader you’ll need to install the @langchain/community integration, and the @mendable/firecrawl-js package. Then create a FireCrawl account and get an API key.

Credentials​

Sign up and get your free FireCrawl API key to start. FireCrawl offers 300 free credits to get you started, and it’s open-source in case you want to self-host.

Once you’ve done this set the FIRECRAWL_API_KEY environment variable:

export FIRECRAWL_API_KEY="your-api-key"

If you want to get automated tracing of your model calls you can also set your LangSmith API key by uncommenting below:

# export LANGCHAIN_TRACING_V2="true"
# export LANGCHAIN_API_KEY="your-api-key"

Installation​

The LangChain FireCrawlLoader integration lives in the @langchain/community package:

yarn add @langchain/community @mendable/firecrawl-js

Instantiation​

Here’s an example of how to use the FireCrawlLoader to load web search results:

Firecrawl offers 2 modes: scrape and crawl. In scrape mode, Firecrawl will only scrape the page you provide. In crawl mode, Firecrawl will crawl the entire website.

Now we can instantiate our model object and load documents:

import "@mendable/firecrawl-js";
import { FireCrawlLoader } from "@langchain/community/document_loaders/web/firecrawl";

const loader = new FireCrawlLoader({
url: "https://firecrawl.dev", // The URL to scrape
apiKey: "...", // Optional, defaults to `FIRECRAWL_API_KEY` in your env.
mode: "scrape", // The mode to run the crawler in. Can be "scrape" for single urls or "crawl" for all accessible subpages
params: {
// optional parameters based on Firecrawl API docs
// For API documentation, visit https://docs.firecrawl.dev
},
});

Load​

const docs = await loader.load();
docs[0];
Document {
pageContent: "Introducing [Smart Crawl!](https://www.firecrawl.dev/smart-crawl)\n" +
" Join the waitlist to turn any web"... 18721 more characters,
metadata: {
title: "Home - Firecrawl",
description: "Firecrawl crawls and converts any website into clean markdown.",
keywords: "Firecrawl,Markdown,Data,Mendable,Langchain",
robots: "follow, index",
ogTitle: "Firecrawl",
ogDescription: "Turn any website into LLM-ready data.",
ogUrl: "https://www.firecrawl.dev/",
ogImage: "https://www.firecrawl.dev/og.png?123",
ogLocaleAlternate: [],
ogSiteName: "Firecrawl",
sourceURL: "https://firecrawl.dev",
pageStatusCode: 500
},
id: undefined
}
console.log(docs[0].metadata);
{
title: "Home - Firecrawl",
description: "Firecrawl crawls and converts any website into clean markdown.",
keywords: "Firecrawl,Markdown,Data,Mendable,Langchain",
robots: "follow, index",
ogTitle: "Firecrawl",
ogDescription: "Turn any website into LLM-ready data.",
ogUrl: "https://www.firecrawl.dev/",
ogImage: "https://www.firecrawl.dev/og.png?123",
ogLocaleAlternate: [],
ogSiteName: "Firecrawl",
sourceURL: "https://firecrawl.dev",
pageStatusCode: 500
}

Additional Parameters​

For params you can pass any of the params according to the Firecrawl documentation.

API reference​

For detailed documentation of all FireCrawlLoader features and configurations head to the API reference: https://api.js.langchain.com/classes/langchain_community_document_loaders_web_firecrawl.FireCrawlLoader.html


Was this page helpful?


You can also leave detailed feedback on GitHub.