Show HN: Crawlee for Python – a web scraping and browser automation library

jancurn · 2024-07-09T08:30:38.000Z

Hey all,This is Jan, the founder of Apify (https://apify.com/) — a full-stack web scraping platform. After the success of Crawlee for JavaScript (https://github.com/apify/crawlee/) and the demand from the Python community, we're launching Crawlee for Python today!The main features are:- A unified programming interface for both HTTP (HTTPX with BeautifulSoup) & headless browser crawling (Playwright)- Automatic parallel crawling based on available system resources- Written in Python with type hints for enhanced developer experience- Automatic retries on errors or when you’re getting blocked- Integrated proxy rotation and session management- Configurable request routing - direct URLs to the appropriate handlers- Persistent queue for URLs to crawl- Pluggable storage for both tabular data and filesFor details, you can read the announcement blog post: https://crawlee.dev/blog/launching-crawlee-pythonOur team and I will be happy to answer here any questions you might have.

mdaniel 1 year ago |

You'll want to prioritize documenting the existing features, since it's no good having a super awesome full stack web scraping platform if only you can use it. I ordinarily would default to a "read the source" response but your cutesy coding style makes that a non-starter

As a concrete example: command-f for "tier" on https://crawlee.dev/python/docs/guides/proxy-management and tell me how anyone could possibly know what `tiered_proxy_urls: list[list[str]] | None = None` should contain and why?

mnmkng 1 year ago | |

Sorry about the confusion. Some features, like the tiered proxies, are not documented properly. You’re absolutely right. Updates will come soon.

We wanted to have as many features in the initial release as possible, because we have a local Python community conference coming up tomorrow and we wanted to have the library ready for that.

More docs will come soon. I promise. And thanks for the shout.

arresin 1 year ago | | |

I literally had to go through the entire codebase the documentation is that lacking. It’s boring to document but imo it’s the lowest hanging fruit to get people moving down that crawlee -> appify funnel.

NoThisIsMe 1 year ago | |

> but your cutesy coding style makes that a non-starter

I don't think this is fair. The code looks pretty readable to me.

ttymck 1 year ago | |

Can you link to some of the cutesy code? I've never heard this before.

Findecanor 1 year ago |

Does it have support for web scraping opt-out protocols, such as Robots.txt, HTTP and content tags? These are getting more important now, especially in the EU after the DSM directive.

jancurn 1 year ago | |

Not yet, but it’s on the roadmap

nobodywillobsrv 1 year ago |

I don't really understand it. Tried it on some fund site and it didn't really do much besides apparently grepping for links.

The example should show how to literally find and target all data as in .csv .xlsx tables etc and actually download it.

Anyone can use requests and just get the text and grep for urls. I don't get it.

Remember: pick an example where you need to parse one thing to get 1000s of other things to then hit some other endpoints to then get the 3-5 things at each of those. Any example that doesn't look like that is not going to impress anyone.

I'm not even clear if this is saying it's a framework or actually some automation tool. Automation meaning it actually autodetects where to look.

renegat0x0 1 year ago |

I have been running my project with selenium for some time.

Now I am using crawlee. Thanks. I will work on it, to better integrate into my project, however I already can tell it works flawlessly.

My project, with crawlee: https://github.com/rumca-js/Django-link-archive

c0brac0bra 1 year ago |

Wanted to say thanks for apify/crawlee. I'm a long-time node.js user and your library has worked better than all the others I've tried.

jancurn 1 year ago | |

Thank you!

intev 1 year ago |

How is this different from Scrapy?

sauain 1 year ago | |

hey intev,

- Crawlee has out-of-the-box support for headless browser crawling (Playwright). You don't have to install any plugin or set up the middleware. - Crawlee has a minimalistic & elegant interface - Set up your scraper with fewer than 10 lines of code. You don't have to care about what middleware, settings, and anything are or need to be changed, on the top that we also have templates which makes the learning curve much smaller. - Complete type hint coverage. Which is something Scrapy hasn't completed yet. - Based on standard Asyncio. Integrating Scrapy into a classic asyncio app requires integration of Twisted and asyncio. Which is possible, but not easy, and can result in troubles.

mdaniel 1 year ago | | |

> You don't have to install any plugin or set up the middleware.

That cuts both ways, in true 80/20 fashion: it also means that anyone who isn't on the happy path of the way that crawlee was designed is going to have to edit your python files (`pip install -e` type business) to achieve their goals

8organicbits 1 year ago | | |

I've been working on a crawler recently and honestly you need the flexibility middleware gives you. You can only get so far with reasonable defaults, crawling isn't a one-size fits all kinda thing.

mnmkng 1 year ago | | |

Crawlee isn’t any less configurable than Scrapy. It just uses different, in my personal opinion more approachable, patterns. It makes it easier to start with, but you can tweak whatever you want. Btw, you can add middleware in Crawlee Router.

mdaniel 1 year ago | | |

> Crawlee isn’t any less configurable than Scrapy.

Oh, then I have obviously overlooked how one would be able to determined if a proxy has been blocked and evict it from the pool <https://github.com/rejoiceinhope/scrapy-proxy-pool/blob/b833...> . Or how to use an HTTP cache independent of the "browser" cache (e.g. to allow short-circuiting the actual request if I can prove it is not stale for my needs, which enables recrawls to fix logic bugs or even downloading the actual request-response payloads for making better tests) https://docs.scrapy.org/en/2.11/topics/downloader-middleware...

Unless you meant what I said about "pip install -e && echo glhf" in which case, yes, it's a "simple matter of programming" into a framework that was not designed to be extended

ranedk 1 year ago |

I found crawlee a few days ago while figuring out a stack for a project. I wanted a python library but found crawlee with typescript so much easier that I ended up coding the entire project in less than a week in Typescript+Crawlee+Playwright

I found the api a lot better than any python scraping api till date. However I am tempted to try out python with Crawlee.

The playwright integration with gotScraping makes the entire programming experience a breeze. My crawling and scraping involves all kinds of frontend rendered websites with a lot of modified XHR responses to be captured. And IT JUST WORKS!

Thanks a ton . I will definitely use the Apify platform to scale given the integration.

3abiton 1 year ago | |

How does it compare to selenium for python?

sauain 1 year ago | |

would love to have your feedback on the python one too :)

marban 1 year ago |

Nice list, but what would be the arguments for switching over from other libraries? I’ve built my own crawler over time, but from what I see, there’s nothing truly unique.

jancurn 1 year ago | |

The main advantage (for now) is that the library has a single interface for both HTTP and headless browsers, and bundled auto scaling. You can write your crawlers using the same base abstraction, and the framework takes care of this heavy lifting. Developers of scrapers shouldn't need to reinvent the wheel, and just focus on building the "business" logic of their scrapers. Having said that, if you wrote your own crawling library, the motivation to use Crawlee might be lower, and that's fair enough.

Please note that this is the first release, and we'll keep adding many more features as we go, including anti-blocking, adaptive crawling, etc. To see where this might go, check https://github.com/apify/crawlee

robertlagrant 1 year ago | | |

Can I ask - what is anti-blocking?

fullspectrumdev 1 year ago | | |

Usually refers to “evading bot detection”.

Detecting when blocked and switching proxy/“browser fingerprint”.

robertlagrant 1 year ago | | |

Is this a good feature to include? Shouldn't we respect the host's settings on this?

nlh 1 year ago | | |

It’s a fair and totally reasonable question but clashes with reality. Many hosts have data that others want/like to scrape (eBay, Amazon, Google, airlines, etc.) and they setup anti-scraping mechanisms to try and prevent scraping. Whether or not to respect those desires is a bigger question but not one for the scraping library - it’s one for those doing the scraping and their lawyers.

The fact is - many many people want to scrape these sites and there is massive demand for tools to help them do that, so if APIFY/Crawlee decide to take the moral ground and not offer a way around bot detection, someone else will.

VagabundoP 1 year ago |

Looks nice, and modern python.

The code example on the front page has this:

`const data = await crawler.get_data()`

That looks like Javascript? Is there a missing underscore?

mnmkng 1 year ago | |

Oh wow, thanks! Will fix it right away. Crawlee is originally a JS library.

fforflo 1 year ago |

I'd suggest bringing more code snippets from the test cases to documentation as examples.

Nice work though.

manishsharan 1 year ago |

Can this work on intranet sites like sharepoint or confluence , which require employee SSO ?

I was trying to build a small Langchain based RAG based on internal documents but getting the documents from sharepoint/confluence (we have both) is very painful.

mnmkng 1 year ago | |

Technically it can. You can log in with the PlaywrightCrawler class without issue. The question is if there’s 2FA as well and how that’s handled. Crawlee does not have any abstraction for handling 2FA as it depends a lot on what verification options are supported on the SSO side. So that part would need a custom implementation within Crawlee.

jancurn 1 year ago | |

For this use case, you might use this ready-made Actor: https://apify.com/apify/website-content-crawler

holoduke 1 year ago |

Does it have event listeners to wait for specific elements based on certain pattern matches. One reason i am still using phantomjs is because it simulates the entire browser and you can compile your own webkit in it.

mnmkng 1 year ago | |

It uses Playwright under the hood, so yes, it can do all of that, and more.

barrenko 1 year ago |

Pretty cool, and any scraping tool is really welcome - I'll try it out for my personal project. At the monment, due to AI, scraping is like selling shovels during a gold rush.

ijustlovemath 1 year ago |

Do you have any plans to monetize this? How are you supporting development?

sauain 1 year ago | |

Crawlee is open source and free to use and we don't have any plans to monetize it in future. It will be always free to use.

We provide Apify platform to publish your scrapers as Actors for the developer community, and developers earn money through it. You can use Crawlee for Python as well :)

tldr; Crawlee is and always will be free to use and open sourced.

renegat0x0 1 year ago |

Can it be used to obtain RSS contents? Most of examples focus on html

sauain 1 year ago | |

I didn't try it, but I don't see a reason why not:

- RSS feed is transferred via HTTP. - BeatifulSoup can parse both HTML & XML.

(RSS uses XML format)

bmitc 1 year ago |

Can you use this to auto-logon to systems?

jancurn 1 year ago | |

For sure, simply store cookies after login and then use them to initiate the crawl.

localfirst 1 year ago |

in one sentence, what does this do that existing web scraping and browser automation doesn't do?

mnmkng 1 year ago | |

In one word. Nothing.

But I personally think it does some things a little easier, a little faster and little more conveniently than the other libraries and tools out there.

Although there’s one thing that the JS version of Crawlee has which unfortunately isn’t in Python yet, but it will be there soon. AFAIK it’s unique among all libraries. It’s automatically detecting whether a headless browser is needed or if HTTP will suffice and using the most performant option.

localfirst 1 year ago | | |

is there anything that uses a computer vision model/ocr locally to extract data?

I find some dynamic sites purposefully make it extremely difficult to parse and they obfuscate the XHR calls to their API

I've also seen some websites pollute the data when it detects scraping which results in garbage data but you don't know until its verified

mnmkng 1 year ago | | |

We tried a self hosted OCR model a few years ago, but the quality and speed wasn’t great. From experience, it’s usually better to reverse engineer the APIs. The more complicated they are, the less they change. So it can sometimes be painful to set up the scrapers, but once they work, they tend to be more stable than other methods.

Data pollution is real. Also location specific results, personalized results, A/B testing, and my favorite, badly implemented websites are real as well.

When you encounter this, you can try scraping the data from different locations, with various tokens, cookies, referrers etc. and often you can find a pattern to make the data consistent. Websites hate scraping, but they hate showing wrong data to human users even more. So if you resemble a legit user, you’ll most likely get correct data. But of course, there are exceptions.

thelastgallon 1 year ago |

I wonder if there are any AI tools that do web scraping for you without having to write any code?

lifesaverluke 1 year ago | |

Which site would you like to scrape?