Show HN: Finic – Open source platform for building browser automations

Show HN: Finic – Open source platform for building browser automations(github.com)

143 points by jasonwcfan 1 year ago | 77 comments

Last year we launched a project called Psychic that did moderately well on hacker news, but was a commercial failure. We were able to find customers, but none with compelling and overlapping use cases. Everyone who was interested was too early to be a real customer.

This was our launch: https://news.ycombinator.com/item?id=36032081

We recently decided to revive and rebrand the project after seeing a sudden spike in interest from people who wanted to connect LLMs to data - but specifically through browsers. It's also a problem we've experienced firsthand, having built scraping features into Psychic and previously working on bot detection at Robinhood.

If you haven’t built a web scraper or browser automation before, you might assume it’s very straightforward. People have been building scrapers for as long as the internet has existed, so there must be many tools for the job.

The truth is that web scraping strategies need to constantly adapt as web standard change, and as companies that don’t want to be scraped adopt new technologies to try and block it. The old standards never completely go away, so the longer the internet exists, the more edge cases you’ll need to account for. This adds up to a LOT of infrastructure that needs to be set up and a lot of schlep developers have to go through to get up and running.

Scraping is no easier today than it was 10 years ago - the problems are just different.

Finic is an open source platform for building and deploying browser agents. Browser agents are bots deployed to the cloud that mimic the behaviour of humans, like web scrapers or remote process automation (RPA) jobs. Simple examples include scripts that scrape static websites like the SEC's EDGAR database. More complex use cases include integrating with legacy applications that don’t have public APIs, where the best way to automate data entry is to just manipulate HTML selectors (EHRs for example).

Our goal is to make Finic the easiest way to deploy a Playwright-based browser automation. With this launch, you can already do so in just 4 steps. Check out our docs for more info: https://docs.finic.io/quickstart

ghxst 1 year ago |

Cool service but how will you deal / how do you plan to deal with anti scraping and anti bot services like Akamai, Arkose, Cloudflare, DataDome etc.? Automation of the web isn't solved by another playwright or puppeteer abstraction, you need to solve more fundemental problems in order to mitigate the issues you run into at scale.

jasonwcfan 1 year ago | |

I mentioned this in another comment, but I know from experience that it's impossible to reliably differentiate bots from humans over a network. And since the right to automate browsers has survived repeated legal challenges, all vendors can do is make it incrementally harder to weed out the low sophistication actors.

This actually creates an evergreen problem that companies need to overcome, and our paid version will probably involve helping companies overcome these barriers.

Also I should clarify that we're explicitly not trying to build a playwright abstraction - we're trying to remain as unopinionated as possible about how developers code the bot, and just help with the network-level infrastructure they'll need to make it reliable and make it scale.

It's good feedback for us, we'll make that point more clear!

ghxst 1 year ago | | |

> but I know from experience that it's impossible to reliably differentiate bots from humans over a network

While this might be true in theory, it doesn't stop them from trying! And believe me, it's getting to a point where the WAF settings on some websites are even annoying the majority of the real users! Some of the issues I am hinting at however are fundemental issues you run into when automating the web using any mainstream browser that hasn't had some source code patches, I'm curious to see if a solution to that will be part of your service if you decide to tackle it.

candiddevmike 1 year ago | | |

Don't take this the wrong way, but this is the kind of unethical behavior that our industry should frown upon IMO. I view this kind of thing on the same level as DDoS-as-a-Service companies.

I wish your company the kind of success it deserves.

suriya-ganesh 1 year ago |

I've been working on browser agent the last week[1]. So this is very exciting. There are also browser agent implementations like Skyvern[2] (Also YC backed) ,or Tarsier[3] Seems like, finic is providing a way to scale/schedule these agents? If that's the case what's the advantage over something like airflow or windmill ?

If I remember correctly, Skyvern also has an implementation of scaling these browser tasks built in.

ps. Is it not called Robotic Process Automation? First time I'm hearing it as Remote process Automation.

[1]https://github.com/ProductLoft/arachne

[2]https://www.skyvern.com/

[3]https://github.com/reworkd/tarsier

mdaniel 1 year ago | |

https://github.com/reworkd/tarsier/pull/115/files represents someone who does not know what git is used for

  Cloning into 'tarsier'...
  remote: Enumerating objects: 15238, done.
  remote: Counting objects: 100% (1613/1613), done.
  remote: Compressing objects: 100% (929/929), done.
  Receiving objects: 100% (15238/15238), 3.01 GiB | 14.82 MiB/s, done.

ayanb9440 1 year ago | | |

Looks like somebody forgot to update the gitignore lol

ayanb9440 1 year ago | |

Yup that's right its Robotic Process Automation.

Based on the feedback in this thread we're going to be releasing an updated version that focuses more around tooling for the browser agents themselves as opposed to scaling/scheduling, so stay tuned for that!

mdaniel 1 year ago | |

And since the other two links are to GH: https://github.com/Skyvern-AI/skyvern (AGPLv3)

dataviz1000 1 year ago |

I build browser automation systems with either Playwright or Chrome Extensions. The biggest issue with automating 3rd party websites is knowing when the 3rd party developer pushes changes which break the automation. The way I dealt with that is run a headless browser in the cloud which checks the behavior of the automated site periodically sending emails and sms messages when it breaks.

If you don't already have this feature for your system, I would recommend it.

ghxst 1 year ago | |

IO between humans and websites can be broken down to only a few fundamental pieces (or elements I should say). This is actually where AI has a lot of opportunity to add value as it has the capability of significantly reducing the possibilty of breakage between changes.

ayanb9440 1 year ago | |

That's a great suggestion! Essentially a cron job to check for website changes before your automation runs and possibly breaks.

What does this check look like for you? Do you just diff the html to see if there are any changes?

dataviz1000 1 year ago | | |

The issue with diffing html is selectors are autogenerated with any update to a website's code. Often website which combat scraping will autogenerate different HTML. First thing is to screen caption a website for comparison. Second, it is possible to determine all the visible elements on a page. With Playwright, inject event listeners to all elements on a page and start automated clicking. If the agent fills out forms, then make sure that all fields are available to populate. There are a lot of heuristics.

Oras 1 year ago |

Don't take this as a negative thing, but I'm confused. Is it a playwright? Is it a residential proxy? It's not clear from your video.

jasonwcfan 1 year ago | |

Proxies are definitely on our roadmap, but for now it just supports stock Playwright.

Thanks for the feedback! I just updated the repo to make it more clear that it's Playwright based. Once my cofounder wakes up I'll see if he can re-record the video as well.

ghxst 1 year ago | | |

What kind of proxies are on your road map, do you have any experience with in-house proxy networks?

mdaniel 1 year ago |

> Finic uses Playwright to interact with DOM elements, and recommends BeautifulSoup for HTML parsing.

I have never, ever understood anyone who goes to the trouble of booting up a browser, and then uses a python library to do static HTML parsing

Anyway, I was surfing around the repo trying to find what, exactly "Safely store and access credentials using Finic’s built-in secret manager" means

ayanb9440 1 year ago | |

We're in the middle of putting this together right now but it's going to be a wrapper around Google Secret Manager for those that don't want to set up a secrets manager themselves.

0x3444ac53 1 year ago | |

Often times websites won't load the HTML without executing the JavaScript. or uses JavaScript running client side to generate the entire page.

mdaniel 1 year ago | | |

I feel that we are in agreement for the cases where one would use Playwright, and for damn sure would not involve BS4 for anything in that case

msp26 1 year ago | |

What would you recommend for parsing instead?

mdaniel 1 year ago | | |

In this specific scenario, where the project is using *automated Chrome* to even bother with the connection, redirects, and bazillions of other "browser-y" things to arrive at HTML to be parsed, the very idea that one would `soup = BeautifulSoup(playright.content())` is crazypants to me

I am open to the fact that html5lib strives to parse correctly, and good for them, but that would be the case where one wished to use python for parsing to avoid the pitfalls of dragging a native binary around with you

ghxst 1 year ago | | |

In python specifically I like lxml (pretty sure that's what BS uses under the hood?), parse5 if you're using node is usually my go to. Ideally though you shouldn't really have to parse anything (or not much at all) when doing browser automation as you have access to the DOM which gives you an interface that accepts query selectors directly (you don't even need the Runtime domain for most of your needs).

krick 1 year ago |

Does anyone know solid (not SaaS, obviously) solution for scraping these days? It's getting pretty hard to get around some pretty harmless cases (like bulk-downloading MY OWN gpx tracks from some fucking fitness-watch servers), with all these js tricks, countless redirects, cloudflare and so on. Even if you already have the cookies, getting non-403 response to any request is very much not trivial. I feel like it's time to upgrade my usual approach of python requests+libxml, but I don't know if there is a library/tool that solves some of the problems for you.

whatnotests2 1 year ago |

With agents like Finic, soon the web will be built for agents, rather than humans.

I can see a few years from now almost all web traffic is agents.

jasonwcfan 1 year ago | |

Yep. I used to be the guy responsible for bot detection at Robinhood so I can tell you firsthand it's impossible to reliably differentiate between humans and machines over a network. So either you accept being automated, or you overcorrect and block legitimate users.

I don't think the dead internet theory is true today, but I think it will be true soon. IMO that's actually a good thing, more agents representing us online = more time spent in the real world.

candiddevmike 1 year ago | | |

That is some bizarre mental gymnastics to justify the work you've done. What about the rest of us who don't want agents representing us?

j0r0b0 1 year ago |

Thank you for sharing!

Your sign up flow might be broken. I tried creating an account (with my own email), received the confirmation email, but couldn't get my account to be verified. I get "Email not confirmed" when I try to log in.

Also, the verification email was sent from accounts@godealwise.com, which is a bit confusing.

jasonwcfan 1 year ago | |

Oops! We tested the Oauth flow but forgot to update the email one. Thanks for the heads up, fixing this now.

ayanb9440 1 year ago | |

This should be fixed now

skeptrune 1 year ago |

I wonder if there are hidden observality problems with scraping with ideal solutions of a different shape than a dashboard. Feels like sentry connection or other common alert monitoring solutions would combine well with the LLM proposed changes and help trams react more quickly to pipeline problems.

ayanb9440 1 year ago | |

We do support sentry. Finic projects are poetry scripts so you can `poetry add` any observability library you need.

computershit 1 year ago |

First, nice work. I'm certainly glad to see such a tool in this space right now. Besides a UI, what does this provide that something like Browserless doesn't?

jasonwcfan 1 year ago | |

Thanks! Wasn't familiar with Browserless but took a quick look. It seems they're very focused on the scraping use case. We're more focused on the agent use case. One of our first customers turned us on to this - they wanted to build an RPA automation to push data to a cloud EHR. The problem was it ran as a single page application with no URL routing, and had an extremely complex API for their backend that was difficult to reverse engineer. So automating the browser was the best way to integrate.

If you're trying to build an agent for a long-running job like that, you run into different problems: - Failures are magnified as a workflow has multiple upstream dependencies and most scraping jobs don't. - You have to account for different auth schemes (Oauth, password, magic link, etc) - You have to implement token refresh logic for when sessions expire, unless you want to manually login several times per day

We don't have most of these features yet, but it's where we plan to focus.

And finally, we've licensed Finic under Apache 2.0 whereas Browserless is only available under a commercial license.

sahmeepee 1 year ago | | |

Sounds like a prooblem that can be solved with a Playwright script with a bit of error checking in it.

I think this needs more elaboration on what the Finic wrapper is adding to stock Playwright that can't just be achieved through more effective use of stock Playwright.

ushakov 1 year ago |

I do not understand what this actually is. Any difference between Browserbase and what you’re building?

Also, curious why your unstructured idea did not pan out?

ayanb9440 1 year ago | |

Looking at their docs, it seems that with Browserbase you would still have to deploy your Playwright script to a long-running job and manage the infra around that yourself.

Our approach is a bit different. With finic you just write the script. We handle the entire job deployment and scaling on our end.

ilrwbwrkhv 1 year ago |

Backed by YC = Not open source. Eventually pressure to exit and hyper scale will take over.

ayanb9440 1 year ago | |

There are quite a few open source YC startups at this point. Our understanding is that:

1. Developer tooling should be open source by default 2. Open source doesn't meaningfully affect revenue/scaling because developers that would use your self-hosted version would build in-house anyway.

ilrwbwrkhv 1 year ago | | |

I know there are quite a few open source by default companies. But the ethos of open source is sharing / building something by the community and getting paid in a way which does not scale the way VC funding expectations work.

So to have some respect for the open source way on top of which you are building all this please stop advertising it as "open source infrastructure" in bold and sell it like a normal software product with "source available" on the footer.

If you do plan to go open source and actually follow its ethos, remove the funded by VC label and have self hosting front and center in the docs with the hosted bit somewhere in the footer.

yard2010 1 year ago | |

I'm curious, can't do both?

slewis 1 year ago |

Is it stateful? Like can I do a run, read the results, and then do another run from that point?

ayanb9440 1 year ago | |

We currently don't save the browser state after the run has completed but that's something we can definitely add as a feature. Could you elaborate on your use case? In which scenarios would it be better to split a run into multiple steps?

mdaniel 1 year ago | | |

Almost any process that involves the word "workflow" (my mental model is one where the user would press alt-tab to look up something else in another window). The very, very common case would be one where they have a stupid SMS-based or "click email link" login flow: one would not wish to do that a ton, versus just leaving the session authenticated for reuse later in the day

Also, if my mental model is correct, the more browsing and mouse-movement telemetry those cloudflare/akamai/etc gizmos encounter, the more likely they are to think the browser is for real, versus encountering a "fresh" one is almost certainly red-alert. Not a panacea, for sure, but I'd guess every little bit helps

sebmellen 1 year ago |

We use https://windmill.dev which is great for this!