Show HN: Simplex: Automate browser workflows using code and natural language

Show HN: Simplex: Automate browser workflows using code and natural language simplex.sh/playground...

Simplex is an SDK that provides low-level functions you can use to create reliable web automations using code and natural language.

Here's a quick video to show you how it works: https://www.loom.com/share/8d27d0f9e0f34c77a81b0ac770151c12

A couple weeks ago, we needed a way to automatically find 3D assets from the internet for one of our work contracts. We didn’t need an AI to choose all the steps for us autonomously — we knew generally what the flow to find items should be and just needed some intelligence to handle different webpage formats.

Playwright couldn’t easily generalize to the different sites we wanted to search over and Claude computer use was tough to use — it was slow, expensive, struggled to click on the correct web elements, and there was no way to correct its actions if it failed at step 20 of our workflow.

So we built a vision-based solution using vision language models that sits on top of Playwright. In our SDK, building a function that can universally search websites is as easy as this:

    from simplex import Simplex
    simplex = Simplex(api_key="api_key")
    
    def search(website, search_query):
        simplex.goto(website)
        simplex.click("the search bar")
        simplex.type(search_query)
        simplex.press_enter()
        
    search("mit.edu", "alyssa p hacker")

You can play around with what we’ve built and see a few more examples at https://simplex.sh/playground.

We'd love feedback!

Great demo, straight to the point. It might be nice to have some kind of feedback mechanism if it can't find the element on the page, or if it's partially cut off. For example, I changed the GitHub profile to my own (https://github.com/spro) in the example and it doesn't scroll down far enough for the whole image. I imagine in general it would be nice to scroll to an element ID (or even element description using the vision models) instead of a hardcoded value.

Side-note: The comment for the frequency graph is wrong, it mentions stars instead.

marcon680 · a year ago

RE: feedback mechanism -- yep, a feedback mechanism is definitely something we're thinking of adding. Since we use VLMs that are trained to always output coordinates (i.e., they don't have a way to say "not on the page"), we're probably going to try fine tuning with some negative examples to see if we can build that feedback mechanism in.

One way to hack the scrolling to an element is to first run extract_bbox on a natural language description (in your case for GitHub it might be "follow button") then take the Y coordinate of that element and scroll that number of pixels. I just wrote this bit of code that I tested and it brings the contribution graph into full view:

    simplex.goto("github.com/spro")
    coords = simplex.extract_bbox("follow button")
    simplex.scroll(coords[1])
    simplex.wait(2000)
    image = simplex.extract_image("green tile graph")

But then it incorrectly picks the code review/submissions/etc. graph as the green tile graph -- we'll look into it!

re: frequency graph typo -- just pushed a fix, thanks!

simplex.goto("www.amazon.com") simplex.wait(2000) simplex.click('the search bar') simplex.type("bicycle") simplex.press_enter() simplex.wait(1000) image = simplex.extract_image("first 3 rows of results")

jackienotchan · a year ago

Would you mind sharing the story behind your pivot from on-demand photorealistic vision datasets[0] to browser automation?

[0] https://www.ycombinator.com/launches/Lbx-simplex-on-demand-p...

Yeah sure! We mentioned a little bit in the post that we stumbled upon this problem when working on a synthetic data contract. Internally, we were building software at the time to fully automate the process of creating synthetic data, all the way from purchasing assets to building scene layouts to rendering images. We realized after 2 months of building that there wasn't a great need for this problem nor could we build something minimal but feature-complete that we could give to people and iterate off of. We also saw the writing on the wall re: building 3D world models (see WorldLabs and NVIDIA's new 3D world models). We're excited to see where this goes!

TypingOutBugs · a year ago

It's quite a crowded market, browser workflow automation. What are you guys trying to do differently? For me, as someone who does a bunch of browser automation tasks, the real issue is stability. So many tools fall over on non-trivial workflows like complex forms with tabs. Fix that and you may open up a large testing automation market.

Mandatum · a year ago

VC funding.

globnomulous · a year ago

Apparently a simplex is a concept in geometry. I'm guessing this is the intended meaning of the term in the name. When most decently well informed, native speakers of English hear "simplex," though, they are likely to hear "herpes simplex" and so are likely to wonder, as I did, whether, or why, the program is named after a sexually transmitted disease.

gashmol · a year ago

I thought it's related to the simplex method in linear programming

TeMPOraL · a year ago

Which IIRC is named for that geometry thingie GP mentioned, that this method uses to represent the problem, with the solutions being vertices/edges of the thingie.

Interesting, TIL herpes is top of mind for me, but not for most of HN.

If you gave me the chance to come up with 500 things related to simplex, herpes would not be there.

IanCal · a year ago

I would never have made that link.

internet_points · a year ago

I for one did not make that connection. Only thing I thought of was https://simplex.chat/ (which is fortunately not marketed as a dating app)

sprobertson · a year ago

thasaleni · a year ago

tried:

get error:

  simplex error processing error

So sorry about this, it looks like our servers went down overnight. Just fixed, should be good now!

xnx · a year ago

Looks pretty cool. How do you distinguish Simplex from Skyvern or UI.Vision?

Hey thanks! We think we're pretty different from Skyvern as Skyvern provides a full agent loop for users + is no code. We wanted to build something at a lower level (i.e., no high-level planner that chooses what tasks to do for you) and we wanted to be able to program our own logic alongside the intelligence part because using code is what's natural for us as developers.

I hadn't heard of UI Vision but just took a look at it -- it also looks like a no-code solution that's a Chrome extension, so I'd say the main differences are the same as the differences w/ Skyvern -- we're lower level and meant to be used by developers.

I'd add that we're also able to directly extract parts of websites that have no official API -- for example, an image of the GitHub contribution graph like I show in the video demo.

johntash · a year ago

Very cool, any chance of either open-sourcing it or allowing the browser part to be self-hosted? i.e. to act on websites hosted in a local lan/vpn?

Also, did you evaluate https://github.com/browser-use/browser-use by any chance and have any comments about it? I'm assuming it was too AI-heavy based on what you said about claude/etc?

Thanks John! We should make that more clear in the docs. You can set browser=None on initialization and Simplex will create a local browser instance that can run on your local websites. We're not planning to be open source right now since a large portion of our product is custom vision models + inference speedups through hosting.

Browser Use is another YC company. Probably the biggest difference is that they're more agent focused while we're more lower level -- in the Claude Computer Use camp like you mentioned.

mkagenius · a year ago

VLMs are great - I have been able to use it for a similar project too [1]. And it's only going to get better. Congratulations on the product launch what's your VLM model for this?

1. A framework to use/control mobile phones via any LLM - https://github.com/BandarLabs/clickclickclick

We finetune our own VLMs for this -- unfortunately prefer not to share which ones we use specifically! ClickClickClick looks awesome, have you heard of FerretUI (https://arxiv.org/pdf/2404.05719)? Pretty similar idea.

Yes, I tried a similar one called "omniparser" - where the issue was it was missing annotating some UI elements sometimes. Moreover, Gemini and Molmo worked right out of the box without needing any fine tune.

I'm surprised you named your framework clickclickclick instead of taptaptap.

vitalets · a year ago

I've tried the following code:

    simplex.goto("github.com/mitchellh")
    num_contribs = simplex.extract_text("blabla")
    print(num_contribs)

It outputs all texts from the page. Is it expected? Maybe it should fail indicating element could not be found?

skarpoor · a year ago

extract_text returns all the lines of text it finds within an element. Looks like for your case it selected a majority of the page from the description, so it's returning most of the text on the page.

We currently don't return failure cases (just closest match) -- but good suggestion! We'll fine tune on some negative cases and see if we can catch them