HN user
with playwright-cli, the agent makes reasoning decisions between each cli call. with browser-tools, the agent can write big playwright code blocks that execute in one go meaning less turns and the agent can select exactly what it needs from the result.
I think the code version of playwright-cli is still supported but we just use the default instructions given by the cli which push it towards per cli call usage
If you’re using a local provider, it can open browsers in headed mode so it can log in. For remote providers, you’ll need to ask the agent for the live url which will be a proxy feed of the page to log in
As stated on the website, full methodology and results are available here: https://github.com/saffron-health/libretto/tree/main/package...
building something similar. libretto.sh
You should check out Libretto. It can take the playwright code and turn it into a script you can deploy
i dont think its detectable if done well. We’ve built involved automations with no problems. i used libretto not browserbase tho
Thanks! Please try it out. Stop by in the Discord or Github Issues if you have any questions!
Thanks! Please try it out. Stop by in the Discord or Github Issues if you have any questions!
Ofc! Please try it out. Stop by in the Discord or Github Issues if you have any questions!
Sorry! Yes, MIT. Forgot to lift it up when I converted to a monorepo, but it's in packages/libretto
glad to hear! Please reach out on Discord or Github issues you run into issues!
playwright-cli is very simple and meant for humans - it basically generates a first draft of a script, and was originally meant for writing e2e tests. You need to do a lot of post-processing on it to get it to be a reliable automation.
libretto gives a similar ability for agents for building scripts but:
- agents automatically run, debug, and test the integrations they write - they have a much better understanding of the semantics of the actions you take (vs. playwright auto-assuming based on where you clicked) - they can parse network requests and use those to make direct API calls instead
there's fundamentally a mismatch where playwright-cli is for building e2e test scripts for your own app but libretto is for building robust web automations
we started using stagehand initially! But it doesn't follow the same model of pre-generating deterministic code. Your code is meant to look like this:
// Let AI click await stagehand.act("click on the comments link for the top story");
the issue with this is that there's now runtime non-determinism. We move the AI work during dev-time: AI explores and crawls the website first, and generates a deterministic legible script.
Tangentially, Stagehand's model may have worked 2 years ago when humans still wrote the code, but it's no longer the case. We want to empower agents to do the heavy lifting of building a browser automation for us but reap the benefits of running deterministic, fast, cheap, straightforward code.
just adding to michael's reply - we took care to make sure no PHI was exposed in our demo video as well.
this is only gonna be an issue until the next gen models where the labs will aggressively post train the models to proactively call skills
I've been using Triplit on my React native app for a while and it works great. Highly recommend. It's the only local-first db that hits these points for me:
- Good sane query language (not SQL) - Great typescript support - Offline support - React native support
The cherries on top is that it's open source and self-hostable.
not the op, but one reason might be to visually distinguish between two events at the same time on adjacent days versus a 2-day event. Without the gap, events blend together. I don’t think the border provides enough contrast to rely on either.
How does the play animation work? Do you get some sort of json blob output? How do you make sure it fits your schema?
holy shit. How does this work? Do you just have a big list?
Looks cool! I love rss. Have you considered also letting people publish to their own rss feeds from directly in the app? I wonder if having a basic content creation side is the missing piece of a lot of rss feeder apps
What are some the popular patterns to avoid in javascript that get you good performance? Or phrased another way, what are the changes to make today that would likely have the most performance impact in a js codebase?
I’m glad it left that impression! My thoughts have clarified a bit since I read that post, and I think what I describe is more declarative, like React. But the best places to read about it (for web devs) are in Elm!
There is also this new thing I found that seems to really lean into the core of what being functional means here: https://github.com/jorgebucaran/hyperapp
After a while, you see that basically all systems can be modeled as event-driven, functional systems. It’s a flexible model, and fits beautiful into web dev where the semantics are very clear: the system is the web app and events are clicks, keyboard events, asynchronous calls...
But that’s my vibe!
All good, I’m glad you liked it! I’ve already posted it once, but it had no traction last time for some reason.
Thanks!
Thank you so much! I’m so glad it gave off that impression. I went through a lot of revision and planning to write it: I had learned a lot, and wanted to organize all of it carefully. I’m glad it paid off!
Im not sure if there’s a good way to use css to animate an arbitrary number of children with a delay, while retaining clean markup. I did this before: .article-child { animation-delay: var(—index)*0.025s; } and while rendering, set style tags on each element style=“—index:5”, etc. That works, but it was annoying to maintain, and every single element has these style tags. So I just use Motion One (WAAPI) now.
Hi! I’m the creator of the post. Thank you for all the comments on my post! I’m glad more people are getting to see it. Thanks for posting, @antidnan!
Reading through, it looks like I have some rough edges of my website to clean up as well...
Ah! That’s unfortunate . All that work of pre-rendering, and it won’t even work with javascript disabled. To be clear, I use a very very tiny bit of javascript only to animate the article elements after render.I guess when that fails, it just doesn’t animate the opacity to 1…hm...