lucgagan

Here goes my project.

Can confirm. My website is flooded with AI bots despite attempts to block crawlers to certain parts of it.

Correct me if I am wrong, but these implementations are all CPU bound?, i.e. if I have a good GPU, I should look for alternatives.

Woah, this is very _fun_!

Something similar I worked on in the past https://github.com/lucgagan/auto-playwright/

Very cool. I've been using a similar approach to summarize videos on my website https://ray.run/videos/112-playwright-installation-2022 Doesn't cost much and getting great feedback about it.

Is there a way to try this before I buy it?

I am building something that would probably benefit from this, but with that price tag (solo indy dev) that's going to be a big ask! might be worth it, just no way of knowing without trying it first

Thanks for the response. I paused for a day responding to everyone to think through all the feedback I've received, and shared an update https://github.com/lucgagan/auto-playwright/issues/15

Theirs is a commercial project with MIT client. I made an open-source version of their project. (Their backend is closed source. I have no knowledge of it.)

If I was the creator of the project, I'd be happy with it. Most companies would not have donated anything.

Most people I would imagine

This looks great. I wish I had this a few months ago! Giving it a try.

Hello HN community!

I'm excited to introduce Auto Playwright, an open-source tool that revolutionizes web testing by integrating AI with Playwright. Imagine writing your web app tests in plain, simple English and having AI turn those instructions into automated tests!

Key Features: - AI-Driven Test Writing: Just write what you need in plain text, and let the AI do the rest. - Playwright Compatibility: Works with all browsers compatible with Playwright. - Open-Source: Dive into the code, contribute, and customize as you see fit.

How It Works: 1. Install `auto-playwright` via npm. 2. Set up your OpenAI API key. 3. Write tests using intuitive, plain-text prompts.

Examples: From testing a search function with a few lines of text to automating complex e-commerce cart tests – it's all made incredibly simple.

    test("auto Playwright example", async ({ page }) => {
      await page.goto("/");

      // `auto` can query data
      // In this case, the result is plain-text contents of the header
      const headerText = await auto("get the header text", { page, test });

      // `auto` can perform actions
      // In this case, auto will find and fill in the search text input
      await auto(`Type "${headerText}" in the search box`, { page, test });

      // `auto` can assert the state of the website
      // In this case, the result is a boolean outcome
      const searchInputHasHeaderText = await auto(`Is the contents of the search box equal to "${headerText}"?` { page, test });

      expect(searchInputHasHeaderText).toBe(true);
    });

Get Involved: I am building a community around Auto Playwright. Join us on Discord (https://ray.run/discord), contribute to the GitHub repo (https://github.com/lucgagan/auto-playwright), and stay updated with our newsletter (https://ray.run/newsletter).

Looking forward to your thoughts, feedback, and seeing how you use Auto Playwright in your projects!

Check out Auto Playwright on GitHub:

https://github.com/lucgagan/auto-playwright

Ended up publishing to mainspace as per the guidance of a few commenters.

https://en.wikipedia.org/wiki/Playwright_(software)

Thanks!

Contributing to Wikipedia has become really not fun. Or maybe it never was. I tried writing an article about Playwright - perhaps the most common test automation tool these days. It first got rejected and now just has been sitting in review state for 3 months.

https://en.wikipedia.org/wiki/Draft:Playwright_(software)

It is highly demotivating to try to write a quality article for Wikipedia because someone can just reject your days of work in seconds and then leave it in draft forever.

I am building _exactly_ the same thing for Playwright over at https://ray.run/. I think this is the future of writing tests no doubt. Planning to launch next week.

By the looks of it, I am taking a slightly different approach than you. I am using LLM only to identify elements, but the actual generators are created using https://ray.run/browser-extension

Their demo crashes when I try to use it

It honestly looks not too bad!

Took a bit of time to customize it, but you can match your website style pretty closely.

https://cse.google.com/cse?cx=210b5e0b95aee4c07&q=test#gsc.t...

The only downside is that it seems you cannot disable promotions/ads

Very interesting. You are even allowed to upload your own search annotations, i.e. suggestions for search results. https://cse.google.com/cse?cx=210b5e0b95aee4c07#gsc.tab=0 The end result is not too bad considering it is entirely free!

Looks like Google has a way of doing it https://programmablesearchengine.google.com/controlpanel/cre...

so I tried migration, but quickly learned that Astro is not great for my use case

I also tried Remix and that went a bit better. However, until they add support RSC, Next.js DX remains superior

Seems like it is disclosed so shouldn't be "news"?

From my experience with Rust, there is nothing fast about developing with it.

Crazy how many things a single thing breaking takes down with it

Why the focus on synthetic monitoring? As a SRE, I actively eschew synthetic monitoring. It's highly error prone and doesn't actually indicate regional availability. I'd like a status site that I could push a certain internally derived SLA for a given service to and the status site reflects the average over time of that windowed SLA.

As an end user, hard disagree.

GitHub is a great example of this. Their status almost always shows 100% uptime while the service is entirely unstable.

It is clear that their uptime SLAs do not align with end user experience.

As an end user, I care whether I can access and use the service. I don't care what broke in between.

I never understood the "de-risk" things angle. Is the idea that you'd self host if the service went under?

This was very interesting.

For example — we at SensusQ are building a piece of software that can ingest a large amount of disparate information pieces, be that drone images, satellite pictures, social media posts or text documents.

Congrats!

Do you have an article about how you got into it?

What are some startups in your area that are software?

What newsletters do you subscribe?

I would love to get more in the loop.

I read (most of) the article and was thinking "got to be the only guy in the World", and the first comment on HN is another guy doing the same thing. World is a big place and HN is def a selection.

As a counter point, I've been actively assessing feasibility of migrating my experiments to Claude with not much success. Trying to migrate prompts from GPT 4 to Claude often yields half-baked results. Claude seems to require a lot more handholding to get the desired result.