[wjnc]

An RPA-team that does work for me as a client told me that with a certain system they just use a 20 second window between actions. The robot is not much faster than colleagues, but a lot more sturdy and appreciative of menial work. I curse all developers of bigcorp software that do not create an API for all functionalities exposed to users. Likewise, we don’t praise those that do enough!

[anonzzzies]

Yes, that's the approach I take, but in that case, the problem is that Playwright with Chrome plugin + our own bag of scripts for detecting spinners etc, is faster AND more accurate than these AI attempts. These RPA things with LLMs would only work in case they can actually venture out alone and Get Shit Done. If I have to sit and wait and reprompt, I might as well just get the Playwright script which will work actually every time.

But I know it's early days! There is a reason I test ALL of these every few months.

[kordlessagain]

I built this with Playwright and OpenAI's function calling stuff (sorry, no time for docs): https://github.com/MittaAI/mitta-community/tree/main/service...

My thought was to put the results of this in a vector store, with any errors that resulted as opposed to wasting time training a model.

[leobg]

Why not combine both worlds? Have the AI (slow) figure out the path to your goal and write the (fast) script. Whenever the script breaks due to changes in the site, the AI will update it. Might take it half an hour. But no more human “janitor” in the loop. And no more broken scripts.

[wruza]

Why use hard timeouts if you can wait for selector and then 100ms more? Worked fine last time I automated some SPAs.

[anonzzzies]

Depends how well they are built; of course you can do that but it’s a lot of manual work for some badly made ones: timeouts is easier/faster for trying things out. But I agree, once all works, you can put selectors. However, AI cannot find those automatically, although that cannot be that hard either, it’s just not something I have seen anywhere yet.