feat: add option to perform GPT prompts over items from dataset without crawling

[Patai5]

Discussed in slack:
https://apify.slack.com/archives/CGZSN9DQC/p1729755239716379

---

We should add an option to perform basic prompting for dataset items as string inputs straight into GPT with no crawling. Basically provide an easy-to-use GPT API straight from this Actor.

I guess we have to add the dataset ID as input field, but what fields should we take from it as input? I suppose that we could make the input field from the dataset configurable as well? We can put both into a standalone “Dataset Input Settings” or something.

It will be pretty annoying to implement into the current "crawler" architecture ://

[jancurn]

Please let's not bake this into this Actor but create a new instead. Each Actor should do one thing and do it well - see https://github.com/apify/actor-whitepaper?tab=readme-ov-file#philosophy

We even have a full specs for an Actor to run GPT prompts against Dataset: https://docs.google.com/document/d/1wVCNDY5Sv8dLcpcP4JEnSbRxgNkX9IHjC9U4ACu1tGc/edit?tab=t.0

CC @tomasjindra

[Patai5]

Right, I see, good points 👍 Yeah, making it a standalone Actor would be better :)

[PavlinaVencovska]

@metalwarrior665 Sounds like out of the scope of our team, who do we delegate this on to make it happen?

[metalwarrior665]

I think the Video team can do it, not huge priority