OpenAI Text Generation API and Bevy Async Tasks

[@kristupasT]

Great video, even though I'm not interested in using OpenAI in bevy, I still learned about async tasks. Keep up the good work.

[@logicprojects]

Thank you! That was a big part of my goals with this video

[@davidlampach1278]

This video rocks. Exactly the kind of documentation for async we need.

[@logicprojects]

Thank you!

[@alexandrecamilleri8640]

Yes my exact though. I have nothing to do with AI but I had a case where I need async usage with Bevy, was wrecking my head for a few days, this solved it. Thank you so much

[@jan4946]

Great idea! I've gotten better results slightly reframing the text: ``` A hungry dwarf named Marc is living in downton. The subway is to your south through a blocked road and the bar is to your left and it's open. A human approaches. The human says to Marc: What's in the south? What does Marc answer? > Marc would answer: "The subway is located to the south, but the road is blocked." ``` But `Where are we?` yields `Marc would answer: You and the human are in downtown` :D

[@jan4946]

okay, this get's much better when you add `Marc enjoys telling jokes` to the statement :D Adding a description of the humans mood and appearance can trigger some more detailed answers as well.

[@logicprojects]

Neat! I had a feeling there were better ways to get good results and I'm glad to see you're experimenting with it!

[@HalfMonty11]

Very cool idea. Would probably take some big triple A title game to implement something like this in a production build. However, since the output is just text which stores quite easily, you could use open AI to generate hundreds of variations of responses for each "personality" in a game for a controlled number of potential input prompts allowing it to appear to the user that they are getting a unique response every time they play the game which could really help with world building.

[@logicprojects]

That sounds cool and could take a lot of work out of doing something like that. Getting raw input from a user probably won't work but I could see a very well structured prompt being used to turn raw gameplay stats into a paragraph description. It might need to wait for more consistent models though

[@andoliu246]

Thanks for the sharing, I try to compile to wasm and I got this error. 45 | commands.spawn(OpenAiTask(task)); | ---------- ^^^^ expected `Task`, found `FakeTask` | | | arguments to this struct are incorrect

[@RSchenal]

Wow, grate idea! (video is great too btw)

[@CuriousSpy]

checkout poll-promise

[@logicprojects]

I see it's readme but what would that get me in this example? It's unclear if this would even play nice with bevy as the main parts of the library seem to be nonsync

[@bibliusz777]

da vinky?

[@logicprojects]

If I actually mispronounced that... Well I guess that's the life of publishing things I can't easily fix