@netikras a bit hard in bash maybe, but could we give the gpt bot the context of the conversation so it knows what it previously answered? In that case we can do simple follow up questions and you can ask things like "Who is right? Is it correct what x says? Is y considered reasonable? Are you sure? Give other options.". So making it more part of the conversation :)

Ofc I could. But in long threads more text = more tokens = more $...

And not necessarily wanted in all cases

@netikras ah yes, won't complain about your generous donation. Do you use gpt4o or gpt4o mini? I'm using gpt4o-mini because there is for my usage nearly no difference. It's way cheaper. And i do quite mass batching. My default bill is ten cent a day because I have one gb of data stored. I still have to figure out if much data made my bot slower or not, or that OpenAI was just having a bad day. The OpenAI api has often bad days tho. When I use my talking assistant, i do notice some delay more very well.

@retoor I think it's 4o. It's paid by my friend's company, I don't want to exploit him :)

@netikras ah ok, but don't worry too much in general :P gpt is very cheap, I think it's really made for batch processing shit. You can talk a whole day (way more than devRant produces) for a while cent. The makes of GPT WANT to break the internet with their content.

I looked into this once but it became tricky quickly 😅

@ScriptCoded you mean such functionality? With python it's very easy. There are two ways with OpenAI, you keep the message history and send it every time with classic completion or they do it for you but you have to use the agent api.

@retoor yeah I tried modifying the bash script. But iirc there was some trouble with the token limit, forcing me to make some cumbersome whatchamacallits. But that was a year or two ago