Hey there, there's a "system" message which is basically instructions on how the bot should behave, sort of an initial prompt, for example "you're a fitness guide that knows about working out and nutrition". now I wanted to ask if this system message counts as tokens when it comes to the api's pricing and if it does, does it count only once in the beginning of the conversation or does it count every single message?
#"System" message regarding tokens for pricing
latent falcon
haughty arch
Hey there, there's a "system" message which is basically instructions on how the...
Counts for every message as it has to be sent with each one.
(Well if you want it to stay on track)
latent falcon
Counts for every message as it has to be sent with each one.
damn that's a shame. do you think its worth it? if you have experience with it
is it manageable for the price?
haughty arch
damn that's a shame. do you think its worth it? if you have experience with it
i...
Is what manageable? Using a system prompt? Both yes and necessary if you want the bot to do what you want it to. If it's to much and you can't or don't want to write your own application you can always use chat.openai.com for free.
latent falcon
Is what manageable? Using a system prompt? Both yes and necessary if you want th...
okay I understand, I do want to create my own app utilising gpt's api, it's just a shame that it's not a thing you can set up once a conversation for the personality instead of having to do it every single message. In case, thanks!
haughty arch
Any consolation, it is kinda a setup once.... cuz usually you put it in a loop....
latent falcon
Any consolation, it is kinda a setup once.... cuz usually you put it in a loop.....
yeah coding-wise it isn't a problem but it kinda sucks that you need to use your tokens in every single message
west delta
the model needs to process it
takes resources
how else can you generate text based on previous text 😄
its not really any different from the usual user/assistant text, just puts more attention to system
about the price, 3.5 is really cheap (4 is very expensive compared)
latent falcon
the model needs to process it
takes resources
how else can you generate text ba...
I wish there was a way to train the model in advance, guessing it is an option but comes out more expensive any way
and the price, as everything else, is relative haha 😄
true ruin