Comment by dmix
7 months ago
Openai api lets you cache the beginning parts of prompts already to save time/money so it's not parsing the same instructions repeatedly, not very different here.
7 months ago
Openai api lets you cache the beginning parts of prompts already to save time/money so it's not parsing the same instructions repeatedly, not very different here.
There is "performance" as in "speed and cost" and performance as in "the model returning quality responses, without getting lost in the weeds". Caching only helps with the former.
If the context window is small enough then only the tail of the prompt matters anyways.
"the model returning quality responses, without getting lost in the weeds"
I should edit, but that would be disingenuous. This is exactly what I meant.
thank you!