How do you handle tokens limit?

[jsifalda]

Hello, I am limited by size of the data the API accepts, as my notes are quite big... how do I go around it?
also, how do I change model to gtp-4?

thanks in advance

[irbull]

There is a settings for max number of generated tokens, but nothing for how many tokens to pass during the system prompt. If you have ideas there, let me know.

Feel free to file an issue to configure the model. Right now it's hard coded to 3.5-turbo.

[Cry0nicS]

Run into a similar issue.

One of the problems that I see is that I run a prompt like

Prompt: Write me a 2-3 paragraph summary of the my activity this week, in the first person. The work completed is below the '# Important Notes' section.

However, what is being passed to ChatGPT is not just the text under # Important Notes, but rather all the text in all 5 files from the current week.
And it's very likely that other sections (like meetings) have much more text that shouldn't go to chatGPT at all.

Now the question is: is it possible to only send the texts from a specific section?

Otherwise, I guess the only solution is to be able to choose ChatGPT 4 (assuming the user as a license)