API assistant is using over 15k tokens

API
,
1

we are building a basic answer system via gpt 4o mini and for some reason every command run uses over 15k tokens. when we remove instructions/files for the ai it just increases the token usage??

if somebody could help us it would be greatly appriciated, Thanks in advance
Dericom.

1 Like
2

Welcome to the forum @Dericom

Does this increase in usage happens in new thread or the existing thread?

1 Like
3

in new threads as well as in used threads

4

How long is the instructions param for your assistant?

5

exactly 422 token in tokenizer

6

Do you have any functions defined on the assistant?

Also are the 15k tokens for in or out?

A user interface displays the "Playground" section with a token count of 1072, including 1023 input tokens and 49 output tokens, along with navigation links to Dashboard, Docs, API reference, and various icons. (Captioned by AI)
A user interface displays the "Playground" section with a token count of 1072, including 1023 input tokens and 49 output tokens, along with navigation links to Dashboard, Docs, API reference, and various icons. (Captioned by AI)1246×380 26.8 KB

1 Like
7

there are no functions to the assistant

the tokens are In’put’

image
image954×537 110 KB

1 Like
8

I can see that you have the file_search tool enabled along with an attached vector store which would explain the 15k input tokens.

1 Like
9

Thank you for the help, we’ll have to deal with it it seems,
but they should make it more obvious for the developers/consumers that the tools could consume this much tokens in one go.
Yours Faithfully, Dericom

1 Like