Assistant File analysis feature only scans the first 1.000 characters of a text file

[burnt fossil]

I'm building an assistant whose task is to read a transcript of a meeting (txt file) and do different tasks with this content
The probem is that it almost always only take the 1 000 (or 500 or 1500..) first characters of this transcript
I've tried to change my prompt in many ways but it keeps reading only a part of the document
Is there a setting somewhere that can fix this ? (same prompt & same files work perfectly when I use a chatgpt Agent on chatgpt interface

Assistant File analysis feature only scans the first 1.000 characters of a text file

[rotund yacht]

what do you mean with "only scans the first 1.000 characters" ... do you want to use mistral-ocr technology?

[burnt fossil]

thanks for your answer

No, I'm building an assistant with gpt-4o model and his job is to read and summarize the content of a txt file
The file contents approximately 18.000 characters, but when I tell my agent to open the file and analyze itts content, il limits itself to reading the first 1.000 characters of the file (sometimes 500, sometimes 1500, rarely the whole document)

[rotund yacht]

you need to be more precise .... are you trying to use the Assisants API or the Agents framework of LC?

 but when I tell my agent to open the file and analyze itts content, il limits itself to reading the first 1.000 characters of the file (sometimes 500, sometimes 1500, rarely the whole document)

and how do you tell your "Assistant" to do this? sharing a screenshot would help here

[burnt fossil]

I'm using the assistant construction feature in Librechat
Herre is a screenshot

The python code clearly indicates the limited number of characters read, while my prompt (in french, but same results in english) ask for a complete reading of the file content

So i wonder if there is somewhere a settings that sets this kind of limit

[rotund yacht]

I still miss context. "assistants constructions" = you call the openAI / Azure Assistant API ?