describe the bug
Just updated to latest enterprise version 2.2.223 and tried to optimise a pipe it stopped approx. 60 secs in.. I followed up with: "you stopped showing your output are you still running in the background?"

only to get: Error: 400 Prompt too long: 41066 tokens exceeds max context window of 32768 tokens

This is a bit interesting considering that:

1. I only just started screenpipe and unless there has been some calculation issues in tokens in screenpipe i dont believe there can be that many tokens in a single pipe optimse... however that said we did just move to oMLX (https://omlx.ai/) as our ai inference and they do have an interesting way to optimze cache for speed and agent use case.... So I am thinking maybe you are calculating tokens from the server end? If that is the case i wouldv'e thought that would be a bit weird. however I thought I would raise it here anyways.
2. our context size for openai connection is 120k (GPT5 size) so unsure why 32768 is now the limit? - I'll check server end for max context if it's being fed back to screenpipe and assigning that.

logs here: https://screenpi.pe/logs/121c3437-7458-4b07-9234-0b4fe956b308

system info

• os: Windows 11
• screenpipe version: v2.2.223

additional context