frdel
Context repetition to the AI
I'm trying to understand why the context is enlarged every time a prompt is sent. The first prompt isn't enough for the AI to know how to respond. I ask this because my configuration with Olama, and AIs in general, have context limitations. In my case, after 40k, I'm already getting errors and can't continue, and this appears after a conversation of 5 or 6 prompts.
How can this be configured so that it doesn't send such long contexts repeatedly?
I don't fully understand the utility model either. Besides getting a title for each conversation, what other function does it have?
As far as I'm concerned, why use so many AIs if one could get everything, since currently AIs have browser mode, thought mode, etc. This greatly limits users' ability to meet the requirements for making it work, at least locally.
Thank you very much for your attention.
