Comment by wuschel

Hi,

You mentioned switching out LLM backends. Would you think that lesser models e.g. one of the recent Qwen variants would work for your use case?