okuvshynov / vimqq

local LLM chat for vim with dynamic warmup, forks and multi-model support
MIT License
3 stars 0 forks source link

prompt master task #2

Open okuvshynov opened 2 months ago

okuvshynov commented 2 months ago
okuvshynov commented 2 months ago

Let's use configurable number of steps for reasoning chain. Let's also be able to 'hide' this number and decide randomly in advance. This will allow us to have ~blind test for 'how much number of reasoning steps affects quality' if I consistently mark answers. This is probably only possible with groq which is fast enough to not notice the difference in time. We might need to add extra sleep