Petr Mironychev
Petr Mironychev
Box2d under MIT license now
I don't think so, same functional I saw in bacon2D framework and it was separate from qml_box2d. Felgo possible took from bacon2d in the past.
For me Qwen3 is ambiguous, in ollama it does not listen at all to what is written in the system promt, in lm studio it seems better. I can remove...
Sorry but you didn't understand me. Qwen3 model don't count system prompt for her, at all. For example as you can see, I sent no_think and ask return code as...
It was, I just disabled text handler for show raw answer from model. It is controlled from here  This answer from qwen-coder2.5 
Thank you, this is good idea, I want to add this like a users saved presets Provider+Url+Model+Template
So, looks you are using FIM here via llama.cpp? If you will run like Instruct model for code completion in QodeAssist you can fix this just with correct system prompt,...
I meant not about error, but about how FIM models works. You will have bigger output answer and pay for that bigger counting of tokens. In case with instruct models...
Okay, merging code with suggest and show is not simple in QtC like I thought. So now I don't have idea for that
I improved code suggestion and merging with existing code in 0.5.10. Please try, should be better. It is not absolutely what did you want. Maybe need new algorithm for that....