Non-ASCII characters in tool outputs are sent as Unicode escapes.

[jusallcaz]

Low-parameter models are unable to parse these properly on their own.
I believe this can be resolved by setting ensure_ascii=False in json.dumps within json_api.py.

[yagil]

Thanks @jusallcaz! Would you want to take a stab at a PR?

[ncoghlan]

This fix has been published in https://github.com/lmstudio-ai/lmstudio-python/releases/tag/1.3.0