Added new v4.1.0 script. Fixed the thinking model issue and squished more nasty bugs.
Limitations:
Some thinking models like GLM and a couple Qwen models give blanks when they shouldn't. May investigate later, but I don't care about chinkope.Fixed with v4.1.0- Some models give blanks no matter what like Grok. That's on The Old LLM, not me.
- P8 Models still do not support streaming and Kimi K2 Thinking may take minutes to respond if it doesn't time out.
- Expect this to be rugpulled quickly if too many people use it.
Consult the README for usage info.
