kimi k2.5

[bitscrafts]

unable to find the proper place to this post

• any example setting kimi k2.5 to work alone in claude code?
  my attempt stops after first iteration (in test woorks ok)

• any example on setting a model via ollama??
  (provider) ollama-lnx, (http://10.10.10.217:11434/v1) (openai)
  moonshot-openai , (https://api.moonshot.ai/v1) (openai)
  (model) xxx-pro, ollama-lnx → qwen3-coder:30b (1)
  moonshot-openai → kimi-k2-thinking (2)
  (router) only default model, xxx-pro
  same situation stops on the first iteraction (when i say to analyse a codebase)
  anyway to check model/request (the model associated to a request)
  any help?

thanks

[elidickinson]

I'm not sure if this repo is still maintained, but I ran into the same thing and fixed it in my fork: https://github.com/elidickinson/claude-code-mux

The issue is that kimi likes to send the tool call before the text block but the code assumed the text block was always index 0. Fix is in 5a372fb

And FYI you can get very verbose output about requests and responses if set RUST_LOG=DEBUG when starting the server