Could you add support for llama.cpp?

[tdzard94]

I’m using llama.cpp to run the Qwen3.6-27B model on my machine with a relatively low context size, and I access it over my LAN because running it through Ollama or LM Studio is too demanding for my system. I noticed that OpenClaude can’t retrieve context usage information from llama.cpp, so I never really know when the context window is getting full and needs compaction. Would it be possible to add this feature?