Users have expressed interest in using hosted or proprietary models (e.g., OpenAI, Gemini) rather than running a local model. The goal is to make the system flexible while keeping the local LLM option ...