Claude Code opened to any LLM — OpenAI, Gemini, DeepSeek, Ollama, and 200+ models via OpenAI-compatible API shim
// Shared frame interval for render throttling and animations (~60fps) export const FRAME_INTERVAL_MS = 16