pub const AI_LLM_RETRY_MIN_DELAY_MS: u64 = 100;
Minimum retry delay for LLM API calls in milliseconds (floor for exponential backoff).