pub fn process_openai_compat_response( raw_content: Option<&str>, finish_reason: &str, prompt_tokens: u64, completion_tokens: u64, model: &str, backend: &str, ) -> LlmResult