Method _post_request

llms/openai_api.py:53–77

(self, kwargs: Dict[str, Any])

    @retry(stop=stop_after_attempt(10), wait=wait_random(min=5, max=10))
    def _post_request(self, kwargs: Dict[str, Any]) -> Tuple[OpenAIMessage, DebugInfo]:
        try:
            with self.get_client_args() as client_args:
                logger.debug(f"getting client args: {client_args}")
                client = openai.OpenAI(
                    api_key=client_args.get("api_key", None),
                    base_url=client_args.get("base_url", None),
                )
                completion = client.chat.completions.create(**kwargs)
                logger.debug(f"{completion}")
        except openai.RateLimitError as e:
            logger.warning(f"OpenAI rate limit error: {e}")
            raise e
        except openai.APIError as e:
            logger.error(f"OpenAI API error: {e}")
            raise e
        except Exception as e:
            logger.error(f"Unexpected error: {e}, type{e}, args{e.args}")
            logger.error(f"Traceback: {traceback.format_exc()}")
            raise e

        new_message = completion.choices[0].message
        return new_message, DebugInfo(
            info={"OpenAICompletion": completion.model_dump()}
        )