OpenAI Responses API vs Chat Completion API

 

LangSmith를 사용하려는데 langsmith_wrapper에서 openai가 대체 무슨 작업인가 깃허브를 찾아보니까 아래 코드가 있었다.

 

from __future__ import annotations

from typing import TYPE_CHECKING, Any

from langsmith.wrappers.base import ModuleWrapper

if TYPE_CHECKING:
    import openai


def __getattr__(name: str) -> Any:
    if name == "openai":
        try:
            import openai as openai_base
        except ImportError:
            raise ImportError(
                "OpenAI SDK is not installed. "
                "Please install it with `pip install openai`."
            )

        return ModuleWrapper(
            openai_base,
            llm_paths={
                "openai.api_resources.chat_completion.create",
                "openai.api_resources.chat_completion.acreate",
                "openai.api_resources.completion.create",
                "openai.api_resources.completion.acreate",
            },
        )
    raise AttributeError(f"module {__name__} has no attribute {name}")


__all__ = [
    "openai",  # noqa: F822
]

 

OpenAI SDK를 openai_base로,

그리고 llm_paths에 chat_completion과 completion을 보내면 된다.

 

 

그런데 Chat Completion과 Response의 차이를 알아보니 다음과 같다.

 

공식 문서에서 해당 내용을 찾아보니 

 

Why the Responses API?
The Responses API is our newest core API and an agentic API primitive

 

라고 한다.

 

즉 Chat_completion은 다소 옛날 API 같다. 

 

 

Capabilities	Chat Completions API	Responses API
Text generation	O	O
Audio	O	Coming soon
Vision	O	O
Structured Outputs	O	O
Function calling	O	O
Web search	O	O
File search	X	O
Computer use	X	O
Code interpreter	X	Coming soon

 

 

Chat Completion API

링크

 

request

curl https://api.openai.com/v1/chat/completions \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer $OPENAI_API_KEY" \
  -d '{
    "model": "gpt-4o",
    "messages": [
      {
        "role": "developer",
        "content": "You are a helpful assistant."
      },
      {
        "role": "user",
        "content": "Hello!"
      }
    ]
  }'

 

response

{
  "id": "chatcmpl-B9MBs8CjcvOU2jLn4n570S5qMJKcT",
  "object": "chat.completion",
  "created": 1741569952,
  "model": "gpt-4o-2024-08-06",
  "choices": [
    {
      "index": 0,
      "message": {
        "role": "assistant",
        "content": "Hello! How can I assist you today?",
        "refusal": null,
        "annotations": []
      },
      "logprobs": null,
      "finish_reason": "stop"
    }
  ],
  "usage": {
    "prompt_tokens": 19,
    "completion_tokens": 10,
    "total_tokens": 29,
    "prompt_tokens_details": {
      "cached_tokens": 0,
      "audio_tokens": 0
    },
    "completion_tokens_details": {
      "reasoning_tokens": 0,
      "audio_tokens": 0,
      "accepted_prediction_tokens": 0,
      "rejected_prediction_tokens": 0
    }
  },
  "service_tier": "default"
}

 

 

Streaming과 다르게 1회성 대화 요청 및 답변이다.

 

 

 

 

References:

https://platform.openai.com/docs/guides/responses-vs-chat-completions

https://github.com/langchain-ai/langsmith-wrappers/blob/main/langsmith/wrappers/openai.py