[python] zimport - 수많은 python 패키지를 압축하여 관리 (github.com/waveware4ai)

python 패키지를 관리하다 보면 수많은 파일들이 부담스러울때 필요한 zimport 를 소개합니다.

소개

• zimport는 Python의 표준 zipimport를 대체하고 개선한 도구입니다.
• zimport는 zip-archives에서 Python 패키지를 로드하고 관리하는 데 사용됩니다. 즉, Java jar처럼 Python 패키지를 관리할 수 있습니다. 또한 동적 라이브러리(.dll, .pyd, .so) 로드도 지원합니다.
• 이 도구를 만들고 저의 python 작업 디렉토리는 약 160gb, 수백만개 파일에서 80G 1만 개 수준으로 줄어 들었습니다.

주요기능

• zip-archive에서 동적 라이브러리 로딩 지원(.pyd, .dll, .so, .dylib)
• zip-archive에서 내부 read() 시에 Java의 getresource처럼 내부 파일(예: 환경 파일) 읽기 지원
• 컴파일된 .pyc 파일 지원(name.cpython-version.pyc 및 pycache 폴더)

사용된기술

• importlib, meta_path, path_hooks
• function intercept (standard open, stat, read, ctypes.WinDLL, ctypes.CDLL 등)

사용환경

• python win/linux/macosx 지원
• python version 3.8~3.12 지원

사용방법

python -m pip install zimport
import zimport

• lib/site-package 디렉토리에서 패키지를 한꺼번에 압축하고, sys.path 에 추가하기만 하면 정상적으로 작동합니다.
• 물론, 시간의 여유가 되신다면 패키지별로 압축하여, java 의 jar 처럼 의존성에 따라서 패키지를 sys.path 에 추가해 주시면 됩니다.
• 또한 한번 압축된 package 는 share 하여 쓸수 있기 때문에 하드디스크 공간의 낭비를 줄여줄수 있다고 생각합니다.

마지막으로

• 소스는 github 에 공개하였으며 현재 버전은 0.1.4 입니다. 몇몇 package (예를들어 transformers) 아직 지원하지 안으며, torch, torchvision, numpy, pandas 와 같은 major 패키지는 이상없이 동작함을 확인하였습니다.

https://github.com/waveware4ai/zimport

• 또한, portable python 과 아주 궁합이 잘맞습니다. 이것도 github 에 업로드 하였습니다. linux 버전은 직접 컴파일하였고, windows 버전은 embeded 를 개작하였습니다.

https://github.com/waveware4ai/PortablePython

• 사용시 발생하는 버그나 문제점들은 리포팅해주시면 개선하도록 하겠습니다.

[python] 유튜브 자막 가져오기 

""" 유튜브 자막 가져오기 
    pip install youtube-transcript-api
    
    -- 최신 버전으로 업데이트
    pip install --upgrade youtube-transcript-api

    https://www.youtube.com/watch?v=XyljmT8dGA4
    
    자막있는 동영상 : https://www.youtube.com/watch?v=zRz9q8dPjC4
"""


from youtube_transcript_api import YouTubeTranscriptApi
# youtube_transcript_api._errors 에서 TooManyRequests를 제외하고 임포트합니다.
# TooManyRequests는 더 이상 직접 임포트할 수 없는 것으로 보입니다.
from youtube_transcript_api._errors import NoTranscriptFound, TranscriptsDisabled, VideoUnavailable

def get_youtube_transcript(video_id, languages=['ko', 'en']):
    """
    주어진 YouTube 동영상 ID와 언어 목록에 대해 자막을 가져옵니다.
    자동 생성 자막과 공식 자막을 모두 시도합니다.
    """
    try:
        # 우선 공식 자막을 시도
        transcript_list = YouTubeTranscriptApi.list_transcripts(video_id)

        # 사용 가능한 언어 목록에서 요청한 언어 중 하나를 찾아 가져옵니다.
        chosen_transcript = None
        for lang_code in languages:
            for transcript in transcript_list:
                if transcript.language_code == lang_code:
                    chosen_transcript = transcript
                    break
            if chosen_transcript:
                break

        if chosen_transcript:
            print(f"[{video_id}] {chosen_transcript.language} ({chosen_transcript.language_code}) 자막을 가져옵니다.")
            # fetch()는 이제 FetchedTranscript 객체를 반환하며, 이는 이터러블합니다.
            transcript_segments = chosen_transcript.fetch()
            return transcript_segments
        else:
            raise NoTranscriptFound(
                f"No suitable official transcript found for video {video_id} in languages {languages}.",
                video_id
            )

    except NoTranscriptFound:
        print(f"[{video_id}] 공식 자막을 찾을 수 없습니다. 자동 생성 자막을 시도합니다.")
        try:
            for lang_code in languages:
                try:
                    # get_transcript() 역시 이터러블한 객체를 반환하는 것으로 가정합니다.
                    transcript_segments = YouTubeTranscriptApi.get_transcript(video_id, languages=[lang_code], preserve_formatting=True)
                    print(f"[{video_id}] {lang_code} 자동 생성 자막을 가져왔습니다.")
                    return transcript_segments
                except NoTranscriptFound:
                    continue # 다음 언어로 시도
            print(f"[{video_id}] 요청된 언어 ({languages})로 자동 생성 자막도 찾을 수 없습니다.")
            return None # 적합한 자막을 찾지 못함
        except TranscriptsDisabled:
            print(f"[{video_id}] 이 동영상은 자막이 비활성화되어 있습니다.")
            return None
        except VideoUnavailable:
            print(f"[{video_id}] 동영상을 사용할 수 없거나 비공개/삭제되었습니다.")
            return None
        except Exception as e: # TooManyRequests를 포함한 모든 예외를 잡습니다.
            # 이 부분에서 TooManyRequests 에러를 포함하여 일반적인 오류를 처리합니다.
            print(f"[{video_id}] 자막을 가져오는 중 예기치 않은 오류가 발생했습니다: {e}")
            return None

    except TranscriptsDisabled:
        print(f"[{video_id}] 이 동영상은 자막이 비활성화되어 있습니다.")
        return None
    except VideoUnavailable:
        print(f"[{video_id}] 동영상을 사용할 수 없거나 비공개/삭제되었습니다.")
        return None
    except Exception as e: # TooManyRequests를 포함한 모든 예외를 잡습니다.
        # 이 부분에서 TooManyRequests 에러를 포함하여 일반적인 오류를 처리합니다.
        print(f"[{video_id}] 자막을 가져오는 중 예기치 않은 오류가 발생했습니다: {e}")
        return None


if __name__ == "__main__":
    # 예시 동영상 ID (실제 존재하는 동영상 ID로 변경해야 합니다)
    # 제가 추천해 드렸던 URL에서 ID를 추출했습니다.
    video_id_with_subtitle = "XyljmT8dGA4" # "모바일 유튜브 자동번역 한글자막 보는 방법"
    video_id_auto_caption = "XyljmT8dGA4" # 짧은 영상 (자동 생성 자막 가능성)
    #video_id_invalid = "invalid_video_id_123"
    video_id_invalid = "XyljmT8dGA4"

    print("--- 예제 1: 자막이 있는 동영상 ---")
    transcript_data = get_youtube_transcript(video_id_with_subtitle, languages=['ko', 'en'])
    if transcript_data:
        # segment.start, segment.duration, segment.text와 같이 속성으로 접근합니다.
        for i, segment in enumerate(transcript_data[:5]): # 슬라이싱은 여전히 가능해야 합니다.
            print(f"[{segment.start:.2f}-{segment.start + segment.duration:.2f}] {segment.text}")
        print(f"... (총 {len(list(transcript_data))}개 세그먼트)") # len()을 위해 list로 변환

        # 전체 자막 텍스트만 추출하고 싶다면:
        full_text = " ".join([segment.text for segment in transcript_data])
        print("\n--- 전체 자막 텍스트 (예제 1) ---")
        print(full_text[:500] + "...")
    else:
        print("자막을 가져오지 못했습니다.")

    print("\n--- 예제 2: 자동 생성 자막을 시도할 수 있는 동영상 ---")
    transcript_data_auto = get_youtube_transcript(video_id_auto_caption, languages=['en', 'ko'])
    if transcript_data_auto:
        for i, segment in enumerate(transcript_data_auto[:5]):
            print(f"[{segment.start:.2f}-{segment.start + segment.duration:.2f}] {segment.text}")
        print(f"... (총 {len(list(transcript_data_auto))}개 세그먼트)")
    else:
        print("자막을 가져오지 못했습니다.")

    print("\n--- 예제 3: 존재하지 않는 동영상 ID ---")
    transcript_data_invalid = get_youtube_transcript(video_id_invalid)
    if transcript_data_invalid:
        print("자막을 가져왔습니다.")
    else:
        print("자막을 가져오지 못했습니다.")

[PYTHON] Streamlit: 데이터 과학자의 쉬운 웹 제작 도구

vscode에서 streamlit 디버깅 하기

// conda env list
{
    "version": "0.2.0",
    "configurations": [
        {
            "name": "streamlit debug (p_ai env)",
            "type": "python",
            "request": "launch",
            "module": "streamlit",
            "args": ["run", "${file}"],
            "justMyCode": true,
            "python": "C:\\ProgramData\\anaconda3\\envs\\p_ai\\python.exe" // <--- 이 줄을 추가합니다. 
        }
    ]
}

https://github.com/ngio/py_ai/blob/main/README.md

[python] streamlit.io - A faster way to build and share data apps

https://www.youtube.com/watch?v=d7fnzDQ5qM8

https://streamlit.io/

Summary for experts

1. Set up your Python development environment.
2. Run:
   
   pip install streamlit
   
   
3. Validate the installation by running our Hello app:
   
   streamlit hello
   
   
4. Jump to our Basic concepts.

파이썬 생태계를 소개합니다.

https://wikidocs.net/226619

파이썬 생태계는 매우 방대하고 다양하며, 이를 통해 개발자들은 다양한 종류의 문제를 해결할 수 있는 도구에 접근할 수 있습니다. 이 생태계의 구체적인 특징은 다음과 같습니다:

1. 풍부한 라이브러리 선택: 파이썬은 데이터 과학, 웹 개발, 머신러닝, 네트워킹, 데이터베이스 관리, 그래픽 디자인, 게임 개발 등 거의 모든 프로그래밍 영역을 커버하는 수많은 라이브러리를 보유하고 있습니다. 이러한 다양성은 파이썬을 매우 다재다능한 언어로 만듭니다.
2. 특화된 도구들: 각각의 파이썬 라이브러리는 특정 작업 또는 문제 영역에 특화되어 있어, 개발자들은 필요에 맞는 최적의 도구를 선택할 수 있습니다. 예를 들어, NumPy는 수치 연산에, Pandas는 데이터 분석에, Matplotlib과 Seaborn은 데이터 시각화에, Scikit-learn은 머신러닝에 특화되어 있습니다.
3. 활발한 커뮤니티와 지원: 파이썬 라이브러리 대부분은 활발한 오픈 소스 커뮤니티에 의해 지원됩니다. 이 커뮤니티는 사용자들이 문제를 해결하고, 새로운 기능을 제안하며, 라이브러리를 개선하는 데 기여할 수 있는 환경을 제공합니다.
4. 지속적인 발전과 혁신: 파이썬 라이브러리는 지속적으로 업데이트되고 개선되어 새로운 기술 동향과 요구 사항을 반영합니다. 이는 파이썬을 최신 기술과 트렌드에 부합하는 유연한 언어로 유지시켜 줍니다.
5. 통합과 확장성: 많은 파이썬 라이브러리들은 서로 통합될 수 있어, 복잡한 작업을 수행하기 위해 여러 라이브러리를 함께 사용할 수 있습니다. 예를 들어, 데이터를 Pandas로 처리하고 Matplotlib 또는 Seaborn으로 시각화할 수 있습니다.

파이썬 생태계의 이러한 특징은 개발자들이 다양한 요구 사항에 맞는 최적의 솔루션을 개발할 수 있게 해주며, 파이썬의 인기와 활용도를 높이는 중요한 요소입니다.

파이썬 생태계가 얼마나 다양한지 알려주기 위해 최대한 한장에 하나의 모듈이 소개되도록 했습니다. 모듈 별 주요특징, 예제 코드 혹은 과련 시각화 이미지들을 포함하여 소개하였습니다. 예제 코드는 직접 돌려보기보다는 어떤 느낌으로 해당 모듈이 사용되는지 확인해보는 용도로 봐주시면 되겠습니다.

[PYTHON] OpenAI python API 라이브러리 - pip install openai

https://pypi.org/project/openai/

OpenAI Python 라이브러리는 Python 3.8 이상 애플리케이션에서 OpenAI REST API에 편리하게 액세스할 수 있도록 합니다. 이 라이브러리는 모든 요청 매개변수와 응답 필드에 대한 유형 정의를 포함하고 있으며, httpx 기반 동기 및 비동기 클라이언트를 모두 제공합니다 .

# install from PyPI
pip install openai

import openai
print(openai.__version__)

It is generated from our OpenAPI specification with Stainless.

Documentation

The REST API documentation can be found on platform.openai.com. The full API of this library can be found in api.md.

Installation

# install from PyPI
pip install openai

Usage

The full API of this library can be found in api.md.

The primary API for interacting with OpenAI models is the Responses API. You can generate text from the model with the code below.

import os
from openai import OpenAI

client = OpenAI(
    # This is the default and can be omitted
    api_key=os.environ.get("OPENAI_API_KEY"),
)

response = client.responses.create(
    model="gpt-4o",
    instructions="You are a coding assistant that talks like a pirate.",
    input="How do I check if a Python object is an instance of a class?",
)

print(response.output_text)

The previous standard (supported indefinitely) for generating text is the Chat Completions API. You can use that API to generate text from the model with the code below.

from openai import OpenAI

client = OpenAI()

completion = client.chat.completions.create(
    model="gpt-4o",
    messages=[
        {"role": "developer", "content": "Talk like a pirate."},
        {
            "role": "user",
            "content": "How do I check if a Python object is an instance of a class?",
        },
    ],
)

print(completion.choices[0].message.content)

While you can provide an api_key keyword argument, we recommend using python-dotenv to add OPENAI_API_KEY="My API Key" to your .env file so that your API key is not stored in source control. Get an API key here.

Vision

With an image URL:

prompt = "What is in this image?"
img_url = "https://upload.wikimedia.org/wikipedia/commons/thumb/d/d5/2023_06_08_Raccoon1.jpg/1599px-2023_06_08_Raccoon1.jpg"

response = client.responses.create(
    model="gpt-4o-mini",
    input=[
        {
            "role": "user",
            "content": [
                {"type": "input_text", "text": prompt},
                {"type": "input_image", "image_url": f"{img_url}"},
            ],
        }
    ],
)

With the image as a base64 encoded string:

import base64
from openai import OpenAI

client = OpenAI()

prompt = "What is in this image?"
with open("path/to/image.png", "rb") as image_file:
    b64_image = base64.b64encode(image_file.read()).decode("utf-8")

response = client.responses.create(
    model="gpt-4o-mini",
    input=[
        {
            "role": "user",
            "content": [
                {"type": "input_text", "text": prompt},
                {"type": "input_image", "image_url": f"data:image/png;base64,{b64_image}"},
            ],
        }
    ],
)

Async usage

Simply import AsyncOpenAI instead of OpenAI and use await with each API call:

import os
import asyncio
from openai import AsyncOpenAI

client = AsyncOpenAI(
    # This is the default and can be omitted
    api_key=os.environ.get("OPENAI_API_KEY"),
)


async def main() -> None:
    response = await client.responses.create(
        model="gpt-4o", input="Explain disestablishmentarianism to a smart five year old."
    )
    print(response.output_text)


asyncio.run(main())

Functionality between the synchronous and asynchronous clients is otherwise identical.

Streaming responses

We provide support for streaming responses using Server Side Events (SSE).

from openai import OpenAI

client = OpenAI()

stream = client.responses.create(
    model="gpt-4o",
    input="Write a one-sentence bedtime story about a unicorn.",
    stream=True,
)

for event in stream:
    print(event)

The async client uses the exact same interface.

import asyncio
from openai import AsyncOpenAI

client = AsyncOpenAI()


async def main():
    stream = await client.responses.create(
        model="gpt-4o",
        input="Write a one-sentence bedtime story about a unicorn.",
        stream=True,
    )

    async for event in stream:
        print(event)


asyncio.run(main())

Realtime API beta

The Realtime API enables you to build low-latency, multi-modal conversational experiences. It currently supports text and audio as both input and output, as well as function calling through a WebSocket connection.

Under the hood the SDK uses the websockets library to manage connections.

The Realtime API works through a combination of client-sent events and server-sent events. Clients can send events to do things like update session configuration or send text and audio inputs. Server events confirm when audio responses have completed, or when a text response from the model has been received. A full event reference can be found here and a guide can be found here.

Basic text based example:

import asyncio
from openai import AsyncOpenAI

async def main():
    client = AsyncOpenAI()

    async with client.beta.realtime.connect(model="gpt-4o-realtime-preview") as connection:
        await connection.session.update(session={'modalities': ['text']})

        await connection.conversation.item.create(
            item={
                "type": "message",
                "role": "user",
                "content": [{"type": "input_text", "text": "Say hello!"}],
            }
        )
        await connection.response.create()

        async for event in connection:
            if event.type == 'response.text.delta':
                print(event.delta, flush=True, end="")

            elif event.type == 'response.text.done':
                print()

            elif event.type == "response.done":
                break

asyncio.run(main())

However the real magic of the Realtime API is handling audio inputs / outputs, see this example TUI script for a fully fledged example.

Realtime error handling

Whenever an error occurs, the Realtime API will send an error event and the connection will stay open and remain usable. This means you need to handle it yourself, as no errors are raised directly by the SDK when an error event comes in.

client = AsyncOpenAI()

async with client.beta.realtime.connect(model="gpt-4o-realtime-preview") as connection:
    ...
    async for event in connection:
        if event.type == 'error':
            print(event.error.type)
            print(event.error.code)
            print(event.error.event_id)
            print(event.error.message)

Using types

Nested request parameters are TypedDicts. Responses are Pydantic models which also provide helper methods for things like:

• Serializing back into JSON, model.to_json()
• Converting to a dictionary, model.to_dict()

Typed requests and responses provide autocomplete and documentation within your editor. If you would like to see type errors in VS Code to help catch bugs earlier, set python.analysis.typeCheckingMode to basic.

Pagination

List methods in the OpenAI API are paginated.

This library provides auto-paginating iterators with each list response, so you do not have to request successive pages manually:

from openai import OpenAI

client = OpenAI()

all_jobs = []
# Automatically fetches more pages as needed.
for job in client.fine_tuning.jobs.list(
    limit=20,
):
    # Do something with job here
    all_jobs.append(job)
print(all_jobs)

Or, asynchronously:

import asyncio
from openai import AsyncOpenAI

client = AsyncOpenAI()


async def main() -> None:
    all_jobs = []
    # Iterate through items across all pages, issuing requests as needed.
    async for job in client.fine_tuning.jobs.list(
        limit=20,
    ):
        all_jobs.append(job)
    print(all_jobs)


asyncio.run(main())

Alternatively, you can use the .has_next_page(), .next_page_info(), or .get_next_page() methods for more granular control working with pages:

first_page = await client.fine_tuning.jobs.list(
    limit=20,
)
if first_page.has_next_page():
    print(f"will fetch next page using these details: {first_page.next_page_info()}")
    next_page = await first_page.get_next_page()
    print(f"number of items we just fetched: {len(next_page.data)}")

# Remove `await` for non-async usage.

Or just work directly with the returned data:

first_page = await client.fine_tuning.jobs.list(
    limit=20,
)

print(f"next page cursor: {first_page.after}")  # => "next page cursor: ..."
for job in first_page.data:
    print(job.id)

# Remove `await` for non-async usage.

Nested params

Nested parameters are dictionaries, typed using TypedDict, for example:

from openai import OpenAI

client = OpenAI()

response = client.chat.responses.create(
    input=[
        {
            "role": "user",
            "content": "How much ?",
        }
    ],
    model="gpt-4o",
    response_format={"type": "json_object"},
)

File uploads

Request parameters that correspond to file uploads can be passed as bytes, or a PathLike instance or a tuple of (filename, contents, media type).

from pathlib import Path
from openai import OpenAI

client = OpenAI()

client.files.create(
    file=Path("input.jsonl"),
    purpose="fine-tune",
)

The async client uses the exact same interface. If you pass a PathLike instance, the file contents will be read asynchronously automatically.

Handling errors

When the library is unable to connect to the API (for example, due to network connection problems or a timeout), a subclass of openai.APIConnectionError is raised.

When the API returns a non-success status code (that is, 4xx or 5xx response), a subclass of openai.APIStatusError is raised, containing status_code and response properties.

All errors inherit from openai.APIError.

import openai
from openai import OpenAI

client = OpenAI()

try:
    client.fine_tuning.jobs.create(
        model="gpt-4o",
        training_file="file-abc123",
    )
except openai.APIConnectionError as e:
    print("The server could not be reached")
    print(e.__cause__)  # an underlying Exception, likely raised within httpx.
except openai.RateLimitError as e:
    print("A 429 status code was received; we should back off a bit.")
except openai.APIStatusError as e:
    print("Another non-200-range status code was received")
    print(e.status_code)
    print(e.response)

Error codes are as follows:

Status CodeError Type

Request IDs

For more information on debugging requests, see these docs

All object responses in the SDK provide a _request_id property which is added from the x-request-id response header so that you can quickly log failing requests and report them back to OpenAI.

response = await client.responses.create(
    model="gpt-4o-mini",
    input="Say 'this is a test'.",
)
print(response._request_id)  # req_123

Note that unlike other properties that use an _ prefix, the _request_id property is public. Unless documented otherwise, all other _ prefix properties, methods and modules are private.

[!IMPORTANT]
If you need to access request IDs for failed requests you must catch the APIStatusError exception

import openai

try:
    completion = await client.chat.completions.create(
        messages=[{"role": "user", "content": "Say this is a test"}], model="gpt-4"
    )
except openai.APIStatusError as exc:
    print(exc.request_id)  # req_123
    raise exc

Retries

Certain errors are automatically retried 2 times by default, with a short exponential backoff. Connection errors (for example, due to a network connectivity problem), 408 Request Timeout, 409 Conflict, 429 Rate Limit, and >=500 Internal errors are all retried by default.

You can use the max_retries option to configure or disable retry settings:

from openai import OpenAI

# Configure the default for all requests:
client = OpenAI(
    # default is 2
    max_retries=0,
)

# Or, configure per-request:
client.with_options(max_retries=5).chat.completions.create(
    messages=[
        {
            "role": "user",
            "content": "How can I get the name of the current day in JavaScript?",
        }
    ],
    model="gpt-4o",
)

Timeouts

By default requests time out after 10 minutes. You can configure this with a timeout option, which accepts a float or an httpx.Timeout object:

from openai import OpenAI

# Configure the default for all requests:
client = OpenAI(
    # 20 seconds (default is 10 minutes)
    timeout=20.0,
)

# More granular control:
client = OpenAI(
    timeout=httpx.Timeout(60.0, read=5.0, write=10.0, connect=2.0),
)

# Override per-request:
client.with_options(timeout=5.0).chat.completions.create(
    messages=[
        {
            "role": "user",
            "content": "How can I list all files in a directory using Python?",
        }
    ],
    model="gpt-4o",
)

On timeout, an APITimeoutError is thrown.

Note that requests that time out are retried twice by default.

Advanced

Logging

We use the standard library logging module.

You can enable logging by setting the environment variable OPENAI_LOG to info.

$ export OPENAI_LOG=info

Or to debug for more verbose logging.

How to tell whether None means null or missing

In an API response, a field may be explicitly null, or missing entirely; in either case, its value is None in this library. You can differentiate the two cases with .model_fields_set:

if response.my_field is None:
  if 'my_field' not in response.model_fields_set:
    print('Got json like {}, without a "my_field" key present at all.')
  else:
    print('Got json like {"my_field": null}.')

Accessing raw response data (e.g. headers)

The "raw" Response object can be accessed by prefixing .with_raw_response. to any HTTP method call, e.g.,

from openai import OpenAI

client = OpenAI()
response = client.chat.completions.with_raw_response.create(
    messages=[{
        "role": "user",
        "content": "Say this is a test",
    }],
    model="gpt-4o",
)
print(response.headers.get('X-My-Header'))

completion = response.parse()  # get the object that `chat.completions.create()` would have returned
print(completion)

These methods return a LegacyAPIResponse object. This is a legacy class as we're changing it slightly in the next major version.

For the sync client this will mostly be the same with the exception of content & text will be methods instead of properties. In the async client, all methods will be async.

A migration script will be provided & the migration in general should be smooth.

.with_streaming_response

The above interface eagerly reads the full response body when you make the request, which may not always be what you want.

To stream the response body, use .with_streaming_response instead, which requires a context manager and only reads the response body once you call .read(), .text(), .json(), .iter_bytes(), .iter_text(), .iter_lines() or .parse(). In the async client, these are async methods.

As such, .with_streaming_response methods return a different APIResponse object, and the async client returns an AsyncAPIResponse object.

with client.chat.completions.with_streaming_response.create(
    messages=[
        {
            "role": "user",
            "content": "Say this is a test",
        }
    ],
    model="gpt-4o",
) as response:
    print(response.headers.get("X-My-Header"))

    for line in response.iter_lines():
        print(line)

The context manager is required so that the response will reliably be closed.

Making custom/undocumented requests

This library is typed for convenient access to the documented API.

If you need to access undocumented endpoints, params, or response properties, the library can still be used.

Undocumented endpoints

To make requests to undocumented endpoints, you can make requests using client.get, client.post, and other http verbs. Options on the client will be respected (such as retries) when making this request.

import httpx

response = client.post(
    "/foo",
    cast_to=httpx.Response,
    body={"my_param": True},
)

print(response.headers.get("x-foo"))

Undocumented request params

If you want to explicitly send an extra param, you can do so with the extra_query, extra_body, and extra_headers request options.

Undocumented response properties

To access undocumented response properties, you can access the extra fields like response.unknown_prop. You can also get all the extra fields on the Pydantic model as a dict with response.model_extra.

Configuring the HTTP client

You can directly override the httpx client to customize it for your use case, including:

• Support for proxies
• Custom transports
• Additional advanced functionality

import httpx
from openai import OpenAI, DefaultHttpxClient

client = OpenAI(
    # Or use the `OPENAI_BASE_URL` env var
    base_url="http://my.test.server.example.com:8083/v1",
    http_client=DefaultHttpxClient(
        proxy="http://my.test.proxy.example.com",
        transport=httpx.HTTPTransport(local_address="0.0.0.0"),
    ),
)

You can also customize the client on a per-request basis by using with_options():

client.with_options(http_client=DefaultHttpxClient(...))

Managing HTTP resources

By default the library closes underlying HTTP connections whenever the client is garbage collected. You can manually close the client using the .close() method if desired, or with a context manager that closes when exiting.

from openai import OpenAI

with OpenAI() as client:
  # make requests here
  ...

# HTTP client is now closed

Microsoft Azure OpenAI

To use this library with Azure OpenAI, use the AzureOpenAI class instead of the OpenAI class.

[!IMPORTANT] The Azure API shape differs from the core API shape which means that the static types for responses / params won't always be correct.

from openai import AzureOpenAI

# gets the API Key from environment variable AZURE_OPENAI_API_KEY
client = AzureOpenAI(
    # https://learn.microsoft.com/azure/ai-services/openai/reference#rest-api-versioning
    api_version="2023-07-01-preview",
    # https://learn.microsoft.com/azure/cognitive-services/openai/how-to/create-resource?pivots=web-portal#create-a-resource
    azure_endpoint="https://example-endpoint.openai.azure.com",
)

completion = client.chat.completions.create(
    model="deployment-name",  # e.g. gpt-35-instant
    messages=[
        {
            "role": "user",
            "content": "How do I output all files in a directory using Python?",
        },
    ],
)
print(completion.to_json())

In addition to the options provided in the base OpenAI client, the following options are provided:

• azure_endpoint (or the AZURE_OPENAI_ENDPOINT environment variable)
• azure_deployment
• api_version (or the OPENAI_API_VERSION environment variable)
• azure_ad_token (or the AZURE_OPENAI_AD_TOKEN environment variable)
• azure_ad_token_provider

An example of using the client with Microsoft Entra ID (formerly known as Azure Active Directory) can be found here.

Versioning

This package generally follows SemVer conventions, though certain backwards-incompatible changes may be released as minor versions:

1. Changes that only affect static types, without breaking runtime behavior.
2. Changes to library internals which are technically public but not intended or documented for external use. (Please open a GitHub issue to let us know if you are relying on such internals.)
3. Changes that we do not expect to impact the vast majority of users in practice.

We take backwards-compatibility seriously and work hard to ensure you can rely on a smooth upgrade experience.

We are keen for your feedback; please open an issue with questions, bugs, or suggestions.