Migración a la biblioteca de la API de OpenAI Python 1.x

Artículo
09/26/2024

OpenAI ha publicado una nueva versión de la Biblioteca de la API de OpenAI para Python. Esta guía es complementaria a la guía de migración de OpenAI y le ayudará a acelerar los cambios específicos de Azure OpenAI.

Actualizaciones

Se trata de una versión completamente nueva de la biblioteca de API de OpenAI Python.
A partir del 6 de noviembre de 2023, pip install openai y pip install openai --upgrade instalarán version 1.x de la biblioteca de OpenAI Python.
La actualización de version 0.28.1 a version 1.x es un cambio importante, tendrá que probar y actualizar el código.
Reintento automático con retroceso si se produce un error
Tipos adecuados (para mypy/pyright/editors)
Ahora puede crear una instancia de cliente, en lugar de usar un valor predeterminado global.
Cambiar a la creación explícita de instancias de cliente
Cambios de nombre

Problemas conocidos

DALL-E3 es totalmente compatible con la versión 1.x más reciente. DALL-E2 se puede usar con 1.x realizando las siguientes modificaciones en el código.
embeddings_utils.py, que se usaba para proporcionar funcionalidades como la similitud de coseno para la búsqueda de texto semántico, ya no forma parte de la biblioteca de la API de OpenAI Python.
También debe comprobar los problemas de GitHub activos para la biblioteca de OpenAI Python.

Probar antes de migrar

Importante

La migración automática del código mediante openai migrate no se admite con Azure OpenAI.

Como se trata de una nueva versión de la biblioteca con cambios importantes, debe probar el código ampliamente en la nueva versión antes de migrar las aplicaciones de producción para que se basen en la versión 1.x. También debe revisar el código y los procesos internos para asegurarse de seguir los procedimientos recomendados y anclar el código de producción solo a las versiones que haya probado por completo.

Para facilitar el proceso de migración, estamos actualizando los ejemplos de código existentes en nuestros documentos para Python a una experiencia con pestañas:

OpenAI Python 1.x
OpenAI Python 0.28.1

pip install openai --upgrade

pip install openai==0.28.1

Esto proporciona contexto para lo que ha cambiado y le permite probar la nueva biblioteca en paralelo mientras se continúa proporcionando compatibilidad con la versión 0.28.1. Si actualiza a 1.x y se da cuenta de que necesita volver temporalmente a la versión anterior, siempre puede volver a pip uninstall openai e instalar como destino 0.28.1 con pip install openai==0.28.1.

Deberá establecer la variable model en el nombre de implementación que eligió al implementar los modelos GPT-3.5-Turbo o GPT-4. Al escribir el nombre del modelo se producirá un error a menos que elija un nombre de implementación idéntico al nombre del modelo subyacente.

import os
from openai import AzureOpenAI

client = AzureOpenAI(
  azure_endpoint = os.getenv("AZURE_OPENAI_ENDPOINT"), 
  api_key=os.getenv("AZURE_OPENAI_API_KEY"),  
  api_version="2024-02-01"
)

response = client.chat.completions.create(
    model="gpt-35-turbo", # model = "deployment_name"
    messages=[
        {"role": "system", "content": "You are a helpful assistant."},
        {"role": "user", "content": "Does Azure OpenAI support customer managed keys?"},
        {"role": "assistant", "content": "Yes, customer managed keys are supported by Azure OpenAI."},
        {"role": "user", "content": "Do other Azure AI services support this too?"}
    ]
)

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

Puede encontrar ejemplos adicionales en nuestro artículo de finalización detallada del chat.

Deberá establecer la variable engine en el nombre de implementación que eligió al implementar los modelos GPT-3.5-Turbo o GPT-4. Si escribe el nombre del modelo, se producirá un error a menos que elija un nombre de implementación idéntico al nombre del modelo subyacente.

import os
import openai
openai.api_type = "azure"
openai.api_base = os.getenv("AZURE_OPENAI_ENDPOINT") 
openai.api_key = os.getenv("AZURE_OPENAI_API_KEY")
openai.api_version = "2024-02-01"

response = openai.ChatCompletion.create(
    engine="gpt-35-turbo", # engine = "deployment_name".
    messages=[
        {"role": "system", "content": "You are a helpful assistant."},
        {"role": "user", "content": "Does Azure OpenAI support customer managed keys?"},
        {"role": "assistant", "content": "Yes, customer managed keys are supported by Azure OpenAI."},
        {"role": "user", "content": "Do other Azure AI services support this too?"}
    ]
)

print(response)
print(response['choices'][0]['message']['content'])

Finalizaciones

OpenAI Python 1.x
OpenAI Python 0.28.1

import os
from openai import AzureOpenAI
    
client = AzureOpenAI(
    api_key=os.getenv("AZURE_OPENAI_API_KEY"),  
    api_version="2024-02-01",
    azure_endpoint = os.getenv("AZURE_OPENAI_ENDPOINT")
)
    
deployment_name='REPLACE_WITH_YOUR_DEPLOYMENT_NAME' #This will correspond to the custom name you chose for your deployment when you deployed a model. 
    
# Send a completion call to generate an answer
print('Sending a test completion job')
start_phrase = 'Write a tagline for an ice cream shop. '
response = client.completions.create(model=deployment_name, prompt=start_phrase, max_tokens=10) # model = "deployment_name"
print(response.choices[0].text)

import os
import openai

openai.api_key = os.getenv("AZURE_OPENAI_API_KEY")
openai.api_base = os.getenv("AZURE_OPENAI_ENDPOINT") # your endpoint should look like the following https://YOUR_RESOURCE_NAME.openai.azure.com/
openai.api_type = 'azure'
openai.api_version = '2024-02-01' # this might change in the future

deployment_name='REPLACE_WITH_YOUR_DEPLOYMENT_NAME' #This will correspond to the custom name you chose for your deployment when you deployed a model. 

# Send a completion call to generate an answer
print('Sending a test completion job')
start_phrase = 'Write a tagline for an ice cream shop. '
response = openai.Completion.create(engine=deployment_name, prompt=start_phrase, max_tokens=10)
text = response['choices'][0]['text'].replace('\n', '').replace(' .', '.').strip()
print(start_phrase+text)

Inserciones

OpenAI Python 1.x
OpenAI Python 0.28.1

import os
from openai import AzureOpenAI

client = AzureOpenAI(
  api_key = os.getenv("AZURE_OPENAI_API_KEY"),  
  api_version = "2024-02-01",
  azure_endpoint =os.getenv("AZURE_OPENAI_ENDPOINT") 
)

response = client.embeddings.create(
    input = "Your text string goes here",
    model= "text-embedding-ada-002"  # model = "deployment_name".
)

print(response.model_dump_json(indent=2))

Puede encontrar ejemplos adicionales, sobre cómo controlar la búsqueda de texto semántico sin embeddings_utils.py en nuestro tutorial de inserción.

import openai

openai.api_type = "azure"
openai.api_key = YOUR_API_KEY
openai.api_base = "https://YOUR_RESOURCE_NAME.openai.azure.com"
openai.api_version = "2024-02-01"

response = openai.Embedding.create(
    input="Your text string goes here",
    engine="YOUR_DEPLOYMENT_NAME"
)
embeddings = response['data'][0]['embedding']
print(embeddings)

Async

OpenAI no admite la llamada a métodos asincrónicos en el cliente de nivel de módulo; en su lugar, debe crear instancias de un cliente asincrónico.

import os
import asyncio
from openai import AsyncAzureOpenAI

async def main():
    client = AsyncAzureOpenAI(  
      api_key = os.getenv("AZURE_OPENAI_API_KEY"),  
      api_version = "2024-02-01",
      azure_endpoint = os.getenv("AZURE_OPENAI_ENDPOINT")
    )
    response = await client.chat.completions.create(model="gpt-35-turbo", messages=[{"role": "user", "content": "Hello world"}]) # model = model deployment name

    print(response.model_dump_json(indent=2))

asyncio.run(main())

Autenticación

from azure.identity import DefaultAzureCredential, get_bearer_token_provider
from openai import AzureOpenAI

token_provider = get_bearer_token_provider(DefaultAzureCredential(), "https://cognitiveservices.azure.com/.default")

api_version = "2024-02-01"
endpoint = "https://my-resource.openai.azure.com"

client = AzureOpenAI(
    api_version=api_version,
    azure_endpoint=endpoint,
    azure_ad_token_provider=token_provider,
)

completion = client.chat.completions.create(
    model="deployment-name",  # model = "deployment_name"
    messages=[
        {
            "role": "user",
            "content": "How do I output all files in a directory using Python?",
        },
    ],
)
print(completion.model_dump_json(indent=2))

Uso de los datos

Para conocer los pasos de configuración completos necesarios para que estos ejemplos de código funcionen, consulte el Inicio rápido sobre el uso de los datos.

OpenAI Python 1.x
OpenAI Python 0.28.1

import os
import openai
import dotenv

dotenv.load_dotenv()

endpoint = os.environ.get("AZURE_OPENAI_ENDPOINT")
api_key = os.environ.get("AZURE_OPENAI_API_KEY")
deployment = os.environ.get("AZURE_OPEN_AI_DEPLOYMENT_ID")

client = openai.AzureOpenAI(
    base_url=f"{endpoint}/openai/deployments/{deployment}/extensions",
    api_key=api_key,
    api_version="2023-08-01-preview",
)

completion = client.chat.completions.create(
    model=deployment, # model = "deployment_name"
    messages=[
        {
            "role": "user",
            "content": "How is Azure machine learning different than Azure OpenAI?",
        },
    ],
    extra_body={
        "dataSources": [
            {
                "type": "AzureCognitiveSearch",
                "parameters": {
                    "endpoint": os.environ["AZURE_AI_SEARCH_ENDPOINT"],
                    "key": os.environ["AZURE_AI_SEARCH_API_KEY"],
                    "indexName": os.environ["AZURE_AI_SEARCH_INDEX"]
                }
            }
        ]
    }
)

print(completion.model_dump_json(indent=2))

import os
import openai
import dotenv
import requests

dotenv.load_dotenv()

openai.api_base = os.environ.get("AZURE_OPENAI_ENDPOINT")
openai.api_version = "2023-08-01-preview"
openai.api_type = 'azure'
openai.api_key = os.environ.get("AZURE_OPENAI_API_KEY")

def setup_byod(deployment_id: str) -> None:
    """Sets up the OpenAI Python SDK to use your own data for the chat endpoint.

    :param deployment_id: The deployment ID for the model to use with your own data.

    To remove this configuration, simply set openai.requestssession to None.
    """

    class BringYourOwnDataAdapter(requests.adapters.HTTPAdapter):

     def send(self, request, **kwargs):
         request.url = f"{openai.api_base}/openai/deployments/{deployment_id}/extensions/chat/completions?api-version={openai.api_version}"
         return super().send(request, **kwargs)

    session = requests.Session()

    # Mount a custom adapter which will use the extensions endpoint for any call using the given `deployment_id`
    session.mount(
        prefix=f"{openai.api_base}/openai/deployments/{deployment_id}",
        adapter=BringYourOwnDataAdapter()
    )

    openai.requestssession = session

aoai_deployment_id = os.environ.get("AZURE_OPEN_AI_DEPLOYMENT_ID")
setup_byod(aoai_deployment_id)

completion = openai.ChatCompletion.create(
    messages=[{"role": "user", "content": "What are the differences between Azure Machine Learning and Azure AI services?"}],
    deployment_id=os.environ.get("AZURE_OPEN_AI_DEPLOYMENT_ID"),
    dataSources=[  # camelCase is intentional, as this is the format the API expects
        {
            "type": "AzureCognitiveSearch",
            "parameters": {
                "endpoint": os.environ.get("AZURE_AI_SEARCH_ENDPOINT"),
                "key": os.environ.get("AZURE_AI_SEARCH_API_KEY"),
                "indexName": os.environ.get("AZURE_AI_SEARCH_INDEX"),
            }
        }
    ]
)
print(completion)

Corrección de DALL-E

Corrección de DALLE
Corrección asincrónica de DALLE

import time
import json
import httpx
import openai


class CustomHTTPTransport(httpx.HTTPTransport):
    def handle_request(
        self,
        request: httpx.Request,
    ) -> httpx.Response:
        if "images/generations" in request.url.path and request.url.params[
            "api-version"
        ] in [
            "2023-06-01-preview",
            "2023-07-01-preview",
            "2023-08-01-preview",
            "2023-09-01-preview",
            "2023-10-01-preview",
        ]:
            request.url = request.url.copy_with(path="/openai/images/generations:submit")
            response = super().handle_request(request)
            operation_location_url = response.headers["operation-location"]
            request.url = httpx.URL(operation_location_url)
            request.method = "GET"
            response = super().handle_request(request)
            response.read()

            timeout_secs: int = 120
            start_time = time.time()
            while response.json()["status"] not in ["succeeded", "failed"]:
                if time.time() - start_time > timeout_secs:
                    timeout = {"error": {"code": "Timeout", "message": "Operation polling timed out."}}
                    return httpx.Response(
                        status_code=400,
                        headers=response.headers,
                        content=json.dumps(timeout).encode("utf-8"),
                        request=request,
                    )

                time.sleep(int(response.headers.get("retry-after")) or 10)
                response = super().handle_request(request)
                response.read()

            if response.json()["status"] == "failed":
                error_data = response.json()
                return httpx.Response(
                    status_code=400,
                    headers=response.headers,
                    content=json.dumps(error_data).encode("utf-8"),
                    request=request,
                )

            result = response.json()["result"]
            return httpx.Response(
                status_code=200,
                headers=response.headers,
                content=json.dumps(result).encode("utf-8"),
                request=request,
            )
        return super().handle_request(request)


client = openai.AzureOpenAI(
    azure_endpoint="<azure_endpoint>",
    api_key="<api_key>",
    api_version="<api_version>",
    http_client=httpx.Client(
        transport=CustomHTTPTransport(),
    ),
)
image = client.images.generate(prompt="a cute baby seal")

print(image.data[0].url)

import time
import asyncio
import json
import httpx
import openai


class AsyncCustomHTTPTransport(httpx.AsyncHTTPTransport):
    async def handle_async_request(
        self,
        request: httpx.Request,
    ) -> httpx.Response:
        if "images/generations" in request.url.path and request.url.params[
            "api-version"
        ] in [
            "2023-06-01-preview",
            "2023-07-01-preview",
            "2023-08-01-preview",
            "2023-09-01-preview",
            "2023-10-01-preview",
        ]:
            request.url = request.url.copy_with(path="/openai/images/generations:submit")
            response = await super().handle_async_request(request)
            operation_location_url = response.headers["operation-location"]
            request.url = httpx.URL(operation_location_url)
            request.method = "GET"
            response = await super().handle_async_request(request)
            await response.aread()

            timeout_secs: int = 120
            start_time = time.time()
            while response.json()["status"] not in ["succeeded", "failed"]:
                if time.time() - start_time > timeout_secs:
                    timeout = {"error": {"code": "Timeout", "message": "Operation polling timed out."}}
                    return httpx.Response(
                        status_code=400,
                        headers=response.headers,
                        content=json.dumps(timeout).encode("utf-8"),
                        request=request,
                    )

                await asyncio.sleep(int(response.headers.get("retry-after")) or 10)
                response = await super().handle_async_request(request)
                await response.aread()

            if response.json()["status"] == "failed":
                error_data = response.json()
                return httpx.Response(
                    status_code=400,
                    headers=response.headers,
                    content=json.dumps(error_data).encode("utf-8"),
                    request=request,
                )

            result = response.json()["result"]
            return httpx.Response(
                status_code=200,
                headers=response.headers,
                content=json.dumps(result).encode("utf-8"),
                request=request,
            )
        return await super().handle_async_request(request)


async def dall_e():
    client = openai.AsyncAzureOpenAI(
        azure_endpoint="<azure_endpoint>",
        api_key="<api_key>",
        api_version="<api_version>",
        http_client=httpx.AsyncClient(
            transport=AsyncCustomHTTPTransport(),
        ),
    )
    image = await client.images.generate(prompt="a cute baby seal")

    print(image.data[0].url)

asyncio.run(dall_e())

Cambios de nombre

Nota:

Se han quitado todos los métodos a*; debe usarse el cliente asincrónico en su lugar.

OpenAI Python 0.28.1	OpenAI Python 1.x
`openai.api_base`	`openai.base_url`
`openai.proxy`	`openai.proxies`
`openai.InvalidRequestError`	`openai.BadRequestError`
`openai.Audio.transcribe()`	`client.audio.transcriptions.create()`
`openai.Audio.translate()`	`client.audio.translations.create()`
`openai.ChatCompletion.create()`	`client.chat.completions.create()`
`openai.Completion.create()`	`client.completions.create()`
`openai.Edit.create()`	`client.edits.create()`
`openai.Embedding.create()`	`client.embeddings.create()`
`openai.File.create()`	`client.files.create()`
`openai.File.list()`	`client.files.list()`
`openai.File.retrieve()`	`client.files.retrieve()`
`openai.File.download()`	`client.files.retrieve_content()`
`openai.FineTune.cancel()`	`client.fine_tunes.cancel()`
`openai.FineTune.list()`	`client.fine_tunes.list()`
`openai.FineTune.list_events()`	`client.fine_tunes.list_events()`
`openai.FineTune.stream_events()`	`client.fine_tunes.list_events(stream=True)`
`openai.FineTune.retrieve()`	`client.fine_tunes.retrieve()`
`openai.FineTune.delete()`	`client.fine_tunes.delete()`
`openai.FineTune.create()`	`client.fine_tunes.create()`
`openai.FineTuningJob.create()`	`client.fine_tuning.jobs.create()`
`openai.FineTuningJob.cancel()`	`client.fine_tuning.jobs.cancel()`
`openai.FineTuningJob.delete()`	`client.fine_tuning.jobs.create()`
`openai.FineTuningJob.retrieve()`	`client.fine_tuning.jobs.retrieve()`
`openai.FineTuningJob.list()`	`client.fine_tuning.jobs.list()`
`openai.FineTuningJob.list_events()`	`client.fine_tuning.jobs.list_events()`
`openai.Image.create()`	`client.images.generate()`
`openai.Image.create_variation()`	`client.images.create_variation()`
`openai.Image.create_edit()`	`client.images.edit()`
`openai.Model.list()`	`client.models.list()`
`openai.Model.delete()`	`client.models.delete()`
`openai.Model.retrieve()`	`client.models.retrieve()`
`openai.Moderation.create()`	`client.moderations.create()`
`openai.api_resources`	`openai.resources`

Quitado

openai.api_key_path
openai.app_info
openai.debug
openai.log
openai.OpenAIError
openai.Audio.transcribe_raw()
openai.Audio.translate_raw()
openai.ErrorObject
openai.Customer
openai.api_version
openai.verify_ssl_certs
openai.api_type
openai.enable_telemetry
openai.ca_bundle_path
openai.requestssession (OpenAI ahora usa httpx)
openai.aiosession (OpenAI ahora usa httpx)
openai.Deployment (Usado anteriormente para Azure OpenAI)
openai.Engine
openai.File.find_matching_files()

Compartir a través de