OpenAI has introduced the beta version of its Python SDK, marking an important step toward improving access to the OpenAI API for Python developers. The OpenAI Python Library provides a simplified way for Python-based applications to interact with the OpenAI API, while providing the opportunity for testing and feedback prior to the official release of version 1.0.
The OpenAI Python library provides convenient access to the OpenAI API from applications written in the Python language. It includes a predefined set of classes for API resources that are dynamically initialized from API responses, making it compatible with a wide range of OpenAI API versions.
For examples of using the OpenAI Python library, see the API reference and the OpenAI cookbook.
Furnishings
First, make sure you have Python 3.7.1 or later. If you just want to use the package, run:
pip install –upgrade openai
After installing the package, import it at the beginning of a file:
To install this package from source and make changes, run the following command from the repository root:
Optional dependencies
Install dependencies for openai.embeddings_utils:
pip install openai[embeddings]
Install weights and biases support: pip install openai[wandb]
Data libraries like Numpy and Pandas are not installed by default due to their size. They are required for some functions of this library, but generally not for communicating with the API. If a MissingDependencyError occurs, install it with: pip install openai[datalib]
Use
The library must be configured with your account’s secret key, which is available on the website. Set it as the OPENAI_API_KEY environment variable before using the library:
export OPENAI_API_KEY=’sk-…’
Or set openai.api_key to its value: openai.api_key = “sk-…”
Examples of using this library to accomplish various tasks can be found in the OpenAI Cookbook. It includes code examples for: classification using fine-tuning, clustering, code searching, customizing embeds, answering questions from a corpus of documents, recommendations, visualizing embeds, and more.
Most endpoints support a request_timeout parameter. This parameter accepts a union[float, Tuple[float, float]]and throws an openai.error.Timeout error if the request exceeds this time in seconds (see: https://requests.readthedocs.io/en/l…tart/#timeouts).
Cat completions
Chat templates like gpt-3.5-turbo and gpt-4 can be accessed from the Chat Completions endpoint.
1 | Completion = openai.ChatCompletion.create(model=”gpt-3.5-turbo”, messages=[{“role”: “user”, “content”: “Hello world”}]) print(completion.choices[0].message content) |
degrees
Text templates like babbage-002 or davinci-002 (and older completion templates) can be accessed from the completion endpoint.
1 | Completion = openai.Completion.create(model=”davinci-002″, prompt=”Hello World”) print(completion.choices[0].Text) |
Incorporations
Embeddings are intended to measure the similarity or relevance between text strings. To get an embed for a text string you can use:
1 | text_string = “Example text” model_id = “text-embedding-ada-002” embedding = openai.Embedding.create(input=text_string, model=model_id)[‘data’][0][’embedding’] |
Fine tuning
Fine-tuning a model using training data can both improve results (by giving the model more examples to learn from) and reduce the cost/latency of API calls by reducing the need to insert training examples into the Record prompts.
1 | # Create a fine-tuning job with an already uploaded file openai.FineTuningJob.create(training_file=”file-abc123″, model=”gpt-3.5-turbo”) # List 10 fine-tuning jobs openai.FineTuningJob.list(limit =10) # The status Retrieve a tune openai.FineTuningJob.retrieve(“ft-abc123”) # Cancel a job openai.FineTuningJob.cancel(“ft-abc123″) # List up to 10 events from a tune Tune job openai.FineTuningJob.list_events(id=” ft-abc123″, limit=10) # Delete a fine-grained model (must be owned by the organization where the model was created) openai.Model.delete(“ft: gpt-3.5-turbo:acemeco:suffix:abc123” ) |
To save training results from fine-tuning weights and biases, use the following:
moderation
OpenAI provides a free moderation endpoint to verify that content complies with OpenAI’s content policy.
moderation_resp = openai.Moderation.create(input=”Here is a completely harmless text that complies with all OpenAI content guidelines.”)
Image generation (DALL-E)
DALL-E is a generative image model that can create new images based on a prompt.
image_resp = openai.Image.create(prompt=”two dogs playing chess, oil painting”, n=4, size=”512×512″)
Audio (whisper)
The Speech-to-Text API provides two endpoints, transcriptions and translations, based on the open source Whisper Large v2 model.
1 | f = open(“path/to/file.mp3”, “rb”) transcript = openai.Audio.transcribe(“whisper-1”, f) transcript = openai.Audio.translate(“whisper-1”, f) |
Asynchronous API
Support for Async is available in the API by adding a network-bound method:
1 | async def create_chat_completion(): chat_completion_resp = waiting openai.ChatCompletion.acreate(model=”gpt-3.5-turbo”, messages=[{“role”: “user”, “content”: “Hello world”}]) |
To make asynchronous requests more efficient, you can pass your own aiohttp.ClientSession, but you will have to manually close the client session at the end of your program/event loop:
1 | from aiohttp import ClientSession openai.aiosession.set(ClientSession()) # At the end of your program, close the http session. waiting openai.aiosession.get().close() |
Command line interface
This library also provides an OpenAI command line utility that makes it easy to interact with the API from your terminal. Run openai api -h to use it.
1 | # List models openai api models.list # create a chat completion (gpt-3.5-turbo, gpt-4, etc.) openai api chat_completions.create -m gpt-3.5-turbo -g user “Hello world” # create a completion (text-davinci-003, text-davinci-002, ada, babbage, curie, davinci, etc.) openai api completions.create -m ada -p “Hello World” # Generate images via the DALL·E API openai api image. create -p “two dogs playing chess, cartoon” -n 1 # using openai via a proxy openai –proxy=http://proxy.com api models.list |
Microsoft Azure endpoints
To use the library with Microsoft Azure endpoints, you must define api_type, api_base, and api_version in addition to the api_key. The api_type must be set to “azure” and the others correspond to the properties of your site. Additionally, the deployment name must be passed as an engine parameter.
1 | import openai openai.api_type = “azure” openai.api_key = “…” openai.api_base = “https://example-endpoint.openai.azure.com” openai.api_version = “2023-05-15″ # create a chat completion chat_completion = openai.ChatCompletion.create(deployment_id=”deployment-name”, model=”gpt-3.5-turbo”, messages=[{“role”: “user”, “content”: “Hello world”}]) # print the completion print(chat_completion.choices[0].message content) |
Please note that Microsoft Azure endpoints can currently only be used for completion, integration, and tuning operations.
Microsoft Azure Active Directory authentication
To use Microsoft Active Directory to authenticate your Azure endpoint, you must set the api_type to azure_ad and pass the purchased api_key authentication token. Other parameters need to be set as stated in the previous section.
1 | from azure.identity import DefaultAzureCredential import openai # Request credentials default_credential = DefaultAzureCredential() token = default_credential.get_token(“https://cognitiveservices.azure.com/.default”) # Setup parameters openai.api_type = “azure_ad” openai. api_key = token.token openai.api_base = “https://example-endpoint.openai.azure.com/” openai.api_version = “2023-05-15” |
Source : Python SDK
And you ?
What do you think ?
Which functions do you find interesting?
See also
OpenAI allows developers to integrate ChatGPT into their applications via an API. However, is the tool ready for use in production environments?
With OpenAI, developers can now bring their own data to customize GPT-3.5 Turbo and build and run better models for their use cases
OpenAI announces the general availability of its GPT-4 API, enabling developers to integrate the latest generation of its generative AI into their applications