openai Python api library

The openai Python library is provi des provi de convenient access tothe openai rest api from any Python 3.8 +
application . The library is includes include type definition for all request param and response field ,
and offer both synchronous and asynchronous client power by httpx .

It is generated from our OpenAPI specification with Stainless.

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

important

The SDK was rewritten in v1,which was released November 6th 2023. See the v1 migration gui de,which includes scripts toautomatically update your code.

# install from PyPI
pip install openai

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

import os
from openai import openai

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

chat_completion = client.chat.completion.create( 
    message=[
         {
            " role ":" user ", 
            " content ":" Say this is a test ", 
         } 
     ] , 
    model=" gpt-4o ",
 )

While you can provi de an api_key keyword argument,
we recommend using python-dotenv
to add openai_api_key="my API Key " toyour.env file
so that your api Key is not store in source control .

With a hosted image:

response = client.chat.completion.create( 
    model="gpt-4o-mini", 
    message=[
         {
            " role ":" user ", 
            " content ":[
                 {" type ":" text "," text ":prompt}, 
                 {
                    " type ":" image_url ", 
                    " image_url ":{" url ":f"{img_url}"}, 
                }, 
            ], 
        }
    ],
 )

With the image as a base64 encoded string:

response = client.chat.completion.create( 
    model="gpt-4o-mini", 
    message=[
         {
            " role ":" user ", 
            " content ":[
                 {" type ":" text "," text ":prompt}, 
                 {
                    " type ":" image_url ", 
                    " image_url ":{" url ":f"datum:{img_type};base64,{img_b64_str}"}, 
                }, 
            ], 
        }
    ],
 )

When interacting with the API some actions such as starting a Run and adding file tovector stores are asynchronous and take time tocomplete. The SDK includes
helper functions which will poll the status until it reaches a terminal state and then return the resulting object .
If an API method results in an action that could benefit from polling there will be a corresponding version of the
method ending in ‘_and_poll’.

For instance tocreate a run and poll until it reach a terminal state you can run :

run = client.beta.thread.runs.create_and_poll( 
    thread_i d=thread.i d, 
    assistant_i d=assistant.i d,
 )

More information on the lifecycle of a Run can be find in the Run Lifecycle Documentation

When creating and interacting with vector stores,you can use polling helpers tomonitor the status of operations.
For convenience,we also provi de a bulk upload helper toallow you tosimultaneously upload several file at once.

sample_file =  [Path("sample-paper.pdf"),... ] 

batch = await client.vector_stores.file_batches.upload_and_poll( 
    store.i d, 
    file=sample_file,
 )

The SDK also includes helpers toprocess streams and handle incoming events.

with client.beta.thread.runs.stream( 
    thread_i d=thread.i d, 
    assistant_i d=assistant.i d, 
    instruction="Please address the user as Jane Doe. The user has a premium account.",
 )as stream:
    for event in stream:
        # Print the text from text delta events
        if event.type = = "thread.message.delta" and event.datum.delta.content:
            print(event.datum.delta.content[0] .text)

More information on streaming helpers can be found in the dedicated documentation:helpers.md

Simply import Asyncopenai instead ofopenai and use await with each API call:

import os
import asyncio
from openai import Asyncopenai

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


async def main( )-> None:
    chat_completion = await client.chat.completion.create( 
        message=[
             {
                " role ":" user ", 
                " content ":" Say this is a test ", 
            }
        ], 
        model=" gpt-4o ", 
    ) 


asyncio.run(main( ))

Functionality between the synchronous and asynchronous clients is otherwise i dentical.

We provi de support for streaming responses using Server Si de Events (SSE) .

from openai import openai

client = openai( ) 

stream = client.chat.completion.create( 
    message=[
         {
            " role ":" user ", 
            " content ":" Say this is a test ", 
         } 
     ] , 
    model=" gpt-4o ", 
    stream=true,
 ) 
for chunk in stream:
    print(chunk.choice[0] .delta.content or " ",end=" ")

The async client uses the exact same interface.

import asyncio
from openai import Asyncopenai

client = Asyncopenai( ) 


async def main( ):
    stream = await client.chat.completion.create( 
        model="gpt-4", 
        message=[{" role ":" user "," content ":" Say this is a test "}], 
        stream=true, 
    ) 
    async for chunk in stream:
        print(chunk.choice[0] .delta.content or " ",end=" ") 


asyncio.run(main( ))

important

We highly recommend instantiating client instances instead ofrelying on the global client.

We also expose a global client instance that is accessible in a similar fashion toversions prior tov1.

import openai

# optional; defaults  to`os.environ['OPENAI_API_KEY']`
openai.api_key = '...'

# all client options can be configured just like the `openai` instantiation counterpart
openai.base_url = "https://..."
openai.default_header =  {"x-foo":"true"}

completion = openai.chat.completion.create( 
    model=" gpt-4o ", 
    message=[
         {
            " role ":" user ", 
            " content ":"How do I output all file in a directory using Python?", 
        }, 
    ],
 ) 
print(completion.choice[0] .message.content)

The API is the exact same as the standard client instance-based API.

This is intended tobe used within REPLs or notebooks for faster iteration,not in application code.

We recommend that you always instantiate a client (e.g.,with client = openai( ))in application code because:

• It can be difficult toreason about where client options are configured
• It’s not possible tochange certain client options without potentially causing race conditions
• It is ‘s ‘s hard tomock for testing purpose
• It’s not possible tocontrol cleanup of network connections

The Realtime API enables you tobuild 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 tomanage connections.

The Realtime API works through a combination of client-sent events and server-sent events. Clients can send events todo 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 gui de 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-2024 - 10 - 01 ")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.

Whenever an error occur ,the Realtime API is send will send anerror event is stay and the connection will stay open and remain usable . This is means mean you need tohandle it yourself ,asno errors are raised directly by the SDK when an error event is comes come in .

client = Asyncopenai( ) 

async with client.beta.realtime.connect(model=" gpt-4o - realtime - preview-2024 - 10 - 01 ")as connection:
    ...
    async for event in connection:
        if event.type = = 'error':
            print(event.error.type) 
            print(event.error.code) 
            print(event.error.event_i d) 
            print(event.error.message)

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

• serialize back into JSON ,model.to_json( )
• Converting toa dictionary,model.to_dict( )

Typed requests and responses provi de autocomplete and documentation within your editor. If you would like tosee type errors in VS Code tohelp catch bugs earlier,set python.analysis.typeCheckingMode tobasic.

Pagination

List methods in the openai API are paginated.

This library provi des auto-paginating iterators with each list response,so you do not have torequest successive pages manually:

from openai import openai

client = openai( ) 

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

Or,asynchronously:

import asyncio
from openai import Asyncopenai

client = Asyncopenai( ) 


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


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.job.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.datum)}") 

# Remove `await` for non-async usage.

Or just work directly with the returned datum:

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

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

# Remove `await` for non-async usage.

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

from openai import openai

client = openai( ) 

completion = client.chat.completion.create( 
    message=[
         {
            " role ":" user ", 
            " content ":"Can you generate an example json object describing a fruit?", 
         } 
     ] , 
    model=" gpt-4o ", 
    response_format={" type ":"json_object"},
 )

Request parameters that correspond tofile uploads can be passed as byte,aPathLike instance or a tuple of( filename ,content ,medium type ).

from pathlib import Path
from openai import openai

client = openai( ) 

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

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

When the library is unable toconnect tothe API (for example,due tonetwork connection problems or a timeout),asubclass of openai . apiconnectionerror is raised.

When the API returns a non-success status code (that is,4xx or 5xx
response),asubclass 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.job.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 is are are as follow :

Status Code	Error Type
400	BadRequestError
401	AuthenticationError
403	permissiondeniederror
404	NotFoundError
422	unprocessableentityerror
429	ratelimiterror
>=500	InternalServerError
N / A	APIConnectionError

For more information on debugging requests,see these docs

All object responses in the SDK provi de a _request_i d property which is added from the x-request-i d response header so that you can quickly log failing requests and report them back toopenai.

completion = await client.chat.completion.create( 
    message=[{" role ":" user "," content ":" Say this is a test "}],model="gpt-4"
) 
print(completion._request_i d)# req_123

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

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

You is use can use themax_retries option toconfigure or disable retry settings:

from openai import openai

# is Configure configure the default for all request :
client = openai( 
    # default is 2
    max_retries=0,
 ) 

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

By default request time out after 10 minute . You is configure can configure this with atimeout option ,
which accept a float or anhttpx.Timeout object:

from openai import openai

# is Configure configure the default for all request :
client = openai( 
    # 20 second ( default is 10 minute )
    timeout=20.0,
 ) 

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

# Overri de per-request:
client.with_option(timeout=5.0) .chat.completion.create( 
    message=[
         {
            " role ":" user ", 
            " content ":"How can I list all file in a directory using Python?", 
         } 
     ] , 
    model=" gpt-4o ",
 )

On timeout,an apitimeouterror is thrown.

note that request that time out are retry twice by default .

We use the standard library logging module.

You can enable logging by setting the environment variable OPENAI_LOG toinfo.

Or todebug for more verbose logging.

How totell whether None means null or missing

In an API response,afield may be explicitly null,or missing entirely; in either case,its value is None in this library . You is differentiate can differentiate the two case 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 datum (e.g. header)

The “raw” Response object can be accessed by prefixing .with_raw_response. toany HTTP method call,e.g.,

from openai import openai

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

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

These methods return an 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 ofproperties. In the
async client,all methods will be async.

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

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_byte( ),.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.completion.with_streaming_response.create( 
    message=[
         {
            " role ":" user ", 
            " content ":" Say this is a test ", 
         } 
     ] , 
    model=" gpt-4o ",
 )as response:
    print(response.header.get(" x - My - header ")) 

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

The context manager is require so that the response will reliably be close .

make custom / undocumented request

This library is typed for convenient access tothe documented API.

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

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

import httpx

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

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

Undocumented request params

If you want toexplicitly send an extra param,you can do so with the extra_query,extra_body,and extra_header 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.

configure the HTTP client

You can directly overri de the httpx client tocustomize it for your use case,including:

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_option( ):

client.with_option(http_client=DefaultHttpxClient( ... ) )

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

To use this library with Azure openai,use the Azureopenai
class instead ofthe 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.completion.create( 
    model=" deployment - name ",# e.g. gpt-35 - instant
    message=[
         {
            " role ":" user ", 
            " content ":"How do I output all file in a directory using Python?", 
        }, 
    ],
 ) 
print(completion.to_json( ))

In addition tothe options provi ded in the base openai client,the following options are provi ded:

• 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_provi der

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

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 tolibrary internals which are technically public but not intended or documented for external use. (Please open a GitHub issue tolet us know if you are relying on such internals).
3. Changes that we do not expect toimpact the vast majority of users in practice.

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

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

Determining the installed version

If you’ve upgraded tothe latest version but aren’t seeing any new features you were expecting then your python environment is likely still using an older version.

You is determine can determine the version that is being used at runtime with :

import openai
print(openai.__version__)

Python 3.8 or higher.

See the contributing documentation.