OpenAI

Dispatch provides a built-in integration with openai to retry failed OpenAI requests and exceptions. It’s enabled by default and it takes no extra code to set it up.

import dispatch
from openai import Client

openai = Client(api_key="<your API key>")


@dispatch.function
def ask_chatgpt(question):
    chat_completion = client.chat.completions.create(
        model="gpt-3.5-turbo",
        messages=[
            {"role": "system", "content": "You are a helpful assistant"},
            {"role": "user", "content": question},
        ],
    )

    return chat_completion.choices[0].message.content


def main():
    ask_chatgpt.dispatch("What is the most loved programming language?")


dispatch.run(main)

Exception handling

APIStatusError

Depending on the status_code attribute, a different Dispatch behavior is applied.

HTTP status code	Description	Dispatch status
1xx	Informational	Status.PERMANENT_ERROR
2xx	Successful	Status.OK
3xx	Redirection	Status.PERMANENT_ERROR
3xx	Redirection	Status.PERMANENT_ERROR
4xx	Client error	Status.PERMANENT_ERROR
400	Bad request	Status.INVALID_ARGUMENT
401	Unauthorized	Status.UNAUTHENTICATED
403	Forbidden	Status.PERMISSION_DENIED
404	Not found	Status.NOT_FOUND
408	Request timeout	Status.TIMEOUT
429	Too many requests	Status.THROTTLED
5xx	Server error	Status.TEMPORARY_ERROR
501	Not implemented	Status.PERMANENT_ERROR

APITimeoutError

Returns Status.TIMEOUT.

OpenAIError

Returns Status.TEMPORARY_ERROR, unless exception is APIStatusError or APITimeoutError, since OpenAIError is a parent class of these. Learn more in openai documentation.