Supporting OpenAI's Flex Processing

[empezarcero]

Description

OpenAI released a new service tier called flex.
It reduces the cost of the tokens at the expense of performance/availability.
Is it possible to add a service_tier parameter to the OpenAIModel and the ability to control the timeout value?

References

OpenAI documentation

[Kludex]

Yeah, it's possible.

We'd want to add the field openai_service_tier on the OpenAIModelSettings. Would you like to contribute?

[empezarcero]

@Kludex, I created #1923

[empezarcero]

Feature merged to main