OpenAI Agents financial research and reasoning examples

[jssmith]

What was changed

Ported financial research and reasoning examples from OpenAI Agents SDK.

Why?

Checklist

1. Closes

2. How was this tested:

3. Any docs updates needed?

[cretz]

. (moved to easily replyable comment, can't delete this unfortunately)

[cretz]

Currently, the top-level of openai_agents has all of the samples mashed together as one project at the top level dir. I think it makes the addition of this as a subdir a bit confusing. I wonder if we should move the current sample set into a subdir (e.g. "simple" or something) or make each sample its own dir as we are doing with these newer ones.

EDIT: I now see #221 will fix this, thanks! Can disregard this comment.

[cretz]

I am a little concerned we are adding lots of OpenAI sample code with no tests to confirm it continues to work. Granted I know this is unfortunately OpenAI's approach to samples, but ideally we can have some test somewhere. We have other samples where we have done this and it comes back to bite us hard when people report it stops working and we have to scramble (this is the case with the Sentry sample which has now remained broken for a long time).

(just adding this general comment, deferring code review to Dan/Tim)

[jssmith]

I agree.

Before we said testing the samples would be redundant because the samples were basically the same as the tests already in the SDK. Now we are adding more samples and they will become unmaintainable without tests.

The best way to do this is with end-to-end tests because we want to validate the sample as-written. We'll need to write a little tool to do this because some samples are interactive and all of them call LLMs, making it difficult to validate they do the right thing. We are probably just asserting that they run to completion without errors.

[cretz]

👍 Should we open an issue to track this?