fix(openai): accept valid responses that are falsy at runtime

[Mattijs Ugen (akaIDIOT)]

When using structured output through model.with_structured_output(MyModel) using a MyModel class definition that happens to implement magic methods that influence the truthiness of an instance, a valid response that evaluates to False causes _oai_structured_outputs_parser to fall through all of the options reading the response, concluding that the response must be faulty. Moving from a truthiness check to a presence check (is not None) fixes this; valid responses of my model class that happen to have length 0 are now returned as expected.

See the added test for a minimal example of this behaviour (the test fails without the single line change in langchain_openai/chat_models/base.py).

[codspeed-hq]

Merging this PR will not alter performance

✅ 6 untouched benchmarks
⏩ 29 skipped benchmarks¹

---

_{Comparing akaIDIOT:akaidiot/accept-falsy-structured-result (93ee302) with master (b004103)}

Footnotes

1. 29 benchmarks were skipped, so the baseline results were used instead. If they were deleted from the codebase, click here and archive them to remove them from the performance reports. ↩

[ccurme (ccurme)]

Thank you!