Implement Parallel Function Calls with List[Union[T]]
#378
Conversation
jxnl
changed the title
List[Union[T]]
![ellipsis-dev[bot]](https://avatars.githubusercontent.com/in/64358?s=60&v=4){kind=small}
There was a problem hiding this comment.
Changes requested.
Something look wrong? You can customize Ellipsis by editing the ellipsis.yaml for this repository.
Generated with ❤️ by ellipsis.dev
instructor/dsl/parallel.py
Outdated
| #! We expect this from the OpenAISchema class, We should address | ||
| #! this with a protocol or an abstract class... @jxnlco | ||
| assert mode == Mode.PARALLEL_TOOLS, "Mode must be PARALLEL_TOOLS" | ||
| print(response.choices[0].message) |
There was a problem hiding this comment.
There is a print statement on line 24 which seems to be for debugging purposes. It's a good practice to remove such statements before merging the code to the main branch.
| print(response.choices[0].message) | |
| # print(response.choices[0].message) |
|
Would a change from Iterable[Union[…]] to tuple[…] or even just (…) be to your liking aesthetically? Happy to figure out the implementation and update this PR if you’re aligned |
its not a tuple. its |
I agree that that's the correct type, I'm updating my suggestion to passing a list of |
we're going to use generics to reflect the type to the response. resp: [T1, T2, T3] = create(response_model=[T1, T2, T3]) is not a thing |
Summary:
This PR introduces parallel function calls using a list of union types, modifies the instructor's patching mechanism to handle these calls, and includes tests for the new functionality.
Key points:
PARALLEL_TOOLStoModeenum ininstructor/function_calls.py.ParallelBaseclass ininstructor/dsl/parallel.pyto handle parallel function calls.instructor/patch.pyto handlePARALLEL_TOOLSmode and useParallelBasefor response processing.tests/openai/test_parallel.py.examples/patching/pcalls.py.Generated with ❤️ by ellipsis.dev