Fix metadata deserialization in async mode for PGVector #125
Conversation
|
Facing this very same issue. Wondering when this commit could be merged. Regards |
| @@ -1058,17 +1058,38 @@ async def asimilarity_search_with_score_by_vector( | |||
|
|
|||
| def _results_to_docs_and_scores(self, results: Any) -> List[Tuple[Document, float]]: | |||
There was a problem hiding this comment.
Could you add a relevant unit test(s) for this?
There was a problem hiding this comment.
Hi @eyurtsev,
Thank you for your feedback. I've added unit tests to cover metadata deserialization in asynchronous operations, specifically targeting the _results_to_docs_and_scores method. The new tests ensure that metadata is correctly deserialized and that Document instances receive the proper metadata when using async_mode=True.
Please review the updated tests and let me know if you have any further comments or suggestions.
Best regards,
|
@eyurtsev, please let me know if there is anything else I need to do. If not, please merge it. |
Problem
When using asynchronous methods with
PGVector(async_mode=True), themetadatafield retrieved from the database may be of typeFragment(fromasyncpg) or other non-dict types. This causes aValidationErrorwhen theDocumentclass expectsmetadatato be a dictionary.Solution
This pull request modifies the
_results_to_docs_and_scoresmethod to ensure thatmetadatais correctly converted into a dictionary before creatingDocumentinstances. The method now handles different possible types ofmetadataand attempts to deserialize it into a dict.Changes
_results_to_docs_and_scoresmethod inPGVectorclass to handlemetadatadeserialization for different types (e.g.,dict,str,Fragment).Testing
async_mode=Trueand confirmed that themetadatafield is correctly deserialized and no longer causes validation errors.async_mode=False.Related Issues