jiggy-ai / hnsqlite Goto Github PK

Thanks for the repo.
I want to know if you have any plans to support haystack

https://docs.haystack.deepset.ai/docs/document_store#choosing-the-right-document-store

Thanks again!

Sometimes when i try to delete a file using metadata, it gives the error label not found.whats does mean by this error? why it does occur? and how can i resolve it?

2023-06-15 10:21:22.928 | INFO | hnsqlite.collection:delete:608 - Deleting items from newcollection with filter: {'file_name': 'xyz.pdf'}

RuntimeError Traceback (most recent call last)
in <cell line: 1>()
----> 1 collection_test.delete(filter=filter_dict)

/usr/local/lib/python3.10/dist-packages/hnsqlite/collection.py in delete(self, doc_ids, filter, delete_all)
613 if filter_item(filter, embedding.metadata_as_dict()):
614 count += 1
--> 615 self.hnsw_ix.mark_deleted(embedding.id)
616 session.delete(embedding)
617 session.commit()

RuntimeError: Label not found

inital tests show that that a random forest regression is effective at predicting near-optimal values of M, ef_construction, ef parameters given sufficient dataset.

How does this package compare to: https://github.com/asg017/sqlite-vss ?

depends on nmslib/hnswlib#444

I wasn't able to find a way to choose the folder where the index is stored, as it is possible with the sqlite.
Could you expose that parameter? If needed, I can try to do it an open a PR.

Great project.

Running your minimal example, I get this error:

collection.py:305
AttributeError: 'ScalarResult' object has no attribute 'yield_per'