You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
[Bug] Incompatible change: array[float] no longer accepted by type signature #511
Is this a new bug?
In pinecone-client 5.0.1, an array[float] could be used when creating indexes.
In version 7.0.2, this may still work in practice with type-checking overridden; however, the type definitions indicate that only List[float] is permitted, and array does not satisfy List, so previously-legal values are no longer accepted.
I believe this is a new bug
I have searched the existing Github issues and Community Forum, and I could not find an existing post for this bug
Steps to reproduce the issue locally
Run the following code through type-checking with pyright:
fromarrayimportarrayimportpineconeembedding_bytes=b''# in real-world scenarios this of course will be non-emptyembedding_arr=array("f", embedding_bytes)
pc=pinecone.Pinecone()
idx=pc.Index(name="test")
idx.upsert(vectors=[("name", embedding_arr)])
It will fail as follows:
error: Argument of type "list[tuple[str, array[float]]]" cannot be assigned to parameter "vectors" of type "List[Vector] | List[VectorTuple] | List[VectorTupleWithMetadata] | List[VectorTypedDict]" in function "upsert"
"tuple[str, array[float]]" is not assignable to "Vector" (reportArgumentType)
1 error, 0 warnings, 0 informations
Activity
kris-gaudel commentedon Jul 16, 2025
Hi is this issue still available? I'm interested in contributing!
fix(types): accept Sequence[float] for vector values in VectorTuple
test(integration): prove legacy create() sugar serves the real vector…