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.
Steps to reproduce the issue locally
Run the following code through type-checking with pyright:
from array import array
import pinecone
embedding_bytes = b'' # in real-world scenarios this of course will be non-empty
embedding_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
Environment
- OS Version: NixOS 25.05
- Python version: 3.12.19
- Python SDK version: 7.0.2
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, andarraydoes not satisfyList, so previously-legal values are no longer accepted.Steps to reproduce the issue locally
Run the following code through type-checking with pyright:
It will fail as follows:
Environment