Hybrid search throws error when prefer_grpc=True

[StreetLamb]

I am running a simple hybrid service with fastembed with prefer_grpc=True:

from qdrant_client import QdrantClient

client = QdrantClient(
    url="http://qdrant.localhost:6333", api_key="xxx", prefer_grpc=True
)
client.set_model("BAAI/bge-small-en-v1.5")
client.set_sparse_model("prithivida/Splade_PP_en_v1")

if not client.collection_exists(collection_name="test"):
    client.create_collection(
        collection_name="test",
        vectors_config=client.get_fastembed_vector_params(),
        sparse_vectors_config=client.get_fastembed_sparse_vector_params(),
    )

client.add(
    collection_name="test",
    documents=["Hello there"],
)

However, I get the following error:

  File "/Users/xxx/Desktop/projects/test/test.py", line 34, in <module>
    client.add(
  File "/Users/xxx/Library/Caches/pypoetry/virtualenvs/app-w4MTL1IK-py3.12/lib/python3.12/site-packages/qdrant_client/qdrant_fastembed.py", line 513, in add
    self._validate_collection_info(collection_info)
  File "/Users/xxx/Library/Caches/pypoetry/virtualenvs/app-w4MTL1IK-py3.12/lib/python3.12/site-packages/qdrant_client/qdrant_fastembed.py", line 386, in _validate_collection_info
    sparse_vector_field_name in collection_info.config.params.sparse_vectors
TypeError: argument of type 'NoneType' is not iterable

The error only occurs if prefer_grpc=True. I checked that in the created collection's config, sparse_vectors has fast-sparse-splade_pp_en_v1 so collection_info.config.params.sparse_vectors should not be None.

[joein]

Hi @StreetLamb

Thanks for pointing out, it's actually a bug from our side
I've proposed a fix in #674

[aiorga-sherpas]

When trying to perform an Hybrid Query as shown in the documentation it fails due to not detecting the dense embeddings as a valid query model.

Snipped of code from the documentation:

from qdrant_client import QdrantClient, models

client = QdrantClient(url="http://localhost:6333")

client.query_points(
    collection_name="{collection_name}",
    prefetch=[
        models.Prefetch(
            query=models.SparseVector(indices=[1, 42], values=[0.22, 0.8]),
            using="sparse",
            limit=20,
        ),
        models.Prefetch(
            query=[0.01, 0.45, 0.67, ...],  # <-- dense vector
            using="dense",
            limit=20,
        ),
    ],
    query=models.FusionQuery(fusion=models.Fusion.RRF),
)

Error:

 qdrant_client/conversions/conversion.py", line 2579, in convert_query
    raise ValueError(f"invalid Query model: {model}")  # pragma: no cover
    ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
ValueError: invalid Query model:

qdrant-client/qdrant_client/conversions/conversion.py

Line 2560 in bbe17b3

def convert_query(cls, model: rest.Query) -> grpc.Query:

There the function receives a dense vector as model so any of the conditions is met.

[joein]

Hi @aiorga-sherpas
Thanks for pointing it out, we'll fix it asap

[joein]

#682

[joein]

fixes should be available with qdrant-client==1.10.1