Find answers from the community
s
F
Y
a
P
search_batch request to Qdrant would be different than the results coming from using a llamaindex retriever that is pointed at the same Qdrant collection? Some overlap in the results but I would have assumed an exact match. No filters being applied during the search.response = self._client.search(
collection_name=self.collection_name,
query_vector=query_embedding,
limit=query.similarity_top_k,
query_filter=query_filter,
)VectorStoreQueryMode.HYBRID), in which case I believe it is using search_batch; although, I would still expect there to be no differences, unless i'm missing something with the hybrid option enabled in the collection..response = self._client.search_batch(
collection_name=self.collection_name,
requests=[
rest.SearchRequest(
vector=rest.NamedVector(
name="text-dense",
vector=query_embedding,
),
limit=query.similarity_top_k,
filter=query_filter,
with_payload=True,
),
],
)retriever.retrieve, correct? I am passing the exact same string to both the retriever and the Qdrant client and getting slightly different results (some overlap of retrieved nodes, albeit with different scores, but also some that don't match); I am not passing a filter to Qdrant and am wondering if that could be the cause somehow, if it is defaulting to something when making the retrieve call?