Hi guys! I am staring to use Qdrant as VectorStore to do some performance test, and I wanted to take advantage of the filtering capabilities using llama-index abstraction
MetadataFilters
. However, I faced an error which leads me to suspect there might be a bug in the piece of code that translates these abstractions to Qdrant's implementation.
I am trying to build a filter with nested OR like (A OR B) AND C. However when building this filter:
MetadataFilters(
filters=[
MetadataFilters(
filters=[
MetadataFilter(
key="A",
operator=FilterOperator.EQ,
value=username,
),
MetadataFilter(
key="B",
operator=FilterOperator.EQ,
value=role,
),
],
condition=FilterCondition.OR,
),
MetadataFilter(
key="C", operator=FilterOperator.NE, value=username
),
],
condition=FilterCondition.AND,
)