I have a Milvus collection with both dense and sparse vector fields for text retrieval. I also want to store a list of preprocessed keywords for each text chunk, for example:

tags = ["milvus", "array", "filter", "hybrid-search"]

My schema has an ARRAY field for these keywords:

schema.add_field(     
field_name="tags",     
datatype=DataType.ARRAY,     
element_type=DataType.VARCHAR,     
max_capacity=20,     
max_length=512,     
nullable=True, )

I also created a scalar index on this field:

index_params.add_index(     
field_name="tags",     
index_type="AUTOINDEX",     
index_name="tags_index", )

Now I want to run a vector search with client.search() and only return records where the tags array contains a required subset of words. For example:

required_tags = ["milvus", "filter"]

I want to return only records where the tags field contains both "milvus" and "filter".

I tried this filter expression:

filter_expr = 'json_contains_all(tags, ["milvus", "filter"])'  
res = client.search(     
collection_name="my_collection",     
data=[query_vector],     
anns_field="dense_vector",     
limit=10,     
filter=filter_expr,     
output_fields=["tags", "text"], )

If I need to add more scalar conditions, I plan to build the expression like this:

and_filters = [     
'json_contains_all(tags, ["milvus", "filter"])',     
'source == "docs"', ]  
filter_expr = " AND ".join(and_filters)

I am using MilvusClient.search() for vector search.

The collection has a dense vector field named dense_vector.

Each entity also has a metadata field named tags.

The tags field is defined as ARRAY.

I created an AUTOINDEX scalar index on the tags field.

I want to combine vector search with an array filter such as json_contains_all(tags, [...]).

I want client.search() to return vector search results only when the tags ARRAY field contains all required words.I am not sure whether json_contains_all(tags, [...]) is the correct filter expression for an ARRAY field, or whether it is only valid for JSON fields.

What I checked：

The tags field is defined as DataType.ARRAY.

The array element type is DataType.VARCHAR.

The field has an AUTOINDEX scalar index.

I want to use the filter inside client.search(), not only in client.query().

I may need to combine this condition with other scalar filters using AND.

so, in Milvus client.search(), how should I write a filter expression to return vector search results where an ARRAY field contains all words from a required subset?