I realized that when I tried to search it returns nothing, how to solve this?

[JenuelDev]

So I have this function, that creates a milvus collection

def __generate_collection(self):
        if self.collection_name not in self.milvusClient.list_collections():
            schema = MilvusClient.create_schema(auto_id=True, enable_dynamic_field=True, primary_field="id")

            schema.add_field(field_name="id", datatype=DataType.INT64, is_primary=True)
            schema.add_field(field_name="source", datatype=DataType.VARCHAR, max_length=50000)
            schema.add_field(field_name="embeddings", datatype=DataType.FLOAT_VECTOR, dim=self.truncate_dim)

            index_params = MilvusClient.prepare_index_params()

            index_params.add_index(field_name="id", index_type="STL_SORT")
            index_params.add_index(
                field_name="embeddings",
                index_type="IVF_FLAT",
                metric_type="L2",
                params={"nlist": 1024},
            )

            self.milvusClient.create_collection(
                collection_name=self.collection_name, schema=schema, index_params=index_params
            )

I also inerted data.. using this:

entities = [
    {"source": "blah blah vlah", "embeddings": [1,2,3,4, etc..]}
]
insert_result = self.milvusClient.insert(collection_name=self.collection_name, data=entities)
print(insert_result)

#outputs {'insert_count': 1, 'ids': [450065725884255501], 'cost': 0}

As you can see it inserted something,, but when I try to search. It does not give me anything

def search_and_query(
        self,
        search_vectors: list,
        search_field: str = "embeddings",
        search_params={"metric_type": "L2", "params": {"nprobe": 10}},
        vector_search_limit=10,
    ):
        result = self.milvusClient.search(
            collection_name=self.collection_name,
            data=search_vectors,
            search_params=search_params,
            anns_field=search_field,
            limit=vector_search_limit,
            output_fields=["source"],
        )
        
        print(result)

        return result

# output: data: ['[]'] , extra_info: {'cost': 0}

I need help, or did I did something wrong here? do I need to await something?

[yhmo]

Data visibility is controlled by consistency_level: https://milvus.io/docs/consistency.md#Consistency

Add consistency_level="Strong" to search() and you will see the result:

from pymilvus import (
    DataType,

    MilvusClient,
)

milvusClient = MilvusClient("http://localhost:19530")

dim = 128
collection_name = "XXX"
schema = MilvusClient.create_schema(auto_id=True, enable_dynamic_field=True, primary_field="id")

schema.add_field(field_name="id", datatype=DataType.INT64, is_primary=True)
schema.add_field(field_name="source", datatype=DataType.VARCHAR, max_length=50000)
schema.add_field(field_name="embeddings", datatype=DataType.FLOAT_VECTOR, dim=dim)

index_params = MilvusClient.prepare_index_params()

index_params.add_index(field_name="id", index_type="STL_SORT")
index_params.add_index(
    field_name="embeddings",
    index_type="IVF_FLAT",
    metric_type="L2",
    params={"nlist": 1024},
)

milvusClient.create_collection(
    collection_name=collection_name, schema=schema, index_params=index_params
)

entities = [
    {"source": "blah blah vlah", "embeddings": [0.5 for _ in range(dim)]}
]
insert_result = milvusClient.insert(collection_name=collection_name, data=entities)
print(insert_result)

search_params={"metric_type": "L2", "params": {"nprobe": 10}}
search_vectors = [[0.5 for _ in range(dim)]]
result = milvusClient.search(
    collection_name=collection_name,
    data=search_vectors,
    search_params=search_params,
    anns_field="embeddings",
    limit=10,
    output_fields=["source"],
    consistency_level="Strong", # comment out this line, search result could be empty
)
print(result)

milvusClient.drop_collection(collection_name=collection_name)