I had thought that num_entities would indicate the number of records (or whatever the correct term is) within a Milvus collection. However, I created 1 file – test_milvus.py to create a simple collection like so:

import numpy as np
from pymilvus import connections, Collection, CollectionSchema, FieldSchema, DataType

connections.connect(alias='default',host='localhost', port='19530')

# Define the schema
schema = CollectionSchema([FieldSchema("id", DataType.INT64, is_primary=True, max_length=100),
                           FieldSchema("vector", DataType.FLOAT_VECTOR, dim=2)])

# Create a collection
collection = Collection("test_collection", schema)

# Insert data
data = [{"id":i, "vector": np.array([i, i],dtype=np.float32)} for i in range(10)]
collection.insert(data)

# Flush data
collection.flush()

# Disconnect from the server
connections.disconnect(alias='default')

and another to get information on collections within a Milvus database = milvus_info.py – like so:

from pymilvus import Collection, connections, db, utility

def get_info (host: str = "localhost", port: str = "19530"):

    # Connect to Milvus (replace with your connection details)
    connections.connect(alias="default", host=host, port=port)  # Replace with your connection parameters

    # Print the list of databases and collections
    db_list = db.list_database()
    for db_name in db_list:
        print(f"Database: {db_name}")
        collection_list = utility.list_collections(using=db_name)
        if len(collection_list) == 0:
            print("  No collections")
        for collection_name in collection_list:
            print(f"  Collection: {collection_name}")
            temp_collection = Collection(name=collection_name)
            for info in temp_collection.describe():
                print(f"    {info}: {temp_collection.describe()[info]}")
            print(f"   Number of entities: {temp_collection.num_entities}")
          
    # Disconnect from Milvus
    connections.disconnect(alias='default')


if __name__ == "__main__":
    get_info()

The first time I ran test_milvus followed by milvus_info.py, I got this output:

  $ python test_milvus.py 
  $ python milvus_info.py 
    Database: default
      Collection: test_collection
        collection_name: test_collection
        auto_id: False
        num_shards: 1
        description: 
        fields: [{'field_id': 100, 'name': 'id', 'description': '', 'type': <DataType.INT64: 5>, 'params': {}, 'is_primary': True}, {'field_id': 101, 'name': 'vector', 'description': '', 'type': <DataType.FLOAT_VECTOR: 101>, 'params': {'dim': 2}}]
        aliases: []
        collection_id: 450687678279804785
        consistency_level: 2
        properties: {}
        num_partitions: 1
        enable_dynamic_field: False
       Number of entities: 10

which struck me as odd, because there were only 2 vectors in the db.

However, if I run `test_milvus.py’ again, the number of entities goes up to 20, even though no new vectors have been added:

$ python test_milvus.py 
$ python milvus_info.py 
Database: default
  Collection: test_collection
    collection_name: test_collection
    auto_id: False
    num_shards: 1
    description: 
    fields: [{'field_id': 100, 'name': 'id', 'description': '', 'type': <DataType.INT64: 5>, 'params': {}, 'is_primary': True}, {'field_id': 101, 'name': 'vector', 'description': '', 'type': <DataType.FLOAT_VECTOR: 101>, 'params': {'dim': 2}}]
    aliases: []
    collection_id: 450687678279804785
    consistency_level: 2
    properties: {}
    num_partitions: 1
    enable_dynamic_field: False
   Number of entities: 20

This happens even though I’ve only tried to add records that were already there. I would’ve expected num_entities to be 10, no matter how many time I run these files. The documentation says it returns the number of rows, but I can drive it arbitrarily high while still having only 10 rows. Is num_entities supposed to track all rows that ever existed???