milvus_lite_local_vector_collection_insert_and_search.py

1from pymilvus import MilvusClient
2import numpy as np
3
4# 1. Set up a local Milvus Lite instance
5# This will create a local file named "milvus_demo.db" in the current directory
6client = MilvusClient("milvus_demo.db")
7
8# 2. Create a collection
9# We use "auto_id" and let Milvus handle the schema for simplicity in this quickstart
10if client.has_collection(collection_name="demo_collection"):
11    client.drop_collection(collection_name="demo_collection")
12
13client.create_collection(
14    collection_name="demo_collection",
15    dimension=5  # The vectors in this example have 5 dimensions
16)
17
18# 3. Prepare data
19# Representing data as a list of dictionaries
20data = [
21    {"id": i, "vector": np.random.random(5).tolist(), "text": f"data_{i}"}
22    for i in range(10)
23]
24
25# 4. Insert data
26res = client.insert(
27    collection_name="demo_collection",
28    data=data
29)
30
31print(f"Successfully inserted {res['insert_count']} entities.")
32
33# 5. Search
34# Define a query vector
35query_vectors = [np.random.random(5).tolist()]
36
37# Execute search
38results = client.search(
39    collection_name="demo_collection",
40    data=query_vectors,
41    limit=3,  # Return top 3 results
42    output_fields=["text"]  # Specify which fields to return
43)
44
45# 6. Display results
46for hits in results:
47    for hit in hits:
48        print(f"Hit: {hit}")