Pythonweaviate_search.py
Copy
1
import weaviate
2
import weaviate.classes as wvc
3
import random
4
 
5
DIM = 384
6
 
7
def random_vector(dim: int) -> list[float]:
8
    v = [random.gauss(0, 1) for _ in range(dim)]
9
    norm = sum(x * x for x in v) ** 0.5
10
    return [x / norm for x in v]
11
 
12
# Connect to local Weaviate
13
client = weaviate.connect_to_local()
14
 
15
try:
16
    # Create collection
17
    COLLECTION = "Article"
18
    if client.collections.exists(COLLECTION):
19
        client.collections.delete(COLLECTION)
20
 
21
    articles = client.collections.create(
22
        name=COLLECTION,
23
        vectorizer_config=wvc.config.Configure.Vectorizer.none(),  # We provide our own vectors
24
        properties=[
25
            wvc.config.Property(name="title", data_type=wvc.config.DataType.TEXT),
26
            wvc.config.Property(name="content", data_type=wvc.config.DataType.TEXT),
27
            wvc.config.Property(name="category", data_type=wvc.config.DataType.TEXT),
28
        ],
29
    )
30
 
31
    # Batch import
32
    categories = ["tech", "health", "finance", "sports"]
33
    with articles.batch.dynamic() as batch:
34
        for i in range(20):
35
            batch.add_object(
36
                properties={
37
                    "title": f"Article {i}: Understanding topic {i % 5}",
38
                    "content": f"This article discusses important aspects of topic {i % 5} in detail.",
39
                    "category": random.choice(categories),
40
                },
41
                vector=random_vector(DIM),
42
            )
43
    print(f"Imported {articles.aggregate.over_all().total_count} articles")
44
 
45
    # Near-vector search (semantic similarity)
46
    query_vec = random_vector(DIM)
47
    results = articles.query.near_vector(
48
        near_vector=query_vec,
49
        limit=3,
50
        return_properties=["title", "category"],
51
        return_metadata=wvc.query.MetadataQuery(distance=True),
52
    )
53
    print("\nNear-vector results:")
54
    for obj in results.objects:
55
        print(f"  {obj.properties['title']} (category: {obj.properties['category']}, dist: {obj.metadata.distance:.4f})")
56
 
57
    # Filtered search
58
    filtered_results = articles.query.near_vector(
59
        near_vector=query_vec,
60
        limit=3,
61
        filters=wvc.query.Filter.by_property("category").equal("tech"),
62
        return_properties=["title", "category"],
63
    )
64
    print("\nFiltered (tech only):")
65
    for obj in filtered_results.objects:
66
        print(f"  {obj.properties['title']}")
67
 
68
finally:
69
    client.close()
70