1

pip install -U pymilvus

1
2
3

from pymilvus import MilvusClient

client = MilvusClient("milvus_demo.db")

1
2

collection_name = "milvus_demo"
client.create_collection(collection_name, dimension=768)

1
2
3

collection_name = "milvus_demo"
if not client.has_collection(collection_name):
    client.create_collection(collection_name, dimension=768)

1
2
3
4
5
6
7
8

from pymilvus import model

embedding_fn = model.DefaultEmbeddingFunction()
docs = ["hello world", "hi", "你好世界", "milvus is great"]

vectors = embedding_fn.encode_documents(docs)

print ("向量数据的维度:", len(vectors[0]))

1
2
3
4
5
6
7

data = [
    {"id": i, "vector": vectors[i], "text": docs[i]}
    for i in range(len(vectors))
]

print("数据包含", len(data), "个条目，每个条目包含字段: ", data[0].keys())
print("vector 维度:", len(data[0]["vector"]))

1
2

response = client.insert(collection_name, data)
print(response)

1
2
3

query = "hi"
query_vectors = embedding_fn.encode_queries([query])

1
2
3
4
5
6
7
8

query_response = client.search(
    collection_name,
    data=query_vectors,  
    limit=2,  
    output_fields=["id", "text"],  
)
print(query_response)

1
2
3
4
5

data: ["[
          {'id': 1, 'distance': 1.0, 'entity': {'text': 'hi', 'id': 1}},
          {'id': 0, 'distance': 0.5280660390853882, 'entity': {'text': 'hello world', 'id': 0}},
          {'id': 2, 'distance': 0.44701865315437317, 'entity': {'text': '你好世界', 'id': 2}}
      ]"]

1

client.delete(collection_name, ids=[1])

1

client.delete(collection_name, filter="id == 1")

1

client.drop_collection(collection_name)