第24篇：向量数据库的未来展望与技术挑战的应对

    etcd:
      image: quay.io/coreos/etcd
      imageTag: v3.4.13
      replicaCount: 3
    
    minio:
      image: minio/minio
      imageTag: RELEASE.2020-12-03T00-03-10Z
      replicaCount: 1
    
    pulsar:
      image: apachepulsar/pulsar
      imageTag: 2.7.0
      replicaCount: 1
    
    milvus:
      image: milvusdb/milvus
      imageTag: latest
      replicaCount: 3
    
    
    yaml
![](https://ad.itadn.com/c/weblog/blog-img/images/2025-08-17/ucwiGFShkqOYRfZnW8b3C1vrEzUs.png)

    from transformers import BertModel, BertTokenizer
    from tensorflow.keras.applications.vgg16 import VGG16, preprocess_input
    import numpy as np
    import torch
    import cv2
    
    # 加载BERT模型和分词器
    bert_model = BertModel.from_pretrained('bert-base-uncased')
    tokenizer = BertTokenizer.from_pretrained('bert-base-uncased')
    
    # 加载VGG16模型
    vgg_model = VGG16(weights='imagenet', include_top=False, input_shape=(224, 224, 3))
    
    # 文本向量化
    def text_to_vector(text):
    inputs = tokenizer(text, return_tensors='pt', padding=True, truncation=True, max_length=128)
    outputs = bert_model(**inputs)
    vector = outputs.last_hidden_state.mean(dim=1).detach().numpy()
    return vector
    
    # 图像向量化
    def image_to_vector(image_path):
    image = cv2.imread(image_path)
    image = cv2.resize(image, (224, 224))
    image = preprocess_input(image)
    image = np.expand_dims(image, axis=0)
    vector = vgg_model.predict(image).flatten()
    return vector
    
    # 示例文本和图像
    text = "This is a sample text."
    image_path = "path_to_image.jpg"
    
    text_vector = text_to_vector(text)
    image_vector = image_to_vector(image_path)
    
    
    python
    
    
![](https://ad.itadn.com/c/weblog/blog-img/images/2025-08-17/fxe4tikZdG7pNgUOwbsEW6JL8uYK.png)

    from pulsar import Client
    
    # 连接到Pulsar集群
    client = Client('pulsar://localhost:6650')
    
    # 创建生产者
    producer = client.create_producer
    ```python
    producer = client.create_producer('persistent://public/default/milvus-logs')
    
    # 创建消费者
    consumer = client.subscribe('persistent://public/default/milvus-logs', subscription_name='milvus-subscription')
    
    # 记录操作日志
    def log_operation(operation):
    producer.send(operation.encode('utf-8'))
    
    # 恢复操作日志
    def recover_operations():
    while True:
        msg = consumer.receive()
        try:
            operation = msg.data().decode('utf-8')
            print(f"Recovered operation: {operation}")
            consumer.acknowledge(msg)
        except:
            consumer.negative_acknowledge(msg)
    
    # 示例操作日志记录
    log_operation("INSERT INTO collection VALUES (1, [0.1, 0.2, 0.3])")
    log_operation("DELETE FROM collection WHERE id = 1")
    
    # 恢复操作日志
    recover_operations()
    
    # 关闭客户端
    client.close()
    
    
    python
    
    
![](https://ad.itadn.com/c/weblog/blog-img/images/2025-08-17/MZrKaOWG6poh9kIidRfemcw0LV2u.png)

    from pymilvus import Collection, CollectionSchema, FieldSchema, DataType, connections
    
    # 连接到Milvus服务器
    connections.connect("default", host="localhost", port="19530")
    
    # 定义Milvus集合的字段
    fields = [
    FieldSchema(name="vector", dtype=DataType.FLOAT_VECTOR, dim=128),
    FieldSchema(name="id", dtype=DataType.INT64, is_primary=True)
    ]
    
    # 创建Milvus集合
    schema = CollectionSchema(fields, "example_collection")
    collection = Collection("example_collection", schema)
    
    # 插入示例数据
    import numpy as np
    
    data = {
    "vector": [np.random.rand(128).tolist() for _ in range(1000)],
    "id": list(range(1000))
    }
    collection.insert(data)
    
    # 创建IVF索引
    index_params = {"index_type": "IVF_FLAT", "params": {"nlist": 100}}
    collection.create_index("vector", index_params)
    
    # 进行相似性搜索
    query_vector = np.random.rand(128).tolist()
    search_params = {"metric_type": "L2", "params": {"nprobe": 10}}
    results = collection.search([query_vector], "vector", search_params, limit=5)
    
    # 输出搜索结果
    for result in results[0]:
    print(f"ID: {result.id}, Distance: {result.distance}")
    
    
    python
    
    
![](https://ad.itadn.com/c/weblog/blog-img/images/2025-08-17/yqhOrSWu9lCE510fLbGcpoUZVIm3.png)