Milvus 使用文档

1 安装

# 安装 Docker Compose
curl -SL https://github.com/docker/compose/releases/download/v2.30.3/docker-compose-linux-x86_64 -o /usr/local/bin/docker-compose
sudo chmod +x /usr/local/bin/docker-compose
sudo ln -s /usr/local/bin/docker-compose /usr/bin/docker-compose

# 下载 Milvus 配置文件
wget https://github.com/milvus-io/milvus/releases/download/v2.6.9/milvus-standalone-docker-compose.yml -O docker-compose.yml

# 启动容器
sudo docker compose up -d

# 验证容器状态
Creating milvus-etcd  ... done
Creating milvus-minio ... done
Creating milvus-standalone ... done

# 添加密码功能
# 修改配置文件添加：
common:
  security:
    authorizationEnabled: true

# Python 客户端连接示例
from pymilvus import MilvusClient
client = MilvusClient(
    uri='http://localhost:19530',
    token="root:Milvus"
)

2 使用案例

2.1 文档搜索（RAG）

import dashscope
from dashscope import TextEmbedding
from pymilvus import MilvusClient

dashscope.api_key = ''
client = MilvusClient("http://127.0.0.1:19530")

def create_collection(collection_name):
    if client.has_collection(collection_name):
        client.drop_collection(collection_name)
    client.create_collection(
        collection_name=collection_name,
        dimension=1024,
        metric_type="IP",
        consistency_level="Bounded"
    )

def emb_text(text):
    return TextEmbedding.call(
        model="text-embedding-v4",
        input=text,
        dimension=1024
    ).output['embeddings']

def insert_data(collection_name, data):
    return client.insert(collection_name=collection_name, data=data)

def search_data(collection_name):
    query_vectors = emb_text(["深度学习"])[0]['embedding']
    res = client.search(
        collection_name=collection_name,
        data=[query_vectors],
        limit=2,
        output_fields=["text", "subject"]
    )
    print(res)

# 使用示例
if __name__ == '__main__':
    create_collection("demo_collection")
    documents = [
        "人工智能是计算机科学的一个分支",
        "机器学习是实现人工智能的重要方法",
        "深度学习是机器学习的一个子领域"
    ]
    data = [
        {"id": i, "vector": emb_text(documents)[i]['embedding'], 
         "text": documents[i], "subject": "demo"}
        for i in range(len(documents))
    ]
    insert_data("demo_collection", data)
    search_data("demo_collection")

2.2 图片检索（以图搜图）

import base64
import os
import dashscope
from pymilvus import MilvusClient

def image_to_base64(image_path):
    with open(image_path, "rb") as f:
        return f"data:image/png;base64,{base64.b64encode(f.read()).decode('utf-8')}"

def emb_image(input):
    return dashscope.MultiModalEmbedding.call(
        model="tongyi-embedding-vision-plus",
        input=input
    ).output['embeddings'][0]['embedding']

# 使用示例
if __name__ == '__main__':
    client = MilvusClient("http://localhost:19530")
    client.create_collection(
        collection_name="image_embeddings",
        dimension=1152,
        metric_type="IP"
    )
    
    # 插入图片数据
    for file in os.listdir("data"):
        if file.endswith(".png"):
            embedding = emb_image([{"image": image_to_base64(f"data/{file}")}])
            client.insert("image_embeddings", {
                "vector": embedding,
                "filename": file
            })
    
    # 搜索示例
    query_embedding = emb_image([{"image": image_to_base64("data/柯基1.png")}])
    results = client.search(
        collection_name="image_embeddings",
        data=[query_embedding],
        limit=2,
        output_fields=["filename"]
    )
    print(results)

2.3 以文搜图

def emb_multimodal(input):
    return dashscope.MultiModalEmbedding.call(
        model="qwen3-vl-embedding",
        input=input,
        dimension=2560
    ).output['embeddings'][0]['embedding']

# 使用示例
if __name__ == '__main__':
    # 文本搜索
    text_embedding = emb_multimodal(["柯基犬"])
    results = client.search(
        collection_name="image_embeddings",
        data=[text_embedding],
        limit=2,
        output_fields=["filename"]
    )
    print(results)

3 集成

后续更新计划，可参考官方集成文档：
https://docs.llamaindex.org.cn/en/stable/examples/vector_stores/MilvusIndexDemo/

Milvus 提供强大的数据建模功能，支持多种数据类型和属性模型，包括数字、字符、向量、数组、集合和JSON等。

Milvus 使用文档

目录

1 安装

2 使用案例

2.1 文档搜索（RAG）

2.2 图片检索（以图搜图）

2.3 以文搜图

3 集成