{"id":365,"date":"2026-09-14T19:55:20","date_gmt":"2026-09-14T11:55:20","guid":{"rendered":"https:\/\/flycloud.io\/?p=365"},"modified":"2026-10-07T19:56:51","modified_gmt":"2026-10-07T11:56:51","slug":"milvus-%e9%80%9a%e4%b9%89%e5%8d%83%e9%97%ae%ef%bc%9a%e5%bf%ab%e9%80%9f%e6%9e%84%e5%bb%ba%e4%b8%93%e5%b1%9e%e7%9f%a5%e8%af%86%e5%ba%93%e9%97%ae%e7%ad%94%e7%b3%bb%e7%bb%9f","status":"publish","type":"post","link":"https:\/\/flycloud.io\/en\/2026\/09\/14\/milvus-%e9%80%9a%e4%b9%89%e5%8d%83%e9%97%ae%ef%bc%9a%e5%bf%ab%e9%80%9f%e6%9e%84%e5%bb%ba%e4%b8%93%e5%b1%9e%e7%9f%a5%e8%af%86%e5%ba%93%e9%97%ae%e7%ad%94%e7%b3%bb%e7%bb%9f\/","title":{"rendered":"Milvus + \u901a\u4e49\u5343\u95ee\uff1a\u5feb\u901f\u6784\u5efa\u4e13\u5c5e\u77e5\u8bc6\u5e93\u95ee\u7b54\u7cfb\u7edf"},"content":{"rendered":"<p id=\"b101247bceqec\">\u672c\u6587\u5c55\u793a\u4e86\u5982\u4f55\u4f7f\u7528\u5411\u91cf\u68c0\u7d22\u670d\u52a1Milvus\u7248\uff08\u7b80\u79f0Milvus\uff09\u548c\u963f\u91cc\u4e91\u767e\u70bc\u63d0\u4f9b\u7684\u5343\u95ee\u5927\u6a21\u578b\u80fd\u529b\uff0c\u5feb\u901f\u6784\u5efa\u4e00\u4e2a\u57fa\u4e8e\u4e13\u5c5e\u77e5\u8bc6\u5e93\u7684\u95ee\u7b54\u7cfb\u7edf\u3002\u5728\u793a\u4f8b\u4e2d\uff0c\u6211\u4eec\u901a\u8fc7\u63a5\u5165\u963f\u91cc\u4e91\u767e\u70bc\u63d0\u4f9b\u7684\u5343\u95eeAPI\u53ca\u6587\u672c\u5d4c\u5165\uff08Embedding\uff09API\u6765\u5b9e\u73b0LLM\u5927\u6a21\u578b\u7684\u76f8\u5173\u529f\u80fd\u3002<\/p>\n\n\n\n<h2 class=\"wp-block-heading\" id=\"7f8597da0eg4c\"><strong>\u524d\u63d0\u6761\u4ef6<\/strong><\/h2>\n\n\n\n<ul id=\"baf119d5a2xlp\" class=\"wp-block-list\">\n<li>\u5df2\u521b\u5efaMilvus\u5b9e\u4f8b\u3002\u5177\u4f53\u64cd\u4f5c\uff0c\u8bf7\u53c2\u89c1<a href=\"https:\/\/help.aliyun.com\/zh\/milvus\/getting-started\/quickly-create-a-milvus-instance\">\u5feb\u901f\u521b\u5efaMilvus\u5b9e\u4f8b<\/a>\u3002<\/li>\n\n\n\n<li>\u5df2\u5f00\u901a\u963f\u91cc\u4e91\u767e\u70bc\u670d\u52a1\u5e76\u83b7\u5f97API-KEY\u3002\u5177\u4f53\u64cd\u4f5c\uff0c\u8bf7\u53c2\u89c1<a href=\"https:\/\/help.aliyun.com\/zh\/model-studio\/get-api-key\">\u83b7\u53d6\u4e0e\u914d\u7f6e API Key<\/a>\u3002<\/li>\n<\/ul>\n\n\n\n<h2 class=\"wp-block-heading\" id=\"ebbd114b60c5y\"><strong>\u4f7f\u7528\u9650\u5236<\/strong><\/h2>\n\n\n\n<p id=\"bbee8d2823x6l\">\u8bf7\u786e\u4fdd\u60a8\u7684\u8fd0\u884c\u73af\u5883\u4e2d\u5df2\u5b89\u88c5Python 3.8\u6216\u4ee5\u4e0a\u7248\u672c\uff0c\u4ee5\u4fbf\u987a\u5229\u5b89\u88c5\u5e76\u4f7f\u7528DashScope\u3002<\/p>\n\n\n\n<h2 class=\"wp-block-heading\" id=\"f3ba0966d1044\"><strong>\u64cd\u4f5c\u6d41\u7a0b<\/strong><\/h2>\n\n\n\n<h3 class=\"wp-block-heading\" id=\"4f74c4257d2ln\"><strong>\u51c6\u5907\u5de5\u4f5c<\/strong><\/h3>\n\n\n\n<ol id=\"9868def1a1da6\" class=\"wp-block-list\">\n<li>\u5b89\u88c5\u76f8\u5173\u7684\u4f9d\u8d56\u5e93\u3002 <code>pip3 install pymilvus tqdm dashscope<\/code><\/li>\n\n\n\n<li>\u4e0b\u8f7d\u6240\u9700\u7684\u77e5\u8bc6\u5e93\u3002\u672c\u6587\u793a\u4f8b\u4f7f\u7528\u4e86\u516c\u5f00\u6570\u636e\u96c6CEC-Corpus\u3002<a class=\"\" href=\"https:\/\/github.com\/shijiebei2009\/CEC-Corpus\">CEC-Corpus<\/a>\u6570\u636e\u96c6\u5305\u542b332\u7bc7\u9488\u5bf9\u5404\u7c7b\u7a81\u53d1\u4e8b\u4ef6\u7684\u65b0\u95fb\u62a5\u9053\uff0c\u8bed\u6599\u548c\u6807\u6ce8\u6570\u636e\uff0c\u8fd9\u91cc\u6211\u4eec\u53ea\u9700\u8981\u63d0\u53d6\u539f\u59cb\u7684\u65b0\u95fb\u7a3f\u6587\u672c\uff0c\u5e76\u5c06\u5176\u5411\u91cf\u5316\u540e\u5165\u5e93\u3002 <code>git clone https:\/\/github.com\/shijiebei2009\/CEC-Corpus.git<\/code><\/li>\n<\/ol>\n\n\n\n<h3 class=\"wp-block-heading\" id=\"104fea22d93cn\"><strong>\u6b65\u9aa4\u4e00\uff1a<\/strong>\u77e5\u8bc6\u5e93\u5411\u91cf\u5316<\/h3>\n\n\n\n<ol class=\"wp-block-list\">\n<li>\u521b\u5efa<code>embedding.py<\/code>\u6587\u4ef6\uff0c\u5185\u5bb9\u5982\u4e0b\u6240\u793a\u3002<\/li>\n<\/ol>\n\n\n\n<pre id=\"84z0ao\" class=\"wp-block-code\"><code>import os\nimport time\nfrom tqdm import tqdm\nimport dashscope\nfrom dashscope import TextEmbedding\nfrom pymilvus import connections, FieldSchema, CollectionSchema, DataType, Collection, utility\n\ndef prepareData(path, batch_size=25):\n    batch_docs = &#91;]\n    for file in os.listdir(path):\n        with open(path + '\/' + file, 'r', encoding='utf-8') as f:\n            batch_docs.append(f.read())\n            if len(batch_docs) == batch_size:\n                yield batch_docs\n                batch_docs = &#91;]\n\n    if batch_docs:\n        yield batch_docs\n\ndef getEmbedding(news):\n    model = TextEmbedding.call(\n        model=TextEmbedding.Models.text_embedding_v1,\n        input=news\n    )\n    embeddings = &#91;record&#91;'embedding'] for record in model.output&#91;'embeddings']]\n    return embeddings if isinstance(news, list) else embeddings&#91;0]\n\nif __name__ == '__main__':\n\n    current_path = os.path.abspath(os.path.dirname(__file__))   # \u5f53\u524d\u76ee\u5f55\n    root_path = os.path.abspath(os.path.join(current_path, '..'))   # \u4e0a\u7ea7\u76ee\u5f55\n    data_path = f'{root_path}\/CEC-Corpus\/raw corpus\/allSourceText'  # \u6570\u636e\u4e0b\u8f7dgit clone https:\/\/github.com\/shijiebei2009\/CEC-Corpus.git\n\n    # \u914d\u7f6eDashscope API KEY\n    dashscope.api_key = '&lt;YOUR_DASHSCOPE_API_KEY>'\n\n    # \u914d\u7f6eMilvus\u53c2\u6570\n    COLLECTION_NAME = 'CEC_Corpus'\n    DIMENSION = 1536\n    MILVUS_HOST = 'c-97a7d8038fb8****.milvus.aliyuncs.com'\n    MILVUS_PORT = '19530'\n    USER = 'root'\n    PASSWORD = '&lt;password>'\n\n    connections.connect(host=MILVUS_HOST, port=MILVUS_PORT, user=USER, password=PASSWORD)\n\n    # Remove collection if it already exists\n    if utility.has_collection(COLLECTION_NAME):\n        utility.drop_collection(COLLECTION_NAME)\n\n    # Create collection which includes the id, title, and embedding.\n    fields = &#91;\n        FieldSchema(name='id', dtype=DataType.INT64, descrition='Ids', is_primary=True, auto_id=False),\n        FieldSchema(name='text', dtype=DataType.VARCHAR, description='Text', max_length=4096),\n        FieldSchema(name='embedding', dtype=DataType.FLOAT_VECTOR, description='Embedding vectors', dim=DIMENSION)\n    ]\n    schema = CollectionSchema(fields=fields, description='CEC Corpus Collection')\n    collection = Collection(name=COLLECTION_NAME, schema=schema)\n\n    # Create an index for the collection.\n    index_params = {\n        'index_type': 'IVF_FLAT',\n        'metric_type': 'L2',\n        'params': {'nlist': 1024}\n    }\n    collection.create_index(field_name=\"embedding\", index_params=index_params)\n\n    id = 0\n    for news in tqdm(list(prepareData(data_path))):\n        ids = &#91;id + i for i, _ in enumerate(news)]\n        id += len(news)\n\n        vectors = getEmbedding(news)\n        # insert Milvus Collection\n        for id, vector, doc in zip(ids, vectors, news):\n            insert_doc = (doc&#91;:498] + '..') if len(doc) > 500 else doc\n            ins = &#91;&#91;id], &#91;insert_doc], &#91;vector]]  # Insert the title id, the text, and the text embedding vector\n            collection.insert(ins)\n            time.sleep(2)<\/code><\/pre>\n\n\n\n<p id=\"0c323577d6f83\">\u5728\u672c\u6587\u793a\u4f8b\u4e2d\uff0c\u6211\u4eec\u5c06Embedding\u5411\u91cf\u548c\u65b0\u95fb\u62a5\u9053\u6587\u7a3f\u4e00\u8d77\u5b58\u5165Milvus\u4e2d\uff0c\u540c\u65f6\u6784\u5efa\u7d22\u5f15\u7c7b\u578b\u91c7\u7528\u4e86IVF_FLAT\uff0c\u5728\u5411\u91cf\u68c0\u7d22\u65f6\uff0c\u540c\u65f6\u53ef\u4ee5\u53ec\u56de\u539f\u59cb\u6587\u7a3f\u3002<\/p>\n\n\n\n<h3 class=\"wp-block-heading\" id=\"7d0064bd7e8ei\"><strong>\u6b65\u9aa4\u4e8c\uff1a<\/strong>\u5411\u91cf\u68c0\u7d22\u4e0e\u77e5\u8bc6\u95ee\u7b54<\/h3>\n\n\n\n<p id=\"68ff561500mm0\">\u6570\u636e\u5199\u5165\u5b8c\u6210\u540e\uff0c\u5373\u53ef\u8fdb\u884c\u5feb\u901f\u7684\u5411\u91cf\u68c0\u7d22\u3002\u5728\u901a\u8fc7\u63d0\u95ee\u641c\u7d22\u5230\u76f8\u5173\u7684\u77e5\u8bc6\u70b9\u540e\uff0c\u6211\u4eec\u53ef\u4ee5\u6309\u7167\u7279\u5b9a\u7684\u6a21\u677f\u5c06\u201c\u63d0\u95ee + \u77e5\u8bc6\u70b9\u201d\u4f5c\u4e3aprompt\u5411LLM\u53d1\u8d77\u63d0\u95ee\u3002\u5728\u8fd9\u91cc\u6211\u4eec\u6240\u4f7f\u7528\u7684LLM\u662f\u5343\u95ee\uff0c\u8fd9\u662f\u963f\u91cc\u5df4\u5df4\u81ea\u4e3b\u7814\u53d1\u7684\u8d85\u5927\u89c4\u6a21\u8bed\u8a00\u6a21\u578b\uff0c\u80fd\u591f\u5728\u7528\u6237\u81ea\u7136\u8bed\u8a00\u8f93\u5165\u7684\u57fa\u7840\u4e0a\uff0c\u901a\u8fc7\u81ea\u7136\u8bed\u8a00\u7406\u89e3\u548c\u8bed\u4e49\u5206\u6790\uff0c\u7406\u89e3\u7528\u6237\u610f\u56fe\u3002\u901a\u8fc7\u63d0\u4f9b\u5c3d\u53ef\u80fd\u6e05\u6670\u8be6\u7ec6\u7684\u6307\u4ee4\uff08prompt)\uff0c\u53ef\u4ee5\u83b7\u5f97\u66f4\u7b26\u5408\u9884\u671f\u7684\u7ed3\u679c\u3002\u8fd9\u4e9b\u80fd\u529b\u90fd\u53ef\u4ee5\u901a\u8fc7\u5343\u95ee\u6765\u83b7\u5f97\u3002<\/p>\n\n\n\n<p id=\"83a100af98r1k\">\u672c\u6587\u793a\u4f8b\u8bbe\u8ba1\u7684\u63d0\u95ee\u6a21\u677f\u683c\u5f0f\u4e3a\uff1a<code><em>\u8bf7\u57fa\u4e8e\u6211\u63d0\u4f9b\u7684\u5185\u5bb9\u56de\u7b54\u95ee\u9898\u3002\u5185\u5bb9\u662f{___}\uff0c\u6211\u7684\u95ee\u9898\u662f{___}<\/em><\/code>\uff0c\u5f53\u7136\u60a8\u4e5f\u53ef\u4ee5\u81ea\u884c\u8bbe\u8ba1\u5408\u9002\u7684\u6a21\u677f\u3002<\/p>\n\n\n\n<p id=\"4ec628bc22z8d\">\u521b\u5efa<code>answer.py<\/code>\u6587\u4ef6\uff0c\u5185\u5bb9\u5982\u4e0b\u6240\u793a\u3002<\/p>\n\n\n\n<pre id=\"11kisl\" class=\"wp-block-code\"><code>import os\nimport dashscope\nfrom dashscope import Generation\nfrom pymilvus import connections, FieldSchema, CollectionSchema, DataType, Collection\nfrom dashscope import TextEmbedding\n\ndef getEmbedding(news):\n    model = TextEmbedding.call(\n        model=TextEmbedding.Models.text_embedding_v1,\n        input=news\n    )\n    embeddings = &#91;record&#91;'embedding'] for record in model.output&#91;'embeddings']]\n    return embeddings if isinstance(news, list) else embeddings&#91;0]\n\ndef getAnswer(query, context):\n    prompt = f'''\u8bf7\u57fa\u4e8e```\u5185\u7684\u62a5\u9053\u5185\u5bb9\uff0c\u56de\u7b54\u6211\u7684\u95ee\u9898\u3002\n\t      ```\n\t      {context}\n\t      ```\n\t      \u6211\u7684\u95ee\u9898\u662f\uff1a{query}\u3002\n    '''\n\n    rsp = Generation.call(model='qwen-turbo', prompt=prompt)\n    return rsp.output.text\n\ndef search(text):\n    # Search parameters for the index\n    search_params = {\n        \"metric_type\": \"L2\"\n    }\n\n    results = collection.search(\n        data=&#91;getEmbedding(text)],  # Embeded search value\n        anns_field=\"embedding\",  # Search across embeddings\n        param=search_params,\n        limit=1,  # Limit to five results per search\n        output_fields=&#91;'text']  # Include title field in result\n    )\n\n    ret = &#91;]\n    for hit in results&#91;0]:\n        ret.append(hit.entity.get('text'))\n    return ret\n\nif __name__ == '__main__':\n\n    current_path = os.path.abspath(os.path.dirname(__file__))   # \u5f53\u524d\u76ee\u5f55\n    root_path = os.path.abspath(os.path.join(current_path, '..'))   # \u4e0a\u7ea7\u76ee\u5f55\n    data_path = f'{root_path}\/CEC-Corpus\/raw corpus\/allSourceText'\n\n    # \u914d\u7f6eDashscope API KEY\n    dashscope.api_key = '&lt;YOUR_DASHSCOPE_API_KEY&gt;'\n\n    # \u914d\u7f6eMilvus\u53c2\u6570\n    COLLECTION_NAME = 'CEC_Corpus'\n    DIMENSION = 1536\n    MILVUS_HOST = 'c-97a7d8038fb8****.milvus.aliyuncs.com'\n    MILVUS_PORT = '19530'\n    USER = 'root'\n    PASSWORD = '&lt;password&gt;'\n\n    connections.connect(host=MILVUS_HOST, port=MILVUS_PORT, user=USER, password=PASSWORD)\n\n    fields = &#91;\n        FieldSchema(name='id', dtype=DataType.INT64, descrition='Ids', is_primary=True, auto_id=False),\n        FieldSchema(name='text', dtype=DataType.VARCHAR, description='Text', max_length=4096),\n        FieldSchema(name='embedding', dtype=DataType.FLOAT_VECTOR, description='Embedding vectors', dim=DIMENSION)\n    ]\n    schema = CollectionSchema(fields=fields, description='CEC Corpus Collection')\n    collection = Collection(name=COLLECTION_NAME, schema=schema)\n\n    # Load the collection into memory for searching\n    collection.load()\n\n    question = '\u5317\u4eac\u4e2d\u592e\u7535\u89c6\u53f0\u5de5\u5730\u53d1\u751f\u5927\u706b\uff0c\u53d1\u751f\u5728\u54ea\u91cc\uff1f\u51fa\u52a8\u4e86\u591a\u5c11\u8f86\u6d88\u9632\u8f66\uff1f\u4eba\u5458\u4f24\u4ea1\u60c5\u51b5\u5982\u4f55\uff1f'\n    context = search(question)\n    answer = getAnswer(question, context)\n    print(answer)\n<\/code><\/pre>\n\n\n\n<p id=\"dfe5d2df6412y\">\u8fd0\u884c\u5b8c\u6210\u540e\uff0c\u9488\u5bf9<code>\u5317\u4eac\u4e2d\u592e\u7535\u89c6\u53f0\u5de5\u5730\u53d1\u751f\u5927\u706b\uff0c\u53d1\u751f\u5728\u54ea\u91cc\uff1f\u51fa\u52a8\u4e86\u591a\u5c11\u8f86\u6d88\u9632\u8f66\uff1f\u4eba\u5458\u4f24\u4ea1\u60c5\u51b5\u5982\u4f55\uff1f<\/code>\u7684\u63d0\u95ee\uff0c\u4f1a\u5f97\u5230\u4ee5\u4e0b\u7ed3\u679c\u3002<\/p>\n\n\n\n<p id=\"a61c7a4b64rge\"><code>\u706b\u707e\u53d1\u751f\u5728\u5317\u4eac\u5e02\u671d\u9633\u533a\u4e1c\u4e09\u73af\u4e2d\u592e\u7535\u89c6\u53f0\u65b0\u5740\u56ed\u533a\u5728\u5efa\u7684\u9644\u5c5e\u6587\u5316\u4e2d\u5fc3\u5927\u697c\u5de5\u5730\u3002\u51fa\u52a8\u4e8654\u8f86\u6d88\u9632\u8f66\u3002\u76ee\u524d\u5c1a\u65e0\u4eba\u5458\u4f24\u4ea1\u62a5\u544a\u3002<\/code><\/p>","protected":false},"excerpt":{"rendered":"<p>\u672c\u6587\u5c55\u793a\u4e86\u5982\u4f55\u4f7f\u7528\u5411\u91cf\u68c0\u7d22\u670d\u52a1Milvus\u7248\uff08\u7b80\u79f0Milvus\uff09\u548c\u963f\u91cc\u4e91\u767e\u70bc\u63d0\u4f9b\u7684\u5343\u95ee\u5927\u6a21\u578b\u80fd\u529b\uff0c\u5feb\u901f\u6784\u5efa\u4e00\u4e2a\u57fa\u4e8e\u4e13\u5c5e\u77e5\u8bc6\u5e93\u7684\u95ee\u7b54\u7cfb\u7edf\u3002\u5728\u793a\u4f8b\u4e2d\uff0c\u6211\u4eec\u901a\u8fc7\u63a5\u5165\u963f\u91cc\u4e91\u767e\u70bc\u63d0\u4f9b\u7684\u5343\u95eeAPI\u53ca\u6587\u672c\u5d4c\u5165\uff08Embedding\uff09API\u6765\u5b9e\u73b0LLM\u5927\u6a21\u578b\u7684\u76f8\u5173\u529f\u80fd\u3002 \u524d\u63d0\u6761\u4ef6 \u4f7f\u7528\u9650\u5236 \u8bf7\u786e\u4fdd\u60a8\u7684\u8fd0\u884c\u73af\u5883\u4e2d\u5df2\u5b89\u88c5Python 3.8\u6216\u4ee5\u4e0a\u7248\u672c\uff0c\u4ee5\u4fbf\u987a\u5229\u5b89\u88c5\u5e76\u4f7f\u7528DashScope\u3002 \u64cd\u4f5c\u6d41\u7a0b \u51c6\u5907\u5de5\u4f5c \u6b65\u9aa4\u4e00\uff1a&hellip;<\/p>","protected":false},"author":1,"featured_media":0,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[5,10],"tags":[],"class_list":["post-365","post","type-post","status-publish","format-standard","hentry","category-ai","category-alibaba-cloud"],"_links":{"self":[{"href":"https:\/\/flycloud.io\/en\/wp-json\/wp\/v2\/posts\/365","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/flycloud.io\/en\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/flycloud.io\/en\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/flycloud.io\/en\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/flycloud.io\/en\/wp-json\/wp\/v2\/comments?post=365"}],"version-history":[{"count":1,"href":"https:\/\/flycloud.io\/en\/wp-json\/wp\/v2\/posts\/365\/revisions"}],"predecessor-version":[{"id":366,"href":"https:\/\/flycloud.io\/en\/wp-json\/wp\/v2\/posts\/365\/revisions\/366"}],"wp:attachment":[{"href":"https:\/\/flycloud.io\/en\/wp-json\/wp\/v2\/media?parent=365"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/flycloud.io\/en\/wp-json\/wp\/v2\/categories?post=365"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/flycloud.io\/en\/wp-json\/wp\/v2\/tags?post=365"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}