Elasticsearch 向量搜索

目标记录

["你好，我的爱人","你好，我的爱妻","你好，我的病人","世界真美丽"]

搜索词

爱人

预期返回

["你好，我的爱人","你好，我的爱妻"]

示例代码：

代码连接 es8以及bge-large-zh模型，

bge-large-zh用来将文本转换为向量数据

es用来存储向量数据，并根据向量来搜索相似度最高的文本(相似度可以用阈值调整)

from flask import Flask, request  # 导入Flask类
from FlagEmbedding import FlagModel
from elasticsearch import Elasticsearch
from elasticsearch.helpers import bulk

app = Flask(__name__)  # 实例化并命名为app实例
model = FlagModel('./models/bge/bge-large-zh', query_instruction_for_retrieval="为这个句子生成表示以用于检索相关文章：")
# 创建Elasticsearch客户端对象
es = Elasticsearch(hosts="http://localhost:9200")
es.ping()


@app.route('/ins', methods=['POST'])
def index():    
    data = request.get_json()
    print(data)
    strs = data["strs"]
    documents = []
    for str in strs:            
        print(str)
        tmp = model.encode(str)
        documents.append({
            "general_text": str,
            "general_text_vector": tmp,
            # "domain":"001"
        })

    documents
    bulk(es, documents, index="demo")

    return success(1)

@app.route('/search', methods=['POST'])
def search():
    data = request.get_json()
    doc_vector = model.encode(data["name"])
    results = es.search(
        index="demo",
        source=[
            "general_text",            
        ],
        min_score= 1.83,
        query={
            "script_score": {
                "query": { "match_all": {} },
                "script": {
                    "source": "cosineSimilarity(params.queryVector, 'general_text_vector') + 1.0",
                    "params": {
                        "queryVector": doc_vector.tolist()
                    }
                }
            }
        },
        size=1000
    )
    # return results
    return results['hits']['hits']


def success(data):
    return {
        "status": "success",
        "result": data
    }


def fail(data):
    return {
        "status": "fail",
        "result": data
    }

if __name__ == "__main__":
    # Run
    port = 5000
    app.run(host='127.0.0.1', port=port, debug=False, use_reloader=False)

es存储数据