|
|
|
|
|
from WxMini.Milvus.Config.MulvusConfig import *
|
|
|
|
|
|
from WxMini.Milvus.Utils.MilvusCollectionManager import MilvusCollectionManager
|
|
|
|
|
|
from WxMini.Milvus.Utils.MilvusConnectionPool import *
|
|
|
|
|
|
from gensim.models import KeyedVectors
|
|
|
|
|
|
|
|
|
|
|
|
import jieba
|
|
|
|
|
|
import time
|
|
|
|
|
|
from openai import OpenAI
|
|
|
|
|
|
|
|
|
|
|
|
# 初始化 OpenAI 客户端
|
|
|
|
|
|
client = OpenAI(
|
|
|
|
|
|
api_key=MODEL_API_KEY,
|
|
|
|
|
|
base_url="https://dashscope.aliyuncs.com/compatible-mode/v1",
|
|
|
|
|
|
)
|
|
|
|
|
|
|
|
|
|
|
|
# 1. 加载预训练的 Word2Vec 模型
|
|
|
|
|
|
model_path = MS_MODEL_PATH
|
|
|
|
|
|
model = KeyedVectors.load_word2vec_format(model_path, binary=False, limit=MS_MODEL_LIMIT)
|
|
|
|
|
|
print(f"模型加载成功，词向量维度: {model.vector_size}")
|
|
|
|
|
|
|
|
|
|
|
|
# 功能：将文本转换为嵌入向量
|
|
|
|
|
|
def text_to_embedding(text):
|
|
|
|
|
|
words = jieba.lcut(text) # 使用 jieba 分词
|
|
|
|
|
|
print(f"文本: {text}, 分词结果: {words}")
|
|
|
|
|
|
embeddings = [model[word] for word in words if word in model]
|
|
|
|
|
|
print(f"有效词向量数量: {len(embeddings)}")
|
|
|
|
|
|
if embeddings:
|
|
|
|
|
|
avg_embedding = sum(embeddings) / len(embeddings)
|
|
|
|
|
|
print(f"生成的平均向量: {avg_embedding[:5]}...") # 打印前 5 维
|
|
|
|
|
|
return avg_embedding
|
|
|
|
|
|
else:
|
|
|
|
|
|
print("未找到有效词，返回零向量")
|
|
|
|
|
|
return [0.0] * model.vector_size
|
|
|
|
|
|
|
|
|
|
|
|
# 2. 使用连接池管理 Milvus 连接
|
|
|
|
|
|
milvus_pool = MilvusConnectionPool(host=MS_HOST, port=MS_PORT, max_connections=MS_MAX_CONNECTIONS)
|
|
|
|
|
|
|
|
|
|
|
|
# 3. 从连接池中获取一个连接
|
|
|
|
|
|
connection = milvus_pool.get_connection()
|
|
|
|
|
|
|
|
|
|
|
|
# 4. 初始化集合管理器
|
|
|
|
|
|
collection_name = MS_COLLECTION_NAME
|
|
|
|
|
|
collection_manager = MilvusCollectionManager(collection_name)
|
|
|
|
|
|
|
|
|
|
|
|
# 5. 输入一个用户问题
|
|
|
|
|
|
user_input = "我给你起的名字叫小云，记住你的名字，你是我的私人助理。"
|
|
|
|
|
|
|
|
|
|
|
|
# 6. 调用大模型生成回复
|
|
|
|
|
|
try:
|
|
|
|
|
|
response = client.chat.completions.create(
|
|
|
|
|
|
model=MODEL_NAME,
|
|
|
|
|
|
messages=[
|
|
|
|
|
|
{"role": "system", "content": "你是一个非常好的聊天伙伴，可以疏导用户，帮他解压,一句控制在20字以内。"},
|
|
|
|
|
|
{"role": "user", "content": user_input}
|
|
|
|
|
|
],
|
|
|
|
|
|
max_tokens=500
|
|
|
|
|
|
)
|
|
|
|
|
|
if response.choices and response.choices[0].message.content:
|
|
|
|
|
|
model_response = response.choices[0].message.content.strip()
|
|
|
|
|
|
else:
|
|
|
|
|
|
model_response = "大模型未返回有效结果"
|
|
|
|
|
|
except Exception as e:
|
|
|
|
|
|
model_response = f"调用大模型失败: {str(e)}"
|
|
|
|
|
|
|
|
|
|
|
|
print(f"大模型回复: {model_response}")
|
|
|
|
|
|
|
|
|
|
|
|
# 7. 获取当前时间和会话 ID
|
|
|
|
|
|
timestamp = time.strftime("%Y-%m-%d %H:%M:%S", time.localtime()) # 当前时间
|
|
|
|
|
|
person_id = "F9D1C319-215D-3B4E-FAA0-0E024AC12D3C" # 会话 ID（可以根据需要动态生成）
|
|
|
|
|
|
|
|
|
|
|
|
# 8. 将用户问题转换为嵌入向量
|
|
|
|
|
|
user_embedding = text_to_embedding(user_input)
|
|
|
|
|
|
|
|
|
|
|
|
# 9. 插入数据，确保字段顺序与集合定义一致
|
|
|
|
|
|
entities = [
|
|
|
|
|
|
[person_id], # person_id
|
|
|
|
|
|
[user_input], # user_input
|
|
|
|
|
|
[model_response], # model_response
|
|
|
|
|
|
[timestamp], # timestamp
|
|
|
|
|
|
[user_embedding] # embedding
|
|
|
|
|
|
]
|
|
|
|
|
|
collection_manager.insert_data(entities)
|
|
|
|
|
|
print("数据插入成功。")
|
|
|
|
|
|
|
|
|
|
|
|
# 10. 释放连接
|
|
|
|
|
|
milvus_pool.release_connection(connection)
|
|
|
|
|
|
|
|
|
|
|
|
# 11. 关闭连接池
|
|
|
|
|
|
milvus_pool.close()
|
