Upload 2 files

langchain_KB.py

@@ -18,15 +18,9 @@ import pathlib
 from pathlib import Path
 import re
 from re import sub
-import matplotlib.pyplot as plt
-from itertools import product
-from tqdm import tqdm_notebook, tqdm, trange
 import time
 from time import sleep
 # import pretty_errors
-import seaborn as sns
-from matplotlib.pyplot import style
-from rich import print
 import warnings
 import PyPDF2
 from openai import OpenAI

rag_reponse_002.py

@@ -12,18 +12,14 @@ from langchain_core.runnables import RunnableParallel
 import streamlit as st
 import re
 import openai
-from openai import OpenAI
 import os
 from langchain.llms.base import LLM
 from langchain.llms.utils import enforce_stop_tokens
 from typing import Dict, List, Optional, Tuple, Union
-import requests
-import json
 # import chatgpt
 import qwen_response
 from dotenv import load_dotenv
 import dashscope
-# dashscope.api_key = "sk-948adb3e65414e55961a9ad9d22d186b"
 
 load_dotenv()
 ### 设置openai的API key
@@ -35,11 +31,10 @@ dashscope.api_key = os.environ['dashscope_api_key']
 
 from langchain.embeddings.openai import OpenAIEmbeddings
 
-embeddings = HuggingFaceEmbeddings(model_name='GanymedeNil/text2vec-large-chinese') ## 这里是联网情况下，部署在Huggingface上后使用。
-
+# embeddings = HuggingFaceEmbeddings(model_name='GanymedeNil/text2vec-large-chinese') ## 这里是联网情况下，部署在Huggingface上后使用。
 # embeddings = OpenAIEmbeddings(disallowed_special=())  ## 这里是联网情况下，部署在Huggingface上后使用。
 # embeddings = HuggingFaceEmbeddings(model_name='/Users/yunshi/Downloads/360Data/Data Center/Working-On Task/演讲与培训/2023ChatGPT/Coding/RAG/bge-large-zh') ## 切换成BGE的embedding。
-vector_store = FAISS.load_local("./faiss_index/", embeddings=embeddings, allow_dangerous_deserialization=True) ## 加载vector store到本地。
+# vector_store = FAISS.load_local("./faiss_index/", embeddings=embeddings, allow_dangerous_deserialization=True) ## 加载vector store到本地。
 # vector_store = FAISS.load_local("./faiss_index/", embeddings=embeddings) ## 加载vector store到本地。 ### original code here.
 
 # ## 配置ChatGLM的类与后端api server对应。
@@ -120,8 +115,13 @@ def rag_source(docs):
     print('source:', source)
     return source
 
-def rag_response(user_input, k=3):
+def rag_response(username, user_input, k=3):
     # docs = vector_store.similarity_search('user_input', k=k) ## Original。
+    
+    embeddings = HuggingFaceEmbeddings(model_name='BAAI/bge-large-zh-v1.5') ## 这里是联网情况下，部署在Huggingface上后使用。
+    # embeddings = HuggingFaceEmbeddings(model_name='GanymedeNil/text2vec-large-chinese') ## 这里是联网情况下，部署在Huggingface上后使用。
+    print('embeddings:', embeddings)
+    vector_store = FAISS.load_local(f"./{username}/faiss_index/", embeddings=embeddings, allow_dangerous_deserialization=True) ## 加载vector store到本地。
     docs = vector_store.similarity_search(user_input, k=k) ##TODO 'user_input' to user_input？
     context = [doc.page_content for doc in docs]
     # print('context: {}'.format(context))