Update app.py

app.py

@@ -1,16 +1,10 @@
-
-import os
-os.environ["TOKENIZERS_PARALLELISM"] = "false"
-
 #####################################
-##  BitsAndBytes
+##   
 #####################################
 
 from transformers import AutoModelForCausalLM, AutoTokenizer, BitsAndBytesConfig
-#from langchain.llms import HuggingFaceHub
 from langchain_community.llms import HuggingFaceHub
 
-model_name = "bn22/Mistral-7B-Instruct-v0.1-sharded"
 
 ###### other models:
 # "Trelis/Llama-2-7b-chat-hf-sharded-bf16"
@@ -18,30 +12,13 @@ model_name = "bn22/Mistral-7B-Instruct-v0.1-sharded"
 # "HuggingFaceH4/zephyr-7b-beta"
 
 # function for loading 4-bit quantized model
-def load_model(model_name: str):
+def load_model( ):
 
     model =  HuggingFaceHub(
         repo_id="mistralai/Mixtral-8x7B-Instruct-v0.1",
         model_kwargs={"max_length": 1048, "temperature":0.2, "max_new_tokens":256, "top_p":0.95, "repetition_penalty":1.0},
     )
-
-    """
-    :param model_name: Name or path of the model to be loaded.
-    :return: Loaded quantized model.
     
-    bnb_config = BitsAndBytesConfig(
-        load_in_4bit=True,
-        bnb_4bit_use_double_quant=True,
-        bnb_4bit_quant_type="nf4",
-        bnb_4bit_compute_dtype=torch.bfloat16
-    )
-
-    model = AutoModelForCausalLM.from_pretrained(
-        model_name,
-        load_in_4bit=True,
-        torch_dtype=torch.bfloat16,
-        quantization_config=bnb_config
-    )"""
     return model
 
 ##################################################
@@ -55,8 +32,6 @@ from langchain_community.document_loaders import WebBaseLoader
 from langchain.text_splitter import RecursiveCharacterTextSplitter
 from langchain_community.vectorstores import Chroma
 
-#from langchain_openai import OpenAIEmbeddings, ChatOpenAI
-#from langchain.embeddings import HuggingFaceBgeEmbeddings
 from langchain_community.embeddings import HuggingFaceBgeEmbeddings
 from langchain.vectorstores.faiss import FAISS
 
@@ -69,6 +44,22 @@ from langchain.chains.combine_documents import create_stuff_documents_chain
 
 load_dotenv()
 
+def get_vectorstore():
+    ''' 
+        FAISS
+        A FAISS vector store containing the embeddings of the text chunks.
+   '''
+    model = "BAAI/bge-base-en-v1.5"
+    encode_kwargs = {
+        "normalize_embeddings": True
+    }  # set True to compute cosine similarity
+    embeddings = HuggingFaceBgeEmbeddings(
+        model_name=model, encode_kwargs=encode_kwargs, model_kwargs={"device": "cpu"}
+    )
+    # load from disk
+    vector_store = Chroma(persist_directory="/home/user/.cache/chroma_db", embedding_function=embeddings)
+    return vector_store
+
 def get_vectorstore_from_url(url):
     # get the text in document form
     loader = WebBaseLoader(url)
@@ -94,34 +85,12 @@ def get_vectorstore_from_url(url):
  
     #vectorstore = FAISS.from_texts(texts=text_chunks, embedding=embeddings)
     vector_store = Chroma.from_documents(document_chunks, embeddings, persist_directory="/home/user/.cache/chroma_db")
- 
-
-
-
-    #######
-    # create a vectorstore from the chunks
 
     return vector_store
 
+def get_context_retriever_chain(vector_store): 
 
-
-
-
-def get_context_retriever_chain(vector_store):
-
-    # specify model huggingface mode name
-    model_name = "anakin87/zephyr-7b-alpha-sharded"
-   # model_name = "bn22/Mistral-7B-Instruct-v0.1-sharded"
-
-    ###### other models:
-    # "Trelis/Llama-2-7b-chat-hf-sharded-bf16"
-    # "bn22/Mistral-7B-Instruct-v0.1-sharded"
-    # "HuggingFaceH4/zephyr-7b-beta"
-
-    # function for loading 4-bit quantized model
-     
-
-    llm = load_model(model_name)
+    llm = load_model( )
     
     retriever = vector_store.as_retriever()
     
@@ -137,7 +106,7 @@ def get_context_retriever_chain(vector_store):
     
 def get_conversational_rag_chain(retriever_chain): 
     
-    llm = load_model(model_name)
+    llm = load_model( )
     
     prompt = ChatPromptTemplate.from_messages([
       ("system", "Du bist eine freundlicher Mitarbeiterin Namens Susie und arbeitest in einenm Call Center. Du beantwortest basierend auf dem Context. Benutze nur den Inhalt des Context. Füge wenn möglich die Quelle hinzu. Antworte mit: Ich bin mir nicht sicher. Wenn die Antwort nicht aus dem Context hervorgeht. Antworte auf Deutsch, bitte? CONTEXT:\n\n{context}"),
@@ -149,44 +118,19 @@ def get_conversational_rag_chain(retriever_chain):
     
     return create_retrieval_chain(retriever_chain, stuff_documents_chain)
 
-#def get_response(user_input):
-#    retriever_chain = get_context_retriever_chain(st.session_state.vector_store)
-#    conversation_rag_chain = get_conversational_rag_chain(retriever_chain)
-#    
-#    response = conversation_rag_chain.invoke({
-#        "chat_history": st.session_state.chat_history,
-#        "input": user_query
-#    })
-    
-    return response
-
-
 
 ###################
 
 ###################
 import gradio as gr
 
-##from langchain_core.runnables.base import ChatPromptValue
-#from torch import tensor
 
-# Create Gradio interface
-#vector_store = None  # Set your vector store here
 chat_history = []     # Set your chat history here
 
 # Define your function here
 def get_response(user_input):
 
-      # Define the prompt as a ChatPromptValue object
-    #user_input = ChatPromptValue(user_input)
-    
-    # Convert the prompt to a tensor
-    #input_ids = user_input.tensor
-    
-
-    #vs = get_vectorstore_from_url(user_url, all_domain)
-    vs = get_vectorstore_from_url("https://huggingface.co/Chris4K")
-   # print("------ here 22 " )
+    vs = get_vectorstor()
     chat_history =[]
     retriever_chain = get_context_retriever_chain(vs)
     conversation_rag_chain = get_conversational_rag_chain(retriever_chain)
@@ -237,7 +181,7 @@ def get_response(message, history):
     
 
     #vs = get_vectorstore_from_url(user_url, all_domain)
-    vs = get_vectorstore_from_url("https://huggingface.co/Chris4K")
+    vs = get_vectorstore()
   
     history =[]
     retriever_chain = get_context_retriever_chain(vs)
@@ -254,11 +198,7 @@ def get_response(message, history):
     last_part = parts[-1]
     return last_part#[-1]['generation']['content']    
 
-    
-#####
-vs = get_vectorstore_from_url("https://www.xing.com/profile/Christof_Kaller/web_profiles")
-#vs = get_vectorstore_from_url("https://www.linkedin.com/in/christof-kaller-6b043733/?originalSubdomain=de")
-vs = get_vectorstore_from_url("https://twitter.com/zX14_7")
+
 
 
  
@@ -304,15 +244,7 @@ def get_all_links_from_domain(domain_url):
     get_links_from_page(domain_url, visited_urls, domain_links)
     return domain_links
 
-# Example usage:
-domain_url = 'https://globl.contact/'
-links = get_all_links_from_domain(domain_url)
-print("Links from the domain:", links)
 
-#########
-# Assuming visited_urls is a list of URLs
-for url in links:
-    vs = get_vectorstore_from_url(url)
 
 
 
@@ -332,4 +264,15 @@ app = gr.ChatInterface(
     clear_btn=None 
 )
 
-app.launch(debug=True, share=True) 
+app.launch(debug=True, share=True) 
+
+def __init__(self):
+    # Example usage:
+    domain_url = 'https://globl.contact/'
+    links = get_all_links_from_domain(domain_url)
+    print("Links from the domain:", links)
+    
+    #########
+    # Assuming visited_urls is a list of URLs
+    for url in links:
+        vs = get_vectorstore_from_url(url)