Update app.py

app.py

@@ -1,36 +1,29 @@
 import gradio as gr
 import os
-from langchain.vectorstores import FAISS
+from langchain.vectorstores.faiss import FAISS  # Direktimport
 from langchain.document_loaders import PyPDFLoader
 from langchain.embeddings import HuggingFaceEmbeddings
 from langchain.chains import ConversationalRetrievalChain
 from langchain.memory import ConversationBufferMemory
 from langchain.llms import HuggingFaceHub
 
-# Liste der Modelle
-list_llm = ["google/flan-t5-small", "distilbert-base-uncased"]  # Leichte Modelle für CPU
+list_llm = ["google/flan-t5-small", "distilbert-base-uncased"]
 list_llm_simple = [os.path.basename(llm) for llm in list_llm]
 
-# PDF-Dokument laden und aufteilen
 def load_doc(list_file_path):
     loaders = [PyPDFLoader(x) for x in list_file_path]
     pages = []
     for loader in loaders:
         pages.extend(loader.load())
-    text_splitter = RecursiveCharacterTextSplitter(
-        chunk_size=512,  # Kleinere Chunks für schnelleres Verarbeiten auf CPU
-        chunk_overlap=32
-    )
+    text_splitter = RecursiveCharacterTextSplitter(chunk_size=512, chunk_overlap=32)
     doc_splits = text_splitter.split_documents(pages)
     return doc_splits
 
-# Erstellen der Vektordatenbank
 def create_db(splits):
     embeddings = HuggingFaceEmbeddings()
     vectordb = FAISS.from_documents(splits, embeddings)
     return vectordb
 
-# Initialisierung des LLM Chains
 def initialize_llmchain(llm_model, temperature, max_tokens, top_k, vector_db):
     llm = HuggingFaceHub(
         repo_id=llm_model,
@@ -40,12 +33,7 @@ def initialize_llmchain(llm_model, temperature, max_tokens, top_k, vector_db):
             "top_k": top_k,
         }
     )
-
-    memory = ConversationBufferMemory(
-        memory_key="chat_history",
-        return_messages=True
-    )
-
+    memory = ConversationBufferMemory(memory_key="chat_history", return_messages=True)
     retriever = vector_db.as_retriever()
     qa_chain = ConversationalRetrievalChain.from_llm(
         llm,
@@ -57,14 +45,12 @@ def initialize_llmchain(llm_model, temperature, max_tokens, top_k, vector_db):
     )
     return qa_chain
 
-# Initialisierung der Datenbank
 def initialize_database(list_file_obj):
     list_file_path = [x.name for x in list_file_obj if x is not None]
     doc_splits = load_doc(list_file_path)
     vector_db = create_db(doc_splits)
     return vector_db, "Datenbank erfolgreich erstellt!"
 
-# Initialisierung des LLMs
 def initialize_LLM(llm_option, llm_temperature, max_tokens, top_k, vector_db):
     llm_name = list_llm[llm_option]
     qa_chain = initialize_llmchain(llm_name, llm_temperature, max_tokens, top_k, vector_db)
@@ -77,7 +63,6 @@ def format_chat_history(message, chat_history):
         formatted_chat_history.append(f"Assistant: {bot_message}")
     return formatted_chat_history
 
-# Chat-Funktion
 def conversation(qa_chain, message, history):
     formatted_chat_history = format_chat_history(message, history)
     response = qa_chain({"question": message, "chat_history": formatted_chat_history})
@@ -85,7 +70,6 @@ def conversation(qa_chain, message, history):
     new_history = history + [(message, response_answer)]
     return qa_chain, gr.update(value=""), new_history
 
-# Gradio App erstellen
 def demo():
     with gr.Blocks() as demo:
         vector_db = gr.State()