Update app.py

app.py

@@ -1,6 +1,6 @@
 import gradio as gr
 import os
-from langchain.vectorstores import Chroma  # Chroma als Vektordatenbank
+from langchain.vectorstores import SimpleVectorStore  # Direkt, ohne zusätzliche Abhängigkeiten
 from langchain.document_loaders import PyPDFLoader
 from langchain.embeddings import HuggingFaceEmbeddings
 from langchain.chains import ConversationalRetrievalChain
@@ -20,8 +20,7 @@ def load_doc(list_file_path):
     return doc_splits
 
 def create_db(splits):
-    embeddings = HuggingFaceEmbeddings()
-    vectordb = Chroma.from_documents(splits, embeddings)
+    vectordb = SimpleVectorStore.from_documents(splits)  # Speichern im Speicher
     return vectordb
 
 def initialize_llmchain(llm_model, temperature, max_tokens, top_k, vector_db):
@@ -30,7 +29,6 @@ def initialize_llmchain(llm_model, temperature, max_tokens, top_k, vector_db):
         model_kwargs={
             "temperature": temperature,
             "max_length": max_tokens,
-            "top_k": top_k,
         }
     )
     memory = ConversationBufferMemory(memory_key="chat_history", return_messages=True)
@@ -41,7 +39,6 @@ def initialize_llmchain(llm_model, temperature, max_tokens, top_k, vector_db):
         chain_type="stuff",
         memory=memory,
         return_source_documents=True,
-        verbose=False
     )
     return qa_chain
 
@@ -51,10 +48,10 @@ def initialize_database(list_file_obj):
     vector_db = create_db(doc_splits)
     return vector_db, "Datenbank erfolgreich erstellt!"
 
-def initialize_LLM(llm_option, llm_temperature, max_tokens, top_k, vector_db):
+def initialize_LLM(llm_option, llm_temperature, max_tokens, vector_db):
     llm_name = list_llm[llm_option]
-    qa_chain = initialize_llmchain(llm_name, llm_temperature, max_tokens, top_k, vector_db)
-    return qa_chain, "LLM erfolgreich initialisiert! Chatbot ist bereit."
+    qa_chain = initialize_llmchain(llm_name, llm_temperature, max_tokens, 3, vector_db)
+    return qa_chain, "Chatbot ist bereit."
 
 def conversation(qa_chain, message, history):
     formatted_chat_history = [(f"User: {m}", f"Assistant: {r}") for m, r in history]
@@ -67,10 +64,10 @@ def demo():
     with gr.Blocks() as demo:
         vector_db = gr.State()
         qa_chain = gr.State()
-        gr.HTML("<center><h1>RAG PDF Chatbot (Kostenlose Version)</h1></center>")
+        gr.HTML("<center><h1>PDF QA Chatbot (Kostenlose Version)</h1></center>")
         with gr.Row():
             with gr.Column():
-                document = gr.Files(height=300, file_count="multiple", file_types=[".pdf"], interactive=True)
+                document = gr.Files(file_types=[".pdf"], interactive=True)
                 db_btn = gr.Button("Erstelle Vektordatenbank")
                 db_progress = gr.Textbox(value="Nicht initialisiert", show_label=False)
                 llm_btn = gr.Radio(["Flan-T5-Small", "MiniLM"], label="Verfügbare Modelle")
@@ -79,7 +76,7 @@ def demo():
                 qachain_btn = gr.Button("Initialisiere QA-Chatbot")
 
             with gr.Column():
-                chatbot = gr.Chatbot(height=400, type="messages")
+                chatbot = gr.Chatbot(height=400)
                 msg = gr.Textbox(placeholder="Frage stellen...")
                 submit_btn = gr.Button("Absenden")