Update app.py

app.py

@@ -101,6 +101,19 @@ def initialize_database(list_file_obj):
     return vector_db, collection_name
 
 
+def initialize_llmchain(vector_db):
+    # Initialize langchain LLM chain
+    llm = HuggingFaceHub(repo_id = llm_model,model_kwargs={"temperature": temperature, 
+                                                           "max_new_tokens": max_tokens, 
+                                                           "top_k": top_k, 
+                                                           "load_in_8bit": True})
+    retriever=vector_db.as_retriever()
+    memory = ConversationBufferMemory(memory_key="chat_history", output_key='answer', return_messages=True)
+    qa_chain = ConversationalRetrievalChain.from_llm(llm,retriever=retriever,chain_type="stuff", 
+                                                         memory=memory,return_source_documents=True,verbose=False,)
+        
+    return qa_chain
+
 def initialize_LLM(vector_db):
     # print("llm_option",llm_option)
     llm_name = llm_model
@@ -146,27 +159,13 @@ def conversation(qa_chain, message, history):
 
 
 
-
 def demo():
     with gr.Blocks(theme='base') as demo:
         #vector_db = gr.State()
         #qa_chain = gr.State()
-        #collection_name = gr.State()
-
-
-        vector_db, collection_name = initialize_database(list_file_obj)
-
-        # Initialize langchain LLM chain
-        llm = HuggingFaceHub(repo_id = llm_model,
-                            model_kwargs={"temperature": temperature, 
-                                          "max_new_tokens": max_tokens, 
-                                          "top_k": top_k, 
-                                          "load_in_8bit": True})
-        retriever=vector_db.as_retriever()
-        memory = ConversationBufferMemory(memory_key="chat_history", output_key='answer', return_messages=True)
-        qa_chain = ConversationalRetrievalChain.from_llm(llm,retriever=retriever,chain_type="stuff", 
-                                                         memory=memory,return_source_documents=True,verbose=False,)
+        #collection_name = gr.State()   
         
+        vector_db, collection_name = initialize_database(list_file_obj)
         chatbot = gr.Chatbot(height=300)
         with gr.Accordion('References', open=True):
             with gr.Row():
@@ -181,9 +180,17 @@ def demo():
         with gr.Row():
             msg = gr.Textbox(placeholder = 'Ask your question', container = True)
         with gr.Row():
+            qa_chain_button = gr.Button('Start Chatbot')
             submit_btn = gr.Button('Submit')
             clear_button = gr.ClearButton([msg, chatbot])
 
+        qa_chain_button.click(initialize_LLM, \
+            inputs=[vector_db], \
+            outputs=[qa_chain]).then(lambda:[None,"",0,"",0,"",0], \
+            inputs=None, \
+            outputs=[chatbot, doc_source1, source1_page, doc_source2, source2_page, doc_source3, source3_page], \
+            queue=False)
+        
         msg.submit(conversation, \
             inputs=[qa_chain, msg, chatbot], \
             outputs=[qa_chain, msg, chatbot, doc_source1, source1_page, doc_source2, source2_page, doc_source3, source3_page], \