Update app.py

app.py

@@ -10,6 +10,7 @@ from langchain.llms import HuggingFacePipeline
 from langchain.chains import ConversationChain
 from langchain.memory import ConversationBufferMemory
 from langchain.llms import HuggingFaceHub
+from langchain.memory import ConversationTokenBufferMemory
 
 from pathlib import Path
 import chromadb
@@ -127,24 +128,13 @@ def initialize_llmchain(temperature, max_tokens, top_k, vector_db, progress=gr.P
                                                           "load_in_8bit": True})
     
     progress(0.75, desc="Defining buffer memory...")
-    memory = ConversationBufferMemory(
-        memory_key="chat_history",
-        output_key='answer',
-        return_messages=True
-    )
+    #memory = ConversationBufferMemory(memory_key="chat_history",output_key='answer',return_messages=True)
+    memory = ConversationTokenBufferMemory(llm = llm, max_token_limit=100)
     # retriever=vector_db.as_retriever(search_type="similarity", search_kwargs={'k': 3})
     retriever=vector_db.as_retriever()
     progress(0.8, desc="Defining retrieval chain...")
-    qa_chain = ConversationalRetrievalChain.from_llm(
-        llm,
-        retriever=retriever,
-        chain_type="stuff", 
-        memory=memory,
-        # combine_docs_chain_kwargs={"prompt": your_prompt})
-        return_source_documents=True,
-        #return_generated_question=False,
-        verbose=False,
-    )
+    qa_chain = ConversationalRetrievalChain.from_llm(llm,retriever=retriever,chain_type="stuff", 
+                                                     memory=memory,return_source_documents=True,verbose=False)
     progress(0.9, desc="Done!")
     return qa_chain
 
@@ -269,7 +259,7 @@ def demo():
             with gr.Row():
                 slider_temperature = gr.Slider(value = 0.1,visible=False)
             with gr.Row():
-                slider_maxtokens = gr.Slider(value = 1000, visible=False)
+                slider_maxtokens = gr.Slider(value = 4000, visible=False)
             with gr.Row():
                 slider_topk = gr.Slider(value = 3, visible=False)