Update app.py

app.py

@@ -10,6 +10,7 @@ from langchain.chains import ConversationalRetrievalChain
 from htmlTemplates import css, bot_template, user_template
 from langchain.llms import HuggingFaceHub, LlamaCpp
 from huggingface_hub import snapshot_download, hf_hub_download
+from prompts import CONDENSE_QUESTION_PROMPT
 
 repo_name = "IlyaGusev/saiga2_7b_gguf"
 model_name = "model-q2_K.gguf"
@@ -30,8 +31,8 @@ def get_pdf_text(pdf_docs):
 def get_text_chunks(text):
     
     text_splitter = CharacterTextSplitter(separator="\n",
-                                          chunk_size=1000,
-                                          chunk_overlap=200,
+                                          chunk_size=500, #1000
+                                          chunk_overlap=30, #200
                                           length_function=len
                                          )
     chunks = text_splitter.split_text(text)
@@ -51,12 +52,22 @@ def get_vectorstore(text_chunks):
 
 def get_conversation_chain(vectorstore, model_name):
 
-    llm = LlamaCpp(model_path=model_name, n_ctx=2048, n_parts=1)
+    llm = LlamaCpp(model_path=model_name,
+                   temperature=0.1,
+                   top_k=30,
+                   top_p=0.9,
+                   streaming=True,
+                   n_ctx=2048,
+                   n_parts=1,
+                   echo=True
+                  )
+    
     #llm = ChatOpenAI()
 
     memory = ConversationBufferMemory(memory_key='chat_history', return_messages=True)
     
     conversation_chain = ConversationalRetrievalChain.from_llm(llm=llm,
+                                                               condense_question_prompt=CONDENSE_QUESTION_PROMPT
                                                                retriever=vectorstore.as_retriever(),
                                                                memory=memory
                                                               )