Update app.py

app.py

@@ -8,52 +8,43 @@ from langchain_community.llms import HuggingFaceHub
 from langchain.memory import ConversationBufferMemory
 
 def make_vectorstore(embeddings):
-    # use glob to find all the pdf files in the data folder in the base directory
+    
     loader = PyPDFDirectoryLoader("data")   
-
-    # load the documents
     documents = loader.load()
-
-    # split the documents into chunks of 1400 characters with 0 overlap
     text_splitter = CharacterTextSplitter(chunk_size=1400, chunk_overlap=0)
-
-    # split the documents into chunks of 1400 characters with 0 overlap
     texts = text_splitter.split_documents(documents)
-
-    # create a vector store from the documents
     docsearch = FAISS.from_documents(texts, embeddings)
     
     return docsearch
 
-def get_conversation(vectorstore):
+def get_qa(vectorstore, llmb):
 
-    # create a memory object to store the conversation history
-    memory = ConversationBufferMemory(memory_key="chat_history",return_messages=True,)
-
-    conversation_chain = ConversationalRetrievalChain.from_chain_type(
-        llm=HuggingFaceHub(repo_id="google/flan-t5-xxl", model_kwargs={"temperature":0.5, "max_length":512}, huggingfacehub_api_token = st.secrets["hf_token"]),
+    qa = RetrievalQA.from_chain_type(
+        llm=llmb,
         chain_type="stuff", 
-        retriever=vectorstore.as_retriever(),
-        memory=memory)
+        retriever=vectorstore.as_retriever())
     
-    return conversation_chain
+    return qa
 
-def get_response(conversation_chain, query):
-    # get the response
-    response = conversation_chain.run(query)
+def get_response(qa, query):
+    
+    response = qa.run(query)
+    
     return response
 
 def main():
+    
     st.title("BetterZila RAG Enabled LLM")
+    llm = HuggingFaceHub(repo_id="google/flan-t5-xxl", model_kwargs={"temperature":0.5, "max_length":512}, huggingfacehub_api_token = st.secrets["hf_token"])
     embeddings = HuggingFaceInstructEmbeddings(model_name="google/t5-v1_1-xl", model_kwargs = {'device': 'cpu'})
     vectorstore = make_vectorstore(embeddings)
-    conversation_chain = get_conversation(vectorstore)
+    qa = get_qa(vectorstore, llm)
     queries = ["Can you give me an example from history where the enemy was crushed totally from the book?", "What's the point of making myself less accessible?", "Can you tell me the story of Queen Elizabeth I from this 48 laws of power book?"]
     for query in queries:
         st.subheader(f"Query: {query}")
-        response = get_response(conversation_chain, query)
+        response = get_response(qa, query)
         st.write(query)
-        st.write(response["llm_response"])
+        st.write(response)
     st.success("Responses generated!")
     
 if __name__ == "__main__":