Update app.py

app.py

@@ -2,6 +2,7 @@ import streamlit as st
 import io
 import fitz
 import requests
+from streamlit_chat import message
 from langchain.llms import LlamaCpp
 from langchain.callbacks.base import BaseCallbackHandler
 from langchain.docstore.document import Document
@@ -65,7 +66,7 @@ def create_chain(_retriever):
     # responses in real time.
     # callback_manager = CallbackManager([stream_handler])
     (repo_id, model_file_name) = ("TheBloke/Mistral-7B-Instruct-v0.1-GGUF",
-                                  "mistral-7b-instruct-v0.1.Q5_0.gguf")
+                                  "mistral-7b-instruct-v0.1.Q4_K_M.gguf")
 
     model_path = hf_hub_download(repo_id=repo_id,
                                  filename=model_file_name,
@@ -77,11 +78,12 @@ def create_chain(_retriever):
     llm = LlamaCpp(
             model_path=model_path,
             n_batch=n_batch,
-            n_ctx=2048,
+            n_ctx=4096,
             max_tokens=2048,
-            temperature=0,
+            temperature=.33,
             # callback_manager=callback_manager,
-            verbose=False,
+            top_p=1, 
+            verbose=True,
             streaming=True,
             )
 
@@ -102,85 +104,48 @@ def create_chain(_retriever):
     return qa_chain
 
 
-# Set the webpage title
-st.set_page_config(
-    page_title="Youtube Aesop Product Reviewer"
-)
 
-# Create a header element
-st.header("Youtube Aesop Product Reviewer")
+def initialize_session_state():
+    if 'history' not in st.session_state:
+        st.session_state['history'] = []
 
-#
-system_prompt = st.text_area(
-    label="System Prompt",
-    value="""
-    DOCUMENT:
-    (document text)
+    if 'generated' not in st.session_state:
+        st.session_state['generated'] = ["Hi, I know what Youtubers said about Aesop's products. Ask me!"]
 
-    QUESTION:
-    (users question)
-    
-    INSTRUCTIONS:
-    Answer the users QUESTION using the DOCUMENT text above.
-    Keep your answer ground in the facts of the DOCUMENT.
-    If the DOCUMENT doesn’t contain the facts to answer the QUESTION return {NONE}""",
-    key="system_prompt"
-)
+    if 'past' not in st.session_state:
+        st.session_state['past'] = ["Hey! 👋"]
 
-if "base_url" not in st.session_state:
-    st.session_state.base_url = ""
+def conversation_chat(query, chain, history):
+    result = chain({"question": query, "chat_history": history})
+    history.append((query, result["answer"]))
+    return result["answer"]
 
-base_url = st.text_input("Enter url for data here. For Aesop product review copy & paste: https://raw.githubusercontent.com/grantjw/product_chatbot_rag/main/data/output_transcripts.csv", key="base_url")
+def display_chat_history(chain):
+    reply_container = st.container()
+    container = st.container()
 
-if st.session_state.base_url != "":
-    
-    retriever = get_retriever(base_url)
-
-    # We store the conversation in the session state.
-    # This will be used to render the chat conversation.
-    # We initialize it with the first message we want to be greeted with.
-    if "messages" not in st.session_state:
-        st.session_state.messages = [
-            {"role": "assistant", "content": "How may I help you today?"}
-        ]
-
-    if "current_response" not in st.session_state:
-        st.session_state.current_response = ""
-
-    # We loop through each message in the session state and render it as
-    # a chat message.
-    for message in st.session_state.messages:
-        with st.chat_message(message["role"]):
-            st.markdown(message["content"])
-
-    # We initialize the quantized LLM from a local path.
-    # Currently most parameters are fixed but we can make them
-    # configurable.
-    llm_chain = create_chain(retriever)
-
-    # We take questions/instructions from the chat input to pass to the LLM
-    if user_prompt := st.chat_input("Your message here", key="user_input"):
-
-        # Add our input to the session state
-        st.session_state.messages.append(
-            {"role": "user", "content": user_prompt}
-        )
-
-        # Add our input to the chat window
-        with st.chat_message("user"):
-            st.markdown(user_prompt)
-
-        # Pass our input to the llm chain and capture the final responses.
-        # It is worth noting that the Stream Handler is already receiving the
-        # streaming response as the llm is generating. We get our response
-        # here once the llm has finished generating the complete response.
-        response = llm_chain.run(user_prompt)
-
-        # Add the response to the session state
-        st.session_state.messages.append(
-            {"role": "assistant", "content": response}
-        )
-
-        # Add the response to the chat window
-        with st.chat_message("assistant"):
-            st.markdown(response)
+    with container:
+        with st.form(key='my_form', clear_on_submit=True):
+            user_input = st.text_input("Question:", placeholder="      ", key='input')
+            submit_button = st.form_submit_button(label='Send')
+
+        if submit_button and user_input:
+            with st.spinner('Generating response...'):
+                output = conversation_chat(user_input, chain, st.session_state['history'])
+
+            st.session_state['past'].append(user_input)
+            st.session_state['generated'].append(output)
+
+    if st.session_state['generated']:
+        with reply_container:
+            for i in range(len(st.session_state['generated'])):
+                message(st.session_state["past"][i], is_user=True, key=str(i) + '_user', avatar_style="thumbs")
+                message(st.session_state["generated"][i], key=str(i), avatar_style="shapes")
+
+base_url = "https://raw.githubusercontent.com/grantjw/product_chatbot_rag/main/data/output_transcripts.csv"
+retriever = get_retriever(base_url)
+llm_chain = create_chain(retriever)
+initialize_session_state()
+st.title("Aesop Product Reviewer from YouTube Reviews")
+st.image("aesop.png", width=550)
+display_chat_history(llm_chain)