app.py

import streamlit as st
import langchain_core
from langchain_core.messages import AIMessage, HumanMessage
from langchain_community.document_loaders import WebBaseLoader
from langchain.text_splitter import RecursiveCharacterTextSplitter
from langchain_community.vectorstores import Chroma
# from langchain_openai import OpenAIEmbeddings, ChatOpenAI
from langchain_core.prompts import ChatPromptTemplate, MessagesPlaceholder
from langchain.chains import create_history_aware_retriever, create_retrieval_chain
from langchain.chains.combine_documents import create_stuff_documents_chain
from langchain_community.embeddings import HuggingFaceBgeEmbeddings
from langchain_community.llms import CTransformers
from ctransformers import AutoModelForCausalLM
from langchain.llms import HuggingFaceHub
from transformers import AutoModelForCausalLM, AutoTokenizer
from langchain.embeddings import HuggingFaceEmbeddings
from langchain.llms import HuggingFacePipeline
from transformers import pipeline
import os
import transformers
import torch
from langchain_community.llms import LlamaCpp
# from langchain_retrieval import BaseRetrieverChain
# from dotenv import load_dotenv

# load_dotenv()



def get_vector_store_from_url(url):
    # model_name = "BAAI/bge-large-en"
    # model_kwargs = {'device': 'cpu'}
    # encode_kwargs = {'normalize_embeddings': False}
    # embeddings = HuggingFaceBgeEmbeddings(
    #     model_name=model_name,
    #     model_kwargs=model_kwargs,
    #     encode_kwargs=encode_kwargs
    # )
    embeddings = HuggingFaceEmbeddings(model_name='thenlper/gte-large',
                                   model_kwargs={'device': 'cpu'})
    
    loader = WebBaseLoader(url)
    document = loader.load()
    
    # split the document into chunks
    text_splitter = RecursiveCharacterTextSplitter()
    document_chunks = text_splitter.split_documents(document)
    
    # create a vectorstore from the chunks
    # vector_store = Chroma.from_documents(document_chunks, OpenAIEmbeddings())
    vector_store = Chroma.from_documents(document_chunks, embeddings)

    return vector_store
    

def get_context_retriever_chain(vector_store,llm):
    # llm = ChatOpenAI()
    llm = llm
    retriever = vector_store.as_retriever()
    
    prompt = ChatPromptTemplate.from_messages([
      MessagesPlaceholder(variable_name="chat_history"),
      ("user", "{input}"),
      ("user", "Given the above conversation, generate a search query to look up in order to get information relevant to the conversation")
    ])
    
    retriever_chain = create_history_aware_retriever(llm, retriever, prompt)
    
    return retriever_chain
    

# def get_conversational_rag_chain(retriever_chain,llm): 
    
#     llm=llm
    
#     template = "Answer the user's questions based on the below context:\n\n{context}"
#     human_template = "{input}"
    
#     prompt = ChatPromptTemplate.from_messages([
#       ("system", template),
#       MessagesPlaceholder(variable_name="chat_history"),
#       ("user", human_template),
#     ])
    
#     stuff_documents_chain = create_stuff_documents_chain(llm,prompt)
    
#     return create_retrieval_chain(retriever_chain, stuff_documents_chain)
def get_conversational_rag_chain(retriever_chain,llm):

    if not retriever_chain:
        raise ValueError("`retriever_chain` cannot be None or an empty object.")

    template = "Answer the user's questions based on the below context:\n\n{context}"
    human_template = "{input}"

    prompt = ChatPromptTemplate.from_messages([
      ("system", template),
      MessagesPlaceholder(variable_name="chat_history"),
      ("user", human_template),
    ])

    def safe_llm(input_str: str) -> str:
        if isinstance(input_str, langchain_core.prompts.chat.ChatPromptValue):
            input_str = str(input_str)
            # input_str = input_str.to_messages()

        # Call the original llm, which should now work correctly
        return llm(input_str)

    stuff_documents_chain = create_stuff_documents_chain(safe_llm, prompt)

    return create_retrieval_chain(retriever_chain, stuff_documents_chain)

def get_response(user_input):
    # llm = CTransformers(
    #                 # model = "TheBloke/Mistral-7B-Instruct-v0.2-GGUF",
    #                 model= "TheBloke/Llama-2-7B-Chat-GGUF",
    #                 model_file = "llama-2-7b-chat.Q3_K_S.gguf",
    #                 model_type="llama",
    #                 max_new_tokens = 300,
    #                 temperature = 0.3,
    #                 lib="avx2", # for CPU
    # )

    # model_name = "TinyLlama/TinyLlama-1.1B-Chat-v1.0"
    # # llm = HuggingFaceHub(
    # #     repo_id=llm_model, 
    # #     model_kwargs={"temperature": 0.3, "max_new_tokens": 250, "top_k": 3}
    # # )

    # llm = transformers.AutoModelForCausalLM.from_pretrained(
    #             model_name,
    #             trust_remote_code=True,
    #             torch_dtype=torch.bfloat16,
    #             device_map='auto'
    # )

    # llm = HuggingFacePipeline.from_model_id(
    #     model_id="google/flan-t5-base",
    #     task="text2text-generation",
    #     # model_kwargs={"temperature": 0.2},
    # )
    # llm = HuggingFacePipeline.from_model_id(
    #     model_id="google-t5/t5-small",
    #     task="text2text-generation",
    #     # model_kwargs={"temperature": 0.2},
    # )
    # llm = pipeline(task="conversational", model="facebook/blenderbot-400M-distill")
    llm = LlamaCpp(
    model_path="tinyllama-1.1b-chat-v1.0.Q4_0.gguf",
    temperature=0.75,
    max_tokens=2000,
    top_p=1,
    # callback_manager=callback_manager,
    # verbose=True,  # Verbose is required to pass to the callback manager
    )

    # llm = HuggingFacePipeline.from_model_id(
    #     model_id="lmsys/fastchat-t5-3b-v1.0",
    #     task="text2text-generation",
    #     # model_kwargs={"temperature": 0.2},
    # )
    
    retriever_chain = get_context_retriever_chain(st.session_state.vector_store,llm)
    conversation_rag_chain = get_conversational_rag_chain(retriever_chain,llm)
    
    response = conversation_rag_chain.invoke({
        "chat_history": st.session_state.chat_history,
        "input": user_query
    })
    
    return response['answer']

    
# app config
st.set_page_config(page_title= "Chat with Websites", page_icon="🤖")
st.title("Chat with Websites")





#sidebar
with st.sidebar:
    st.header("Settings")
    website_url = st.text_input("Website URL")
    # openai_apikey = st.text_input("Enter your OpenAI API key")

if (website_url is None or website_url == ""):
    st.info("Please ensure if website URL is entered")
    

else:
    
    if "chat_history" not in st.session_state:
        st.session_state.chat_history = [ 
        AIMessage(content = "Hello, I am a bot. How can I help you"),
        ]

    if "vector_store" not in st.session_state:
        st.session_state.vector_store = get_vector_store_from_url(website_url)

    
    #user_input
    user_query = st.chat_input("Type your message here...")
    if user_query is not None and user_query !="":
        response = get_response(user_query)
        st.session_state.chat_history.append(HumanMessage(content=user_query))
        st.session_state.chat_history.append(AIMessage(content=response))
        
                
    #conversation
    for message in st.session_state.chat_history:
        if isinstance(message, AIMessage): # checking if the messsage is the instance of an AI message
            with st.chat_message("AI"):
                st.write(message.content)
        elif isinstance(message, HumanMessage): # checking if the messsage is the instance of a Human
            with st.chat_message("Human"):
                st.write(message.content)