app.py

import streamlit as st
from streamlit_chat import message
from langchain.chains import ConversationalRetrievalChain
from langchain.document_loaders import PyPDFLoader, DirectoryLoader
from langchain.embeddings import HuggingFaceEmbeddings
from langchain.llms import CTransformers
from langchain.text_splitter import RecursiveCharacterTextSplitter
from langchain.vectorstores import FAISS
from langchain.memory import ConversationBufferMemory

from huggingface_hub import login
login(token =st.secrets["HF"])

# Function to load documents 
def load_documents():
    loader = DirectoryLoader('data/', glob="*.pdf", loader_cls=PyPDFLoader)
    documents = loader.load()
    return documents

# Function to split text into chunks
def split_text_into_chunks(documents):
    text_splitter = RecursiveCharacterTextSplitter(chunk_size=500, chunk_overlap=50)
    text_chunks = text_splitter.split_documents(documents)
    return text_chunks

# Function to create embeddings
def create_embeddings():
    embeddings = HuggingFaceEmbeddings(model_name="sentence-transformers/paraphrase-multilingual-MiniLM-L12-v2", model_kwargs={'device': "cpu"})
    #embeddings = HuggingFaceEmbeddings(model_name="sentence-transformers/all-MiniLM-L6-v2", model_kwargs={'device': "cpu"})
    return embeddings

# Function to create vector store
def create_vector_store(text_chunks, embeddings, nombre_vector):
    vector_store = FAISS.from_documents(text_chunks, embeddings)
    vector_store.save_local("cache") #Guardarlo en un 
    return vector_store

# Function to create vector store
def load_vector_store(nombre_vector, embeddings):
    return FAISS.load_local(nombre_vector, embeddings)

# Function to create LLMS model
def create_llms_model():
    llm = CTransformers(model='TheBloke/Mistral-7B-Instruct-v0.1-GGUF', config={'max_new_tokens': 128, 'temperature': 0.01})
    return llm

# Initialize Streamlit app
st.title("Chatbot usando mistral")

# loading of documents
documents = load_documents()

# Split text into chunks
text_chunks = split_text_into_chunks(documents)

# Create embeddings
embeddings = create_embeddings()

try:#load vector store from local
    vector_store = load_vector_store("cache",embeddings)
except:# Create vector store
    vector_store = create_vector_store(text_chunks, embeddings, "cache")
# Create LLMS model
llm = create_llms_model()

# Initialize conversation history
if 'history' not in st.session_state:
    st.session_state['history'] = []

if 'generated' not in st.session_state:
    st.session_state['generated'] = ["¡Hola! Pregúntame sobre cualquier cosa 🤗"]

if 'past' not in st.session_state:
    st.session_state['past'] = ["¡Hola! 👋"]

# Create memory
memory = ConversationBufferMemory(memory_key="chat_history", return_messages=True)

# Create chain
chain = ConversationalRetrievalChain.from_llm(llm=llm, chain_type='stuff',
                                              retriever=vector_store.as_retriever(search_kwargs={"k": 2}),
                                              memory=memory)

# Define chat function
def conversation_chat(query):
    result = chain({"question": query, "chat_history": st.session_state['history']})
    st.session_state['history'].append((query, result["answer"]))
    return result["answer"]

# Display chat history
reply_container = st.container()
container = st.container()

with container:
    with st.form(key='my_form', clear_on_submit=True):
        user_input = st.text_input("Question:", placeholder="Ask about your Job Interview", key='input')
        submit_button = st.form_submit_button(label='Send')

    if submit_button and user_input:
        output = conversation_chat(user_input)
        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="fun-emoji")