RohanSardar/simpleRAG

1import os2import streamlit as st3from dotenv import load_dotenv4from langchain_groq import ChatGroq5from langchain_community.vectorstores import FAISS6from langchain_community.embeddings import HuggingFaceEmbeddings7from langchain.text_splitter import RecursiveCharacterTextSplitter8from langchain.chains.combine_documents import create_stuff_documents_chain9from langchain_core.prompts import ChatPromptTemplate10from langchain.chains import create_retrieval_chain11from langchain_community.document_loaders import PyPDFDirectoryLoader12 13load_dotenv()14groq_api_key = os.getenv('GROQ_API_KEY')15 16model = ChatGroq(groq_api_key=groq_api_key, model='Llama3-8b-8192')17 18prompt = ChatPromptTemplate.from_template(19"""20Answer the questions based on the context only.21Provide the answer accurately and briefly to the question22<context>23{context}24<context>25Question:{input}26"""27)28 29st.set_page_config(page_title = 'Simple RAG', page_icon = '⛓️', initial_sidebar_state = 'collapsed')30 31st.sidebar.header('About')32st.sidebar.markdown(33"""34Embeddings: Craig/paraphrase-MiniLM-L6-v2  35VectorDB: FAISS  36LLM: Llama3-8b-819237"""38)39 40st.title('Simple RAG Application')41 42st.warning('This is a simple RAG demonstration application. It uses open-source models for embeddings and \43inference. So it can be slow and ineffecient.', icon='⚠️')44 45 46def create_vector_embedding():47    if 'vectors' not in st.session_state:48        st.session_state.embeddings = HuggingFaceEmbeddings(model_name='Craig/paraphrase-MiniLM-L6-v2')49        st.session_state.loader = PyPDFDirectoryLoader('documents')50        st.session_state.docs = st.session_state.loader.load()51        st.session_state.text_splitter = RecursiveCharacterTextSplitter(chunk_size=1000, chunk_overlap=200)52        st.session_state.final_documents = st.session_state.text_splitter.split_documents(st.session_state.docs[:50])53        st.session_state.vectors = FAISS.from_documents(st.session_state.final_documents, st.session_state.embeddings)54        st.rerun()55 56if 'vectors' not in st.session_state:57    st.write('The vector store database is not yet ready')58    if st.button('Create'):59        with st.spinner('Working...'):60            create_vector_embedding()61 62if 'vectors' in st.session_state:63    user_prompt = st.text_input('Enter your query here')64    if user_prompt:65        document_chain = create_stuff_documents_chain(model, prompt)66        retriever = st.session_state.vectors.as_retriever()67        retrieval_chain = create_retrieval_chain(retriever, document_chain)68        response = retrieval_chain.invoke({'input': user_prompt})69        st.write(response['answer'])70 71        with st.expander('Context'):72            for i, doc in enumerate(response['context']):73                st.write(doc.page_content)74                st.write('\n\n')