app.py

# JB:
# LangChainDeprecationWarning: Importing embeddings from langchain is deprecated. 
# Importing from langchain will no longer be supported as of langchain==0.2.0.
# Please import from langchain-community instead:
# `from langchain_community.embeddings import FastEmbedEmbeddings`.
# To install langchain-community run `pip install -U langchain-community`.
from langchain_community.embeddings import FastEmbedEmbeddings

import os
import streamlit as st
from langchain_groq import ChatGroq
from langchain_community.document_loaders import WebBaseLoader
# JB:
from langchain_community.document_loaders import PyPDFLoader
from langchain_community.embeddings import OllamaEmbeddings

# JB:
from langchain.embeddings import FastEmbedEmbeddings
from langchain_community.document_loaders import PyPDFDirectoryLoader

# JB:
# File Directory
# This covers how to load all documents in a directory.
# Under the hood, by default this uses the UnstructuredLoader.
from langchain_community.document_loaders import DirectoryLoader
from langchain_community.document_loaders import TextLoader
import chardet

from langchain_community.vectorstores import FAISS
# from langchain.vectorstores import Chroma
# from langchain_community.vectorstores import Chroma

from langchain.text_splitter import RecursiveCharacterTextSplitter
from langchain.chains.combine_documents import create_stuff_documents_chain
from langchain_core.prompts import ChatPromptTemplate
from langchain.chains import create_retrieval_chain
import time
from dotenv import load_dotenv

load_dotenv()  #

# groq_api_key = os.environ['GROQ_API_KEY']
groq_api_key = "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" # os.environ['GROQ_API_KEY']
print("groq_api_key: ", groq_api_key)


if "vector" not in st.session_state:

    # st.session_state.embeddings = OllamaEmbeddings() # ORIGINAL
    st.session_state.embeddings = FastEmbedEmbeddings() # JB


    # st.session_state.loader = WebBaseLoader("https://paulgraham.com/greatwork.html") # ORIGINAL
    # st.session_state.docs = st.session_state.loader.load()                           # ORIGINAL
    # https://api.python.langchain.com/en/latest/document_loaders/langchain_community.document_loaders.pdf.PyPDFLoader.html
    # https://python.langchain.com/docs/integrations/document_loaders/merge_doc
    # from langchain_community.document_loaders import PyPDFLoader
    # loader_pdf = PyPDFLoader("../MachineLearning-Lecture01.pdf")
    #
    # https://stackoverflow.com/questions/60215731/pypdf-to-read-each-pdf-in-a-folder
    # 
    # https://api.python.langchain.com/en/latest/document_loaders/langchain_community.document_loaders.pdf.PyPDFDirectoryLoader.html
    # https://python.langchain.com/docs/modules/data_connection/document_loaders/pdf#pypdf-directory 
    # !!!!!
    # PyPDF Directory
    # Load PDFs from directory
    # from langchain_community.document_loaders import PyPDFDirectoryLoader
    # loader = PyPDFDirectoryLoader("example_data/")
    # docs = loader.load()
    #
    # ZIE OOK:
    # https://python.langchain.com/docs/modules/data_connection/document_loaders/pdf#using-pypdf
    # Using MathPix
    # Inspired by Daniel Gross's https://gist.github.com/danielgross/3ab4104e14faccc12b49200843adab21
    # from langchain_community.document_loaders import MathpixPDFLoader
    # loader = MathpixPDFLoader("example_data/layout-parser-paper.pdf")
    # data = loader.load()
    # pdf_file_path = "*.pdf"                                                 # JB
    # st.session_state.loader = PyPDFLoader(file_path=pdf_file_path).load()   # JB
    # st.session_state.loader = PyPDFLoader(*.pdf).load()                     # JB syntax error *.pdf !
    # st.session_state.loader = PyPDFDirectoryLoader("*.pdf")                 # JB PyPDFDirectoryLoader("example_data/")   
    # chunks = self.text_splitter.split_documents(docs)
    # chunks = filter_complex_metadata(chunks)

    # JB:
    # https://python.langchain.com/docs/modules/data_connection/document_loaders/pdf#pypdf-directory
    # st.session_state.docs = st.session_state.loader.load()
    # loader = PyPDFDirectoryLoader(".")
    # docs = loader.load()
    # st.session_state.docs = docs

    # JB:
    # https://python.langchain.com/docs/modules/data_connection/document_loaders/file_directory
    # text_loader_kwargs={'autodetect_encoding': True}
    text_loader_kwargs={'autodetect_encoding': False}
    path = '../'
    # loader = DirectoryLoader(path, glob="**/*.pdf", loader_cls=TextLoader, loader_kwargs=text_loader_kwargs)
    # PyPDFDirectoryLoader (TEST):
    # loader = PyPDFDirectoryLoader(path, glob="**/*.pdf", loader_cls=TextLoader, loader_kwargs=text_loader_kwargs)
    # loader = PyPDFDirectoryLoader(path, glob="**/*.pdf", loader_kwargs=text_loader_kwargs)
    loader = PyPDFDirectoryLoader(path, glob="**/*.pdf")
    docs = loader.load()
    st.session_state.docs = docs
    
    st.session_state.text_splitter = RecursiveCharacterTextSplitter(chunk_size=1000, chunk_overlap=200)
    st.session_state.documents = st.session_state.text_splitter.split_documents(st.session_state.docs)
    # st.session_state.vector = FAISS.from_documents(st.session_state.documents, st.session_state.embeddings) # ORIGINAL
    st.session_state.vector = FAISS.from_documents(st.session_state.documents, st.session_state.embeddings) # ORIGINAL
    # ZIE: 
    # ZIE VOOR EEN APP MET CHROMADB:
    # https://github.com/vndee/local-rag-example/blob/main/rag.py
    # https://raw.githubusercontent.com/vndee/local-rag-example/main/rag.py
    # Chroma.from_documents(documents=chunks, embedding=FastEmbedEmbeddings())
    # st.session_state.vector = Chroma.from_documents(st.session_state.documents, st.session_state.embeddings) # JB


    
# st.title("Chat with Docs - Groq Edition :) ")
st.title("Literature Based Research (LBR) - A. Unzicker and J. Bours - Chat with Docs - Groq Edition (Very Fast!) - VERSION 3 - March 8 2024")

llm = ChatGroq(
            groq_api_key=groq_api_key, 
            model_name='mixtral-8x7b-32768'
    )

prompt = ChatPromptTemplate.from_template("""
Answer the following question based only on the provided context. 
Think step by step before providing a detailed answer. 
I will tip you $200 if the user finds the answer helpful. 
<context>
{context}
</context>
Question: {input}""")

document_chain = create_stuff_documents_chain(llm, prompt)

retriever = st.session_state.vector.as_retriever()
retrieval_chain = create_retrieval_chain(retriever, document_chain)

prompt = st.text_input("Input your prompt here")


# If the user hits enter
if prompt:
    # Then pass the prompt to the LLM
    start = time.process_time()
    response = retrieval_chain.invoke({"input": prompt})
    print(f"Response time: {time.process_time() - start}")

    st.write(response["answer"])

    # With a streamlit expander
    with st.expander("Document Similarity Search"):
        # Find the relevant chunks
        for i, doc in enumerate(response["context"]):
            # print(doc)
            # st.write(f"Source Document # {i+1} : {doc.metadata['source'].split('/')[-1]}")
            st.write(doc.page_content)
            st.write("--------------------------------")