app.py

import gradio as gr
import os

from langchain_community.document_loaders import PyPDFLoader
from langchain.text_splitter import RecursiveCharacterTextSplitter
from langchain_community.vectorstores import Chroma
from langchain.chains import ConversationalRetrievalChain
from langchain_community.embeddings import HuggingFaceEmbeddings
from langchain_community.llms import HuggingFacePipeline
from langchain.chains import ConversationChain
from langchain.memory import ConversationBufferMemory
from langchain_community.llms import HuggingFaceEndpoint

from pathlib import Path
import chromadb
from unidecode import unidecode

from transformers import AutoTokenizer
import transformers
import torch
import tqdm
import accelerate
import re

# Static PDF file link
static_pdf_link = "https://huggingface.co/spaces/CCCDev/PDFChat/resolve/main/Data-privacy-policy.pdf"

list_llm = ["mistralai/Mistral-7B-Instruct-v0.2", "mistralai/Mixtral-8x7B-Instruct-v0.1",
            "mistralai/Mistral-7B-Instruct-v0.1", "google/gemma-7b-it", "google/gemma-2b-it",
            "HuggingFaceH4/zephyr-7b-beta", "HuggingFaceH4/zephyr-7b-gemma-v0.1",
            "meta-llama/Llama-2-7b-chat-hf", "microsoft/phi-2",
            "TinyLlama/TinyLlama-1.1B-Chat-v1.0", "mosaicml/mpt-7b-instruct", "tiiuae/falcon-7b-instruct",
            "google/flan-t5-xxl"]
list_llm_simple = [os.path.basename(llm) for llm in list_llm]


# Load PDF document and create doc splits
def load_doc(file_path, chunk_size, chunk_overlap):
    loader = PyPDFLoader(file_path)
    pages = loader.load()
    text_splitter = RecursiveCharacterTextSplitter(
        chunk_size=chunk_size,
        chunk_overlap=chunk_overlap)
    doc_splits = text_splitter.split_documents(pages)
    return doc_splits


# Create vector database
def create_db(splits, collection_name):
    embedding = HuggingFaceEmbeddings()
    new_client = chromadb.EphemeralClient()
    vectordb = Chroma.from_documents(
        documents=splits,
        embedding=embedding,
        client=new_client,
        collection_name=collection_name,
    )
    return vectordb


# Initialize langchain LLM chain
def initialize_llmchain(llm_model, temperature, max_tokens, top_k, vector_db, progress=gr.Progress()):
    progress(0.1, desc="Initializing HF tokenizer...")

    progress(0.5, desc="Initializing HF Hub...")
    if llm_model == "mistralai/Mixtral-8x7B-Instruct-v0.1":
        llm = HuggingFaceEndpoint(
            repo_id=llm_model,
            temperature=temperature,
            max_new_tokens=max_tokens,
            top_k=top_k,
            load_in_8bit=True,
        )
    elif llm_model in ["HuggingFaceH4/zephyr-7b-gemma-v0.1", "mosaicml/mpt-7b-instruct"]:
        raise gr.Error("LLM model is too large to be loaded automatically on free inference endpoint")
    elif llm_model == "microsoft/phi-2":
        llm = HuggingFaceEndpoint(
            repo_id=llm_model,
            temperature=temperature,
            max_new_tokens=max_tokens,
            top_k=top_k,
            trust_remote_code=True,
            torch_dtype="auto",
        )
    elif llm_model == "TinyLlama/TinyLlama-1.1B-Chat-v1.0":
        llm = HuggingFaceEndpoint(
            repo_id=llm_model,
            temperature=temperature,
            max_new_tokens=250,
            top_k=top_k,
        )
    elif llm_model == "meta-llama/Llama-2-7b-chat-hf":
        raise gr.Error("Llama-2-7b-chat-hf model requires a Pro subscription...")
    else:
        llm = HuggingFaceEndpoint(
            repo_id=llm_model,
            temperature=temperature,
            max_new_tokens=max_tokens,
            top_k=top_k,
        )

    progress(0.75, desc="Defining buffer memory...")
    memory = ConversationBufferMemory(
        memory_key="chat_history",
        output_key='answer',
        return_messages=True
    )
    retriever = vector_db.as_retriever()
    progress(0.8, desc="Defining retrieval chain...")
    qa_chain = ConversationalRetrievalChain.from_llm(
        llm,
        retriever=retriever,
        chain_type="stuff",
        memory=memory,
        return_source_documents=True,
        verbose=False,
    )
    progress(0.9, desc="Done!")
    return qa_chain


# Generate collection name for vector database
def create_collection_name(filepath):
    collection_name = Path(filepath).stem
    collection_name = collection_name.replace(" ", "-")
    collection_name = unidecode(collection_name)
    collection_name = re.sub('[^A-Za-z0-9]+', '-', collection_name)
    collection_name = collection_name[:50]
    if len(collection_name) < 3:
        collection_name = collection_name + 'xyz'
    if not collection_name[0].isalnum():
        collection_name = 'A' + collection_name[1:]
    if not collection_name[-1].isalnum():
        collection_name = collection_name[:-1] + 'Z'
    print('Filepath: ', filepath)
    print('Collection name: ', collection_name)
    return collection_name


# Initialize database
def initialize_database(chunk_size, chunk_overlap, progress=gr.Progress()):
    file_path = static_pdf_link
    progress(0.1, desc="Creating collection name...")
    collection_name = create_collection_name(file_path)
    progress(0.25, desc="Loading document...")
    doc_splits = load_doc(file_path, chunk_size, chunk_overlap)
    progress(0.5, desc="Generating vector database...")
    vector_db = create_db(doc_splits, collection_name)
    progress(0.9, desc="Done!")
    return vector_db, collection_name, "Complete!"


def initialize_LLM(llm_option, llm_temperature, max_tokens, top_k, vector_db, progress=gr.Progress()):
    llm_name = list_llm[llm_option]
    print("llm_name: ", llm_name)
    qa_chain = initialize_llmchain(llm_name, llm_temperature, max_tokens, top_k, vector_db, progress)
    return qa_chain, "Complete!"


def format_chat_history(message, chat_history):
    formatted_chat_history = []
    for user_message, bot_message in chat_history:
        formatted_chat_history.append(f"User: {user_message}")
        formatted_chat_history.append(f"Assistant: {bot_message}")
    return formatted_chat_history


def conversation(qa_chain, message, history):
    formatted_chat_history = format_chat_history(message, history)
    response = qa_chain({"question": message, "chat_history": formatted_chat_history})
    response_answer = response["answer"]
    if response_answer.find("Helpful Answer:") != -1:
        response_answer = response_answer.split("Helpful Answer:")[-1]
    response_sources = response["source_documents"]
    response_source1 = response_sources[0].page_content.strip()
    response_source2 = response_sources[1].page_content.strip()
    response_source3 = response_sources[2].page_content.strip()
    response_source1_page = response_sources[0].metadata["page"] + 1
    response_source2_page = response_sources[1].metadata["page"] + 1
    response_source3_page = response_sources[2].metadata["page"] + 1

    new_history = history + [(message, response_answer)]
    return qa_chain, gr.update(
        value=""), new_history, response_source1, response_source1_page, response_source2, response_source2_page, response_source3, response_source3_page


def demo():
    with gr.Blocks(theme="base") as demo:
        vector_db = gr.State()
        qa_chain = gr.State()
        collection_name = gr.State()

        gr.Markdown(
            """<center><h2>PDF-based chatbot</center></h2>
            <h3>Ask any questions about your PDF documents</h3>""")
        gr.Markdown(
            """<b>Note:</b> This AI assistant, using Langchain and open-source LLMs, performs retrieval-augmented generation (RAG) from your PDF documents. \
            The user interface explicitely shows multiple steps to help understand the RAG workflow. 
            This chatbot takes past questions into account when generating answers (via conversational memory), and includes document references for clarity purposes.<br>
            <br><b>Warning:</b> This space uses the free CPU Basic hardware from Hugging Face. Some steps and LLM models used below (free inference endpoints) can take some time to generate a reply.
            """)

        with gr.Tab("Step 2 - Process document"):
            with gr.Row():
                db_btn = gr.Radio(["ChromaDB"], label="Vector database type", value="ChromaDB", type="index",
                                  info="Choose your vector database")
            with gr.Accordion("Advanced options - Document text splitter", open=False):
                with gr.Row():
                    chunk_size = gr.Slider(64, 4096, value=512, step=32, label="Text chunk size",
                                           info="Text length of each document chunk being embedded into the vector database. Default is 512.")
                    chunk_overlap = gr.Slider(0, 1024, value=24, step=8, label="Text chunk overlap",
                                              info="Text overlap between each document chunk being embedded into the vector database. Default is 24.")

            initialize_db = gr.Button("Process document")

            with gr.Row():
                output_db = gr.Textbox(label="Database initialization steps", placeholder="", show_label=False)
                with gr.Accordion("Vector database collection details", open=False):
                    collection = gr.Textbox(label="Collection name", placeholder="", show_label=False)

        with gr.Tab("Step 3 - Initialize LLM"):
            with gr.Row():
                llm_options = gr.Dropdown(list_llm_simple, label="Choose open-source LLM",
                                          value="Mistral-7B-Instruct-v0.2",
                                          info="Choose among the proposed open-source LLMs")
            with gr.Accordion("Advanced LLM options", open=False):
                with gr.Row():
                    llm_temperature = gr.Slider(0.01, 1.0, value=0.1, step=0.01, label="LLM temperature",
                                                info="LLM sampling temperature, in [0.01,1.0] range. Default is 0.1")
                    llm_max_tokens = gr.Slider(32, 1024, value=512, step=16, label="Max tokens",
                                               info="Maximum number of new tokens to be generated, in [32,1024] range. Default is 512")
                    llm_top_k = gr.Slider(1, 40, value=20, step=1, label="Top K",
                                          info="The number of highest probability vocabulary tokens to keep for top-k-filtering. Default is 20.")

            initialize_llm = gr.Button("Initialize LLM")

            with gr.Row():
                output_llm = gr.Textbox(label="LLM initialization steps", placeholder="", show_label=False)

        with gr.Tab("Step 4 - Start chatting"):
            chatbot = gr.Chatbot(label="PDF chatbot", height=500)
            msg = gr.Textbox(label="Your question", placeholder="Type your question here...", show_label=False)
            clear = gr.Button("Clear chat")

            with gr.Accordion("Document sources (3)", open=False):
                gr.Markdown("Source 1")
                response_src1 = gr.Textbox(label="Source 1", placeholder="", show_label=False)
                response_src1_page = gr.Number(label="Page number", value=0, precision=0, interactive=False)
                gr.Markdown("Source 2")
                response_src2 = gr.Textbox(label="Source 2", placeholder="", show_label=False)
                response_src2_page = gr.Number(label="Page number", value=0, precision=0, interactive=False)
                gr.Markdown("Source 3")
                response_src3 = gr.Textbox(label="Source 3", placeholder="", show_label=False)
                response_src3_page = gr.Number(label="Page number", value=0, precision=0, interactive=False)

        initialize_db.click(initialize_database,
                            inputs=[chunk_size, chunk_overlap],
                            outputs=[vector_db, collection_name, output_db])
        initialize_llm.click(initialize_LLM,
                             inputs=[llm_options, llm_temperature, llm_max_tokens, llm_top_k, vector_db],
                             outputs=[qa_chain, output_llm])
        msg.submit(conversation,
                   inputs=[qa_chain, msg, chatbot],
                   outputs=[chatbot, msg, chatbot, response_src1, response_src1_page, response_src2, response_src2_page,
                            response_src3, response_src3_page])
        clear.click(lambda: None, None, chatbot, queue=False)
        clear.click(lambda: None, None, msg, queue=False)

    return demo.queue().launch(debug=True)


# demo().launch(server_name="0.0.0.0")
if __name__ == "__main__":
    demo()