app.py

import os 
import gradio as gr
import time 
from langchain.document_loaders import PDFMinerLoader,CSVLoader ,UnstructuredWordDocumentLoader,TextLoader,OnlinePDFLoader
from langchain.text_splitter import CharacterTextSplitter
from langchain.embeddings import SentenceTransformerEmbeddings
from langchain.vectorstores import FAISS
from langchain.chains import RetrievalQA
from langchain.prompts import PromptTemplate
from langchain.llms.huggingface_pipeline import HuggingFacePipeline


DEVICE = 'cpu'
FILE_EXT = ['pdf','text','csv','word']
DEFAULT_SYSTEM_PROMPT = "As a chatbot you are answering set of questions being requested ."
MAX_NEW_TOKENS = 4096
DEFAULT_TEMPERATURE = 0.1
DEFAULT_MAX_NEW_TOKENS = 2048
MAX_INPUT_TOKEN_LENGTH = 4000

def loading_file():
    return "Loading..."


def process_documents(documents,data_chunk=1500,chunk_overlap=100):
    text_splitter = CharacterTextSplitter(chunk_size=data_chunk, chunk_overlap=chunk_overlap,separator='\n')
    texts = text_splitter.split_documents(documents)
    return texts

def get_hugging_face_model(model_id,temperature=0.1,max_tokens=4096,API_key=None):
    chat_llm = HuggingFacePipeline.from_model_id(
        model_id=model_id,
        task="text-generation",
        pipeline_kwargs={"max_new_tokens": max_tokens,"temperature": temperature,},
    )
    # chat_llm = HuggingFaceHub(huggingfacehub_api_token=API_key,
    #                                   repo_id=model_id,
    #                                   model_kwargs={"temperature": temperature, "max_new_tokens": max_tokens})
    return chat_llm

def chat_application(temperature=0.1, max_tokens=1024):

    llm = get_hugging_face_model(model_id='tiiuae/falcon-7b-instruct',temperature=temperature, max_tokens=max_tokens)
    return llm 


def document_loader(file_path,doc_type='pdf',temperature=0.1,max_tokens=2048):
    document = None
    if doc_type == 'pdf':
        document = process_pdf_document(document_file=file_path)
    elif doc_type == 'text':
        document = process_text_document(document_file=file_path)
    elif doc_type == 'csv':
        document = process_csv_document(document_file=file_path)
    elif doc_type == 'word':
        document = process_word_document(document_file=file_path)
    
    embedding_model = SentenceTransformerEmbeddings(model_name='thenlper/gte-base',model_kwargs={"device": DEVICE})
    
    texts = process_documents(documents=document)
    global vector_db
    vector_db = FAISS.from_documents(documents=texts, embedding= embedding_model)
    global qa
    qa = RetrievalQA.from_chain_type(llm=chat_application(temperature=temperature,
                                                            max_tokens=max_tokens
                                                        ),
                                chain_type='stuff',
                                retriever=vector_db.as_retriever(),
                                #  chain_type_kwargs=chain_type_kwargs,
                                return_source_documents=True
                            )
    return "Document Processing completed ..."
        
def process_text_document(document_file):
    loader = TextLoader(document_file.name)
    document = loader.load()
    return document

def process_csv_document(document_file):
    loader = CSVLoader(file_path=document_file.name)
    document = loader.load()
    return document

def process_word_document(document_file):
    loader = UnstructuredWordDocumentLoader(file_path=document_file.name)
    document = loader.load()
    return document

def process_pdf_document(document_file):
    print("Document File Name :",document_file.name)
    loader = PDFMinerLoader(document_file.name)
    document = loader.load()
    return document

def clear_chat():
    return []

def infer(question, history):
    # res = []
    # # for human, ai in history[:-1]:
    # #     pair = (human, ai)
    # #     res.append(pair)
    
    # chat_history = res
    print("Question in infer :",question)
    result = qa({"query": question})
    matching_docs_score = vector_db.similarity_search_with_score(question)
    
    print(" Matching_doc ",matching_docs_score)
    
    return result["result"]

def bot(history):
   
    response = infer(history[-1][0], history)
    history[-1][1] = ""
    
    for character in response:     
        history[-1][1] += character
        time.sleep(0.05)
        yield history

def add_text(history, text):
    history = history + [(text, None)]
    return history, ""

css="""
#col-container {max-width: 700px; margin-left: auto; margin-right: auto;}
"""

title = """
<div style="text-align: center;max-width: 700px;">
    <h1>Chat with Data • OpenAI/HuggingFace</h1>
    <p style="text-align: center;">Upload a file from system,UpLoad file and generate embeddings, <br />
    once status is ready, you can start asking questions about the data you uploaded without chat history <br />
    and gives you option to use HuggingFace/OpenAI as LLM's, make sure to add your key.
    </p>
</div>
"""

with gr.Blocks(css=css) as demo:
    with gr.Column(elem_id="col-container"):
        gr.HTML(title)
        
    with gr.Group():
        chatbot = gr.Chatbot(height=300)
    with gr.Row():
        question = gr.Textbox(label="Type your question !",lines=1)
        submit_btn = gr.Button(value="Send message", variant="primary", scale = 1)
        clean_chat_btn =  gr.Button("Delete Chat")

    with gr.Column():
        with gr.Box():
            LLM_option = gr.Dropdown(['tiiuae/falcon-7b-instruct'],label='Large Language Model Selection',info='LLM Service')
                 
        with gr.Column():
            with gr.Box():
                file_extension = gr.Dropdown(FILE_EXT, label="File Extensions", info="Select type of file to upload !")
                pdf_doc = gr.File(label="Upload File to start QA", file_types=FILE_EXT, type="file")
                with gr.Accordion(label='Advanced options', open=False):
                    max_new_tokens = gr.Slider(
                        label='Max new tokens',
                        minimum=2048,
                        maximum=MAX_NEW_TOKENS,
                        step=1024,
                        value=DEFAULT_MAX_NEW_TOKENS,
                        )
                    temperature = gr.Slider(
                    label='Temperature',
                    minimum=0.1,
                    maximum=4.0,
                    step=0.1,
                    value=DEFAULT_TEMPERATURE,
                    )
                with gr.Row():
                    langchain_status = gr.Textbox(label="Status", placeholder="", interactive = False)
                    load_pdf = gr.Button("Upload File & Generate Embeddings",).style(full_width = False)

        # chatbot = gr.Chatbot()l̥
        # question = gr.Textbox(label="Question", placeholder="Type your question and hit Enter")
        # submit_button = gr.Button("Send Message")

    if pdf_doc:
        load_pdf.click(loading_file, None, langchain_status, queue=False)    
        load_pdf.click(document_loader, inputs=[pdf_doc,file_extension,temperature,max_new_tokens], outputs=[langchain_status], queue=False)

    question.submit(add_text, inputs=[chatbot, question], outputs=[chatbot, question]).then(bot, chatbot, chatbot)
    submit_btn.click(add_text, inputs=[chatbot, question], outputs=[chatbot, question]).then(bot, chatbot, chatbot)
    # submit_btn.then(chatf.highlight_found_text, [chatbot, sources], [sources])
    clean_chat_btn.click(clear_chat, [], chatbot)


demo.launch()