app.py

import gradio as gr
import torch as th

from langchain.document_loaders import PDFMinerLoader,CSVLoader ,UnstructuredWordDocumentLoader,TextLoader
from langchain.text_splitter import RecursiveCharacterTextSplitter
from langchain.embeddings import SentenceTransformerEmbeddings
from langchain.vectorstores import Chroma, FAISS
from langchain import HuggingFaceHub


DEVICE = 'cpu '
FILE_EXT = ['pdf','text','csv','word','wav']


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


def process_documents(documents,data_chunk=1000,chunk_overlap=50):
    text_splitter = RecursiveCharacterTextSplitter(chunk_size=data_chunk, chunk_overlap=chunk_overlap)
    texts = text_splitter.split_documents(documents[0])
    return texts

def get_hugging_face_model(model_id,API_key,temperature=0.1):
    chat_llm = HuggingFaceHub(huggingfacehub_api_token=API_key,
                                      repo_id=model_id,
                                      model_kwargs={"temperature": temperature, "max_new_tokens": 2048})
    return chat_llm

def document_loading(file_data,doc_type='pdf',key=None):
    
    embedding_model = SentenceTransformerEmbeddings(model_name='all-mpnet-base-v2',model_kwargs={"device": DEVICE})

    document = None
    if doc_type == 'pdf':
        document = process_pdf_document(document_file_name=file_data)
    elif doc_type == 'text':
        document = process_text_document(document_file_name=file_data)
    elif doc_type == 'csv':
        document = process_csv_document(document_file_name=file_data)
    elif doc_type == 'word':
        document = process_word_document(document_file_name=file_data)
    
    texts = process_documents(documents=document)
    vectordb = FAISS.from_documents(documents=texts, embedding= embedding_model)

        
def process_text_document(document_file_name):
    loader = TextLoader(document_file_name)
    document = loader.load()
    return document


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


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


def process_pdf_document(document_file_name):
    loader = PDFMinerLoader(document_file_name)
    document = loader.load()[0]
    return document





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 your computer, click the "Load data to LangChain" button, <br />
    when everything is ready, you can start asking questions about the data you uploaded ;) <br />
    This version is just for QA retrival so it will not use chat history, and uses Hugging face as LLM, 
    so you don't need any key</p>
</div>
"""

with gr.Blocks(css=css) as demo:
    with gr.Column(elem_id="col-container"):
        gr.HTML(title)
        
        with gr.Column():
            with gr.Box():
                LLM_option = gr.Dropdown(['HuggingFace','OpenAI'],label='LLM',info='select the LLM to be used')
                API_key = gr.Textbox(label="You OpenAI/Huggingface API key", type="password")
            with gr.Column():
                file_extension = gr.Dropdown(FILE_EXT, label="File Extensions", info="Select your files extensions!")
                pdf_doc = gr.File(label="Load a File", file_types=FILE_EXT, type="file")
                with gr.Row():
                    langchain_status = gr.Textbox(label="Status", placeholder="", interactive=False)
                    load_pdf = gr.Button("Load file to langchain")
        
        chatbot = gr.Chatbot([], elem_id="chatbot").style(height=350)
        question = gr.Textbox(label="Question", placeholder="Type your question and hit Enter ")