app.py

import gradio as gr
from langchain.embeddings.openai import OpenAIEmbeddings
from langchain.text_splitter import CharacterTextSplitter, RecursiveCharacterTextSplitter
from langchain.vectorstores import DocArrayInMemorySearch
from langchain.chains import RetrievalQA,  ConversationalRetrievalChain
from langchain.memory import ConversationBufferMemory
from langchain.chat_models import ChatOpenAI
from langchain.embeddings import HuggingFaceEmbeddings
from langchain import HuggingFaceHub
from langchain.llms import LlamaCpp
from huggingface_hub import hf_hub_download
from langchain.document_loaders import (
    EverNoteLoader,
    TextLoader,
    UnstructuredEPubLoader,
    UnstructuredHTMLLoader,
    UnstructuredMarkdownLoader,
    UnstructuredODTLoader,
    UnstructuredPowerPointLoader,
    UnstructuredWordDocumentLoader,
    PyPDFLoader,
)
import param
import os
import torch
from conversadocs.bones import DocChat

dc = DocChat()

##### GRADIO CONFIG ####

if torch.cuda.is_available():
    print("CUDA is available on this system.")
    os.system('CMAKE_ARGS="-DLLAMA_CUBLAS=on" FORCE_CMAKE=1 pip install llama-cpp-python --force-reinstall --upgrade --no-cache-dir --verbose')
else:
    print("CUDA is not available on this system.")
    os.system('pip install llama-cpp-python')

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

title = """
<div style="text-align: center;max-width: 700px;">
    <h1>Chat with Documents 📚 - Falcon, Llama-2</h1>
    <p style="text-align: center;">Upload txt, pdf, doc, docx, enex, epub, html, md, odt, ptt, pttx; click the "Click to Upload Files" button, <br />
    Wait for the Status to show Loaded documents, start typing your questions. <br />
    The app is set to store chat-history</p>
</div>
"""

theme='aliabid94/new-theme'

def flag():
  return "PROCESSING..."

def upload_file(files, max_docs):
    file_paths = [file.name for file in files]
    return dc.call_load_db(file_paths, max_docs)

def predict(message, chat_history, max_k):
        print(message)
        bot_message = dc.convchain(message, max_k)
        print(bot_message)
        return "", dc.get_chats()

def convert():
  docs = dc.get_sources()
  data_docs = ""
  for i in range(0,len(docs),2):
    txt = docs[i][1].replace("\n","<br>")
    sc = "Archive: " + docs[i+1][1]["source"]
    try:
      pg = "Page: " + str(docs[i+1][1]["page"])
    except:
      pg = "Document Data"
    data_docs += f"<hr><h3 style='color:red;'>{pg}</h2><p>{txt}</p><p>{sc}</p>"
  return data_docs



with gr.Blocks(theme=theme, css=css) as demo:
  with gr.Tab("Chat"):

    with gr.Column(elem_id="col-container"):
        gr.HTML(title)
        upload_button = gr.UploadButton("Click to Upload Files", file_types=["pdf"], file_count="multiple")
        file_output = gr.HTML()
        chatbot = gr.Chatbot([], elem_id="chatbot").style(height=300)
        msg = gr.Textbox(label="Question", placeholder="Type your question and hit Enter ")

    with gr.Column():
        sou = gr.HTML("")

  with gr.Tab("Chat Options"):
    max_docs = gr.inputs.Slider(1, 10, default=3, label="Maximum querys to the DB.", step=1)
    row_table = gr.HTML("<hr><h4> </h2>")
    clear_button = gr.Button("CLEAR CHAT HISTORY", )
    link_output = gr.HTML("")
    clear_button.click(flag,[],[link_output]).then(dc.clr_history,[], [link_output]).then(lambda: None, None, chatbot, queue=False)

    upload_button.upload(flag,[],[file_output]).then(upload_file, [upload_button, max_docs], file_output)

  with gr.Tab("Change model"):
    gr.HTML("<h3>Only models from the GGML library are accepted.</h3>")
    repo_ = gr.Textbox(label="Repository" ,value="TheBloke/Llama-2-7B-Chat-GGML")
    file_ = gr.Textbox(label="File name" ,value="llama-2-7b-chat.ggmlv3.q2_K.bin")
    max_tokens = gr.inputs.Slider(1, 2048, default=16, label="Max new tokens", step=1)
    temperature = gr.inputs.Slider(0.1, 1., default=0.2, label="Temperature", step=0.1)
    top_k = gr.inputs.Slider(0.01, 1., default=0.95, label="Top K", step=0.01)
    top_p = gr.inputs.Slider(0, 100, default=50, label="Top P", step=1)
    repeat_penalty = gr.inputs.Slider(0.1, 100., default=1.2, label="Repeat penalty", step=0.1)
    change_model_button = gr.Button("Load Model")
    model_verify = gr.HTML("Loaded model Falcon 7B-instruct")
    default_model = gr.HTML("<hr><h4>Default Model</h2>")
    falcon_button = gr.Button("FALCON 7B-Instruct")

  msg.submit(predict,[msg, chatbot, max_docs],[msg, chatbot]).then(convert,[],[sou])

  change_model_button.click(dc.change_llm,[repo_, file_, max_tokens, temperature, top_p, top_k, repeat_penalty, max_docs],[model_verify])
  falcon_button.click(dc.default_falcon_model, [], [model_verify])

demo.launch(enable_queue=True)