app.py

import os
import csv
import json
import docx
import pptx
import re
import nltk
import time
import PyPDF2
import tempfile
import openpyxl
import requests
import gradio as gr
from bs4 import BeautifulSoup
import xml.etree.ElementTree as ET
from nltk.tokenize import word_tokenize
from langchain_community.vectorstores import FAISS
from youtube_transcript_api import YouTubeTranscriptApi
from langchain_community.llms import HuggingFaceEndpoint
from langchain.schema import SystemMessage, HumanMessage, AIMessage
from langchain_community.chat_models.huggingface import ChatHuggingFace
from langchain_community.embeddings import SentenceTransformerEmbeddings
from youtube_transcript_api._errors import NoTranscriptFound, TranscriptsDisabled, VideoUnavailable
nltk.download('punkt')
nltk.download('omw-1.4')
nltk.download('wordnet')

def read_csv(file_path):
    with open(file_path, 'r', encoding='utf-8', errors='ignore', newline='') as csvfile:
        csv_reader = csv.reader(csvfile)
        csv_data = [row for row in csv_reader]
    return ' '.join([' '.join(row) for row in csv_data])

def read_text(file_path):
    with open(file_path, 'r', encoding='utf-8', errors='ignore',  newline='') as f:
        return f.read()

def read_pdf(file_path):
    text_data = []
    with open(file_path, 'rb') as pdf_file:
        pdf_reader = PyPDF2.PdfReader(pdf_file)
        for page in pdf_reader.pages:
            text_data.append(page.extract_text())
    return '\n'.join(text_data)

def read_docx(file_path):
    doc = docx.Document(file_path)
    return '\n'.join([paragraph.text for paragraph in doc.paragraphs])

def read_pptx(file_path):
    ppt = pptx.Presentation(file_path)
    text_data = ''
    for slide in ppt.slides:
        for shape in slide.shapes:
            if hasattr(shape, "text"):
                text_data += shape.text + '\n'
    return text_data

def read_xlsx(file_path):
    workbook = openpyxl.load_workbook(file_path)
    sheet = workbook.active
    text_data = ''
    for row in sheet.iter_rows(values_only=True):
        text_data += ' '.join([str(cell) for cell in row if cell is not None]) + '\n'
    return text_data

def read_json(file_path):
    with open(file_path, 'r') as f:
        json_data = json.load(f)
    return json.dumps(json_data)

def read_html(file_path):
    with open(file_path, 'r') as f:
        html_content = f.read()
    soup = BeautifulSoup(html_content, 'html.parser')
    return soup

def read_xml(file_path):
    tree = ET.parse(file_path)
    root = tree.getroot()
    return ET.tostring(root, encoding='unicode')

def process_youtube_video(url, languages=['en', 'ar']):
    if 'youtube.com/watch' in url or 'youtu.be/' in url:
        try:
            if "v=" in url:
                video_id = url.split("v=")[1].split("&")[0]
            elif "youtu.be/" in url:
                video_id = url.split("youtu.be/")[1].split("?")[0]
            else:
                return "Invalid YouTube video URL. Please provide a valid YouTube video link."

            response = requests.get(f"http://img.youtube.com/vi/{video_id}/mqdefault.jpg")
            if response.status_code != 200:
                return "Video doesn't exist."

            transcript_data = []
            for lang in languages:
                try:
                    transcript = YouTubeTranscriptApi.get_transcript(video_id, languages=[lang])
                    transcript_data.append(' '.join([entry['text'] for entry in transcript]))
                except (NoTranscriptFound, TranscriptsDisabled, VideoUnavailable):
                    continue

            return ' '.join(transcript_data) if transcript_data else "Please choose a YouTube video with available English or Arabic transcripts."

        except Exception as e:
            return f"An error occurred: {e}"
    else:
        return "Invalid YouTube URL. Please provide a valid YouTube link."

def read_web_page(url):
    result = requests.get(url)
    if result.status_code < 400:
        src = result.text
        soup = BeautifulSoup(src, 'html.parser')
        
        text_data = ''
        div_elements = soup.find_all('div')
        for div in div_elements:
            text_data += div.get_text() + '\n'
        
        return text_data
    else:
        return "Please provide a valid webpage link"

def read_data(file_path_or_url, languages=['en', 'ar']):
    if file_path_or_url.endswith('.csv'):
        return read_csv(file_path_or_url)
    elif file_path_or_url.endswith('.txt'):
        return read_text(file_path_or_url)
    elif file_path_or_url.endswith('.pdf'):
        return read_pdf(file_path_or_url)
    elif file_path_or_url.endswith('.docx'):
        return read_docx(file_path_or_url)
    elif file_path_or_url.endswith('.pptx'):
        return read_pptx(file_path_or_url)
    elif file_path_or_url.endswith('.xlsx'):
        return read_xlsx(file_path_or_url)
    elif file_path_or_url.endswith('.json'):
        return read_json(file_path_or_url)
    elif file_path_or_url.endswith('.html'):
        return read_html(file_path_or_url)
    elif file_path_or_url.endswith('.xml'):
        return read_xml(file_path_or_url)
    elif 'youtube.com/watch' in file_path_or_url or 'youtu.be/' in file_path_or_url:
        return process_youtube_video(file_path_or_url, languages)
    elif file_path_or_url.startswith('http'):
        return read_web_page(file_path_or_url)
    else:
        return "Unsupported type or format."

def normalize_text(text):
    text = re.sub("\*?", "", text)
    text = text.lower()
    text = text.strip()
    punctuation = '''!()[]{};:'"\<>/?$%^&*_`~='''
    for punc in punctuation:
        text = text.replace(punc, "")
    text = re.sub(r'[A-Za-z0-9]*@[A-Za-z]*\.?[A-Za-z0-9]*', "", text)
    words = word_tokenize(text)
    return ' '.join(words)

llm = HuggingFaceEndpoint(
    repo_id="HuggingFaceH4/starchat2-15b-v0.1",
    task="text-generation",
    max_new_tokens=4096,
    temperature=0.6,
    top_p=0.9,
    top_k=40,
    repetition_penalty=1.2,
    do_sample=True,
)
chat_model = ChatHuggingFace(llm=llm)

model_name = "sentence-transformers/all-mpnet-base-v2"
embedding_llm = SentenceTransformerEmbeddings(model_name=model_name)
db = FAISS.load_local("faiss_index", embedding_llm, allow_dangerous_deserialization=True)

def print_like_dislike(x: gr.LikeData):
    print(x.index, x.value, x.liked)

def user(user_message, history):
  if not len(user_message):
    raise gr.Error("Chat messages cannot be empty")
  return "", history + [[user_message, None]]

def user2(user_message, history, link):
    if not len(user_message) or not len(link):
        raise gr.Error("Chat messages or links cannot be empty")
    combined_message = f"{link}\n{user_message}"
    return "", history + [[combined_message, None]], link

def user3(user_message, history, file_path):
    if not len(user_message) or not file_path:
        raise gr.Error("Chat messages or flies cannot be empty")
    combined_message = f"{file_path}\n{user_message}"
    return "", history + [[combined_message, None]], file_path

messages1_state = [
  SystemMessage(content="You are a helpful assistant."),
  HumanMessage(content="Hi AI, how are you today?"),
AIMessage(content="I'm great thank you. How can I help you?")]

    
def Chat_Message(history, messages1):

    message=HumanMessage(content=history[-1][0])
    if isinstance(messages1[-1], HumanMessage):
        messages1=messages1[:-2]
        
    messages1.append(message)
    if len(messages1) >= 8:
      messages1 = messages1[-8:]
        
    try:
        response = chat_model.invoke(messages1)
    except Exception as e:
        error_message = str(e)
        start_index = error_message.find("Input validation error:")
        end_index = error_message.find("and 4096 `max_new_tokens`")
        if start_index != -1 and end_index != -1:
            raise gr.Error(error_message[start_index:end_index].strip()) from e
        else:
            raise gr.Error("Error occurred during response") from e
            
    messages1.append(AIMessage(content=response.content))

    history[-1][1] = ""
    for character in response.content:
        history[-1][1] += character
        time.sleep(0.0025)
        yield history, messages1

def Internet_Search(history, messages2):

    message=history[-1][0]
    if isinstance(messages2[-1], HumanMessage):
        messages2=messages2[:-2]

    similar_docs = db.similarity_search(message, k=3)

    if similar_docs:
        source_knowledge = "\n".join([x.page_content for x in similar_docs])
    else:
        source_knowledge = ""

    augmented_prompt = f"""
    You are an AI designed to help understand and extract information from provided Search Content. Based on the user's Query, you may need to summarize the text, answer specific questions, or provide guidance.
    Query: {message}
    Search Content:
    {source_knowledge}
    
    #If the query is not related to specific Search Content, engage in general conversation or provide relevant information from other sources.
    """
            
    msg=HumanMessage(content=augmented_prompt)
    messages2.append(msg)

    if len(messages2) >= 4:
        messages2 = messages2[-4:]
        
    try:
        response = chat_model.invoke(messages2)
    except Exception as e:
        error_message = str(e)
        start_index = error_message.find("Input validation error:")
        end_index = error_message.find("and 4096 `max_new_tokens`")
        if start_index != -1 and end_index != -1:
            raise gr.Error(error_message[start_index:end_index ].strip()) from e
        else:
            raise gr.Error("Error occurred during response") from e
            
    messages2.append(AIMessage(content=response.content))

    history[-1][1] = ""
    for character in response.content:
        history[-1][1] += character
        time.sleep(0.0025)
        yield history, messages2

def Chart_Generator(history, messages3):

    message = history[-1][0]
    if isinstance(messages3[-1], HumanMessage):
        messages3=messages3[:-2]
        
    if '#chart' in message:
        message = message.split('#chart', 1)[1].strip()
        chart_url = f"https://quickchart.io/natural/{message}"
        response = requests.get(chart_url)

        if response.status_code == 200:
            image_html = f'<img src="{chart_url}" alt="Generated Chart" style="display: block; margin: auto; max-width: 100%; max-height: 100%;">'
            message_with_description = f"Describe and analyse the content of this chart: {chart_url}"
    
            prompt = HumanMessage(content=message_with_description)
            messages3.append(prompt)

            if len(messages3) >= 6:
                messages3 = messages3[-6:]
    
            try:
                response = chat_model.invoke(messages3)
            except Exception as e:
                error_message = str(e)
                start_index = error_message.find("Input validation error:")
                end_index = error_message.find("and 4096 `max_new_tokens`")
                if start_index != -1 and end_index != -1:
                    raise gr.Error(error_message[start_index:end_index ].strip()) from e
                else:
                    raise gr.Error("Error occurred during response") from e
            
            messages3.append(AIMessage(content=response.content))
    
            combined_content = f'{image_html}<br>{response.content}'
        else:
            response_text = "Can't generate this image. Please provide valid chart details."
            combined_content = response_text
    else:
        prompt = HumanMessage(content=message)
        messages3.append(prompt)

        if len(messages3) >= 6:
            messages3 = messages3[-6:]
    
        try:
            response = chat_model.invoke(messages3)
        except Exception as e:
            error_message = str(e)
            start_index = error_message.find("Input validation error:")
            end_index = error_message.find("and 4096 `max_new_tokens`")
            if start_index != -1 and end_index != -1:
                raise gr.Error(error_message[start_index:end_index ].strip()) from e
            else:
                raise gr.Error("Error occurred during response") from e
        
        messages3.append(AIMessage(content=response.content))

        combined_content=response.content
        
    history[-1][1] = ""
    for character in combined_content:
        history[-1][1] += character
        time.sleep(0.0025)
        yield history, messages3

def Link_Scratch(history, messages4):

    combined_message = history[-1][0]
    if isinstance(messages4[-1], HumanMessage):
        messages4=messages4[:-2]

    link = ""
    user_message = ""
    if "\n" in combined_message:
        link, user_message = combined_message.split("\n", 1)
        link = link.strip()
        user_message = user_message.strip()

    result = read_data(link)

    if result in ["Unsupported type or format.", "Please provide a valid webpage link",
                  "Invalid YouTube URL. Please provide a valid YouTube link.",
                  "Please choose a YouTube video with available English or Arabic transcripts.",
                  "Invalid YouTube video URL. Please provide a valid YouTube video link."]:
        response_message = result
    else:
        content_data = normalize_text(result)
        if not content_data:
            response_message = "The provided link is empty or does not contain any meaningful words."
        else:
            augmented_prompt = f"""
            You are an AI designed to help understand and extract information from provided Link Content. Based on the user's Query, you may need to summarize the text, answer specific questions, or provide guidance.
            Query: {user_message}
            Link Content:
            {content_data}
            
            #If the query is not related to specific Link Content, engage in general conversation or provide relevant information from other sources.
            """
            message = HumanMessage(content=augmented_prompt)
            messages4.append(message)

            if len(messages4) >= 1:
                messages4 = messages4[-1:]

            try:
                response = chat_model.invoke(messages4)
            except Exception as e:
                error_message = str(e)
                start_index = error_message.find("Input validation error:")
                end_index = error_message.find("and 4096 `max_new_tokens`")
                if start_index != -1 and end_index != -1:
                    raise gr.Error(error_message[start_index:end_index ].strip()) from e
                else:
                    raise gr.Error("Error occurred during response") from e
            
            messages4.append(AIMessage(content=response.content))

            response_message = response.content

    history[-1][1] = ""
    for character in response_message:
        history[-1][1] += character
        time.sleep(0.0025)
        yield history, messages4

def insert_line_breaks(text, every=8):
    return '\n'.join(text[i:i+every] for i in range(0, len(text), every))

def display_file_name(file):
    supported_extensions = ['.csv', '.txt', '.pdf', '.docx', '.pptx', '.xlsx', '.json', '.html', '.xml']
    file_extension = os.path.splitext(file.name)[1]
    if file_extension.lower() in supported_extensions:
      file_name = os.path.basename(file.name)
      file_name_with_breaks = insert_line_breaks(file_name)
      icon_url = "https://img.icons8.com/ios-filled/50/0000FF/file.png"
      return f"<div style='display: flex; align-items: center;'><img src='{icon_url}' alt='file-icon' style='width: 20px; height: 20px; margin-right: 5px;'><b style='color:blue;'>{file_name_with_breaks}</b></div>"
    else:
      raise gr.Error("( Supported File Types Only : PDF , CSV , TXT , DOCX , PPTX , XLSX , JSON , HTML , XML )")

def File_Interact(history, filepath, messages5):

    combined_message = history[-1][0]
    if isinstance(messages5[-1], HumanMessage):
        messages5=messages5[:-2]
    
    link = ""
    user_message = ""
    if "\n" in combined_message:
      link, user_message = combined_message.split("\n", 1)
      user_message = user_message.strip()

    result = read_data(filepath)

    if result == "Unsupported type or format.":
        response_message = result
    else:
        content_data = normalize_text(result)
        if not content_data:
            response_message = "The file is empty or does not contain any meaningful words."
        else:
            augmented_prompt = f"""
            You are an AI designed to help understand and extract information from provided File Content. Based on the user's Query, you may need to summarize the text, answer specific questions, or provide guidance.
            Query: {user_message}
            File Content:
            {content_data}
            
            #If the query is not related to specific File Content, engage in general conversation or provide relevant information from other sources.
            """
            message = HumanMessage(content=augmented_prompt)
            messages5.append(message)

            if len(messages5) >= 1:
                messages5 = messages5[-1:]
                            
            try:
                response = chat_model.invoke(messages5)
            except Exception as e:
                error_message = str(e)
                start_index = error_message.find("Input validation error:")
                end_index = error_message.find("and 4096 `max_new_tokens`")
                if start_index != -1 and end_index != -1:
                    raise gr.Error(error_message[start_index:end_index ].strip()) from e
                else:
                    raise gr.Error("Error occurred during response") from e
            
            messages5.append(AIMessage(content=response.content))

            response_message = response.content

    history[-1][1] = ""
    for character in response_message:
        history[-1][1] += character
        time.sleep(0.0025)
        yield history, messages5

def Explore_WebSite(history, messages6):

    message=history[-1][0]
    if isinstance(messages6[-1], HumanMessage):
        messages6=messages6[:-2]

    links = [
        'https://huggingface.co/mou3az'
    ]
    
    result = "\n".join([read_data(link) for link in links])
    
    content_data = normalize_text(result)
    
    augmented_prompt = f"""
    You are an AI designed to help understand and extract information from provided WebSite Content. Based on the user's Query, you may need to summarize the text, answer specific questions, or provide guidance.
    Query: {message}
    WebSite Content:
    {content_data}
    
    #If the query is not related to specific WebSite Content, engage in general conversation or provide relevant information from other sources.
    """

    msg=HumanMessage(content=augmented_prompt)
    messages6.append(msg)

    if len(messages6) >= 4:
        messages6 = messages6[-4:]
        
    try:
        response = chat_model.invoke(messages6)
    except Exception as e:
        error_message = str(e)
        start_index = error_message.find("Input validation error:")
        end_index = error_message.find("and 4096 `max_new_tokens`")
        if start_index != -1 and end_index != -1:
            raise gr.Error(error_message[start_index:end_index ].strip()) from e
        else:
            raise gr.Error("Error occurred during response") from e
    
    messages6.append(AIMessage(content=response.content))

    history[-1][1] = ""
    for character in response.content:
        history[-1][1] += character
        time.sleep(0.0025)
        yield history, messages6
        
with gr.Blocks(theme=gr.themes.Soft()) as demo:
  with gr.Row():
    gr.Markdown("""<span style='font-weight: bold; color: blue; font-size: large;'>Choose Your Mode</span>""")
    gr.Markdown("""<div style='margin-left: -120px;'><span style='font-weight: bold; color: blue; font-size: xx-large;'>IT ASSISTANT</span></div>""")

  with gr.Tab("Chat-Message"):
    messages1=gr.State(messages1_state)
    chatbot = gr.Chatbot(
          [],
          elem_id="chatbot",
          bubble_full_width=False,
          height=500,
          placeholder="<span style='font-weight: bold; color: blue; font-size: x-large;'>Feel Free To Ask Me Anything Or Start A Conversation On Any Topic...</span>"
      )
    with gr.Row():
      msg = gr.Textbox(show_label=False, placeholder="Type a message...", scale=10, container=False)
      submit = gr.Button("➡️Send", scale=1)

    clear = gr.ClearButton([msg, chatbot, messages1])

    msg.submit(user, [msg, chatbot], [msg, chatbot], queue=True).then(Chat_Message, [chatbot,messages1], [chatbot,messages1])
    submit.click(user, [msg, chatbot], [msg, chatbot], queue=True).then(Chat_Message, [chatbot,messages1], [chatbot,messages1])
    chatbot.like(print_like_dislike, None, None)

  with gr.Tab("Internet-Search"):
    messages2=gr.State(messages1_state)
    chatbot = gr.Chatbot(
        [],
        elem_id="chatbot",
        bubble_full_width=False,
        height=500,
        placeholder="<span style='font-weight: bold; color: blue; font-size: x-large;'>Demand What You Seek, And I'll Search The Internet For The Most Relevant Information...</span>"
    )
    with gr.Row():
      msg = gr.Textbox(show_label=False, placeholder="Type a message...", scale=10, container=False)
      submit = gr.Button("➡️Send", scale=1)

    clear = gr.ClearButton([msg, chatbot, messages2])

    msg.submit(user, [msg, chatbot], [msg, chatbot], queue=True).then(Internet_Search, [chatbot,messages2], [chatbot,messages2])
    submit.click(user, [msg, chatbot], [msg, chatbot], queue=True).then(Internet_Search, [chatbot,messages2], [chatbot,messages2])
    chatbot.like(print_like_dislike, None, None)

  with gr.Tab("Chart-Generator"):
    messages3=gr.State(messages1_state)
    chatbot = gr.Chatbot(
        [],
        elem_id="chatbot",
        bubble_full_width=False,
        height=500,
        placeholder="<span style='font-weight: bold; color: blue; font-size: x-large;'>Request Any Chart Or Graph By Giving The Data Or A Description, And I'll Create It...</span>"
    )

    with gr.Row():
      msg = gr.Textbox(show_label=False, placeholder="To generate a chart: type #chart [your chart description ]. To discuss the chart: type your message directly...", scale=10, container=False)
      submit = gr.Button("➡️Send", scale=1)

    clear = gr.ClearButton([msg, chatbot, messages3])

    msg.submit(user, [msg, chatbot], [msg, chatbot], queue=True).then(Chart_Generator, [chatbot,messages3], [chatbot,messages3])
    submit.click(user, [msg, chatbot], [msg, chatbot], queue=True).then(Chart_Generator, [chatbot,messages3], [chatbot,messages3])
    chatbot.like(print_like_dislike, None, None)

  with gr.Tab("Link-Scratch"):
    messages4=gr.State(messages1_state)
    chatbot = gr.Chatbot(
        [],
        elem_id="chatbot",
        bubble_full_width=False,
        height=500,
        placeholder="<span style='font-weight: bold; color: blue; font-size: x-large;'>Provide A Link Of Web page Or YouTube Video And Inquire About Its Details...</span>"
    )

    with gr.Row():
        msg1 = gr.Textbox(show_label=False, placeholder="Paste your link...", scale=4, container=False)
        msg2 = gr.Textbox(show_label=False, placeholder="Type a message...", scale=7, container=False)
        submit = gr.Button("➡️Send", scale=1)

    clear = gr.ClearButton([msg2, chatbot, msg1, messages4])

    msg1.submit(user2, [msg2, chatbot, msg1], [msg2, chatbot, msg1], queue=True).then(Link_Scratch, [chatbot,messages4], [chatbot,messages4])
    msg2.submit(user2, [msg2, chatbot, msg1], [msg2, chatbot, msg1], queue=True).then(Link_Scratch, [chatbot,messages4], [chatbot,messages4])
    submit.click(user2, [msg2, chatbot, msg1], [msg2, chatbot, msg1], queue=True).then(Link_Scratch, [chatbot,messages4], [chatbot,messages4])
    chatbot.like(print_like_dislike, None, None)

  with gr.Tab("File-Interact"):
    messages5=gr.State(messages1_state)
    chatbot = gr.Chatbot(
        [],
        elem_id="chatbot",
        bubble_full_width=False,
        height=500,
        placeholder="<span style='font-weight: bold; color: blue; font-size: x-large;'>Upload A File And Explore Questions Related To Its Content...</span><br>( Supported File Types Only : PDF , CSV , TXT , DOCX , PPTX , XLSX , JSON , HTML , XML )"
    )

    with gr.Column():
        with gr.Row():
            filepath = gr.UploadButton("Upload a file", file_count="single", scale=1)
            msg = gr.Textbox(show_label=False, placeholder="Wait until the file has been uploaded, then type a message....", scale=7, container=False)
            submit = gr.Button("➡️Send", scale=1)
        with gr.Row():
            file_output = gr.HTML("<div style='height: 20px; width: 30px;'></div>")
            clear = gr.ClearButton([msg, filepath, chatbot,file_output, messages5],scale=6)

    filepath.upload(display_file_name, inputs=filepath, outputs=file_output)

    msg.submit(user3, [msg, chatbot, file_output], [msg, chatbot, file_output], queue=True).then(File_Interact, [chatbot, filepath, messages5],[chatbot, messages5])
    submit.click(user3, [msg, chatbot, file_output], [msg, chatbot, file_output], queue=True).then(File_Interact, [chatbot, filepath, messages5],[chatbot, messages5])
    chatbot.like(print_like_dislike, None, None)

  with gr.Tab("Explore-WebSite"):
    messages6=gr.State(messages1_state)
    chatbot = gr.Chatbot(
        [],
        elem_id="chatbot",
        bubble_full_width=False,
        height=500,
        placeholder="<span style='font-weight: bold; color: blue; font-size: x-large;'>Explore Any Information About Courses or Blogs In Our Web Site...</span>"
    )
    with gr.Row():
      msg = gr.Textbox(show_label=False, placeholder="Type a message...", scale=10, container=False)
      submit = gr.Button("➡️Send", scale=1)

    clear = gr.ClearButton([msg, chatbot, messages6])

    msg.submit(user, [msg, chatbot], [msg, chatbot], queue=True).then(Explore_WebSite, [chatbot, messages6], [chatbot, messages6])
    submit.click(user, [msg, chatbot], [msg, chatbot], queue=True).then(Explore_WebSite, [chatbot, messages6], [chatbot, messages6])
    chatbot.like(print_like_dislike, None, None)

demo.queue(max_size=10, default_concurrency_limit=4)
demo.launch(max_file_size="5mb", show_api=False, max_threads=50)