import PyPDF2 import os from bs4 import BeautifulSoup import tempfile import csv import json import xml.etree.ElementTree as ET import docx import pptx import openpyxl import re import nltk import time import requests import gradio as gr from nltk.tokenize import word_tokenize from langchain_community.embeddings import SentenceTransformerEmbeddings from langchain_community.vectorstores import FAISS from langchain_huggingface import HuggingFaceEndpoint from langchain.schema import SystemMessage, HumanMessage, AIMessage from langchain_community.chat_models.huggingface import ChatHuggingFace from youtube_transcript_api import YouTubeTranscriptApi 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 == 200: src = result.content soup = BeautifulSoup(src, 'html.parser') text_data = '' for p in soup.find_all('p'): text_data += p.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 def Chat_Message(history): messages = [ 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?")] message=HumanMessage(content=history[-1][0]) messages.append(message) response = chat_model.invoke(messages) messages.append(response.content) if len(messages) >= 8: messages = messages[-8:] history[-1][1] = "" for character in response.content: history[-1][1] += character time.sleep(0.0025) yield history def Web_Search(history): messages = [ 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?")] message=history[-1][0] 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""" If the answer to the next query is not contained in the Search, say 'No Answer Is Available' and then just give guidance for the query. Query: {message} Search: {source_knowledge} """ msg==HumanMessage(content=augmented_prompt) messages.append(msg) response = chat_model.invoke(msg) messages.append(response.content) if len(messages) >= 8: messages = messages[-8:] history[-1][1] = "" for character in response.content: history[-1][1] += character time.sleep(0.0025) yield history def Chart_Generator(history): messages = [ 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?") ] message = history[-1][0] chart_url = f"https://quickchart.io/natural/{message}" response = requests.get(chart_url) if response.status_code == 200: image_html = f'Generated Chart' message_with_description = f"Describe and analyse the content of this chart: {chart_url}" prompt = HumanMessage(content=message_with_description) messages.append(prompt) res = chat_model.invoke(messages) messages.append(res.content) if len(messages) >= 8: messages = messages[-8:] combined_content = f'{image_html}
{res.content}' else: response_text = "Can't generate this image. Please provide valid chart details." combined_content = response_text history[-1][1] = "" for character in combined_content: history[-1][1] += character time.sleep(0.0025) yield history def Link_Scratch(history): messages = [ 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?") ] combined_message = history[-1][0] 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""" If the answer to the next query is not contained in the Link Content, say 'No Answer Is Available' and then just give guidance for the query. Query: {user_message} Link Content: {content_data} """ message = HumanMessage(content=augmented_prompt) messages.append(message) response = chat_model.invoke(messages) messages.append(response.content) if len(messages) >= 1: messages = messages[-1:] response_message = response.content history[-1][1] = "" for character in response_message: history[-1][1] += character time.sleep(0.0025) yield history 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"
file-icon{file_name_with_breaks}
" else: raise gr.Error("( Supported File Types Only : PDF , CSV , TXT , DOCX , PPTX , XLSX , JSON , HTML , XML )") def File_Interact(history,filepath): messages = [ 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?")] combined_message = history[-1][0] 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""" If the answer to the next query is not contained in the File Content, say 'No Answer Is Available' and then just give guidance for the query. Query: {user_message} File Content: {content_data} """ message = HumanMessage(content=augmented_prompt) messages.append(message) response = chat_model.invoke(messages) messages.append(response.content) if len(messages) >= 1: messages = messages[-1:] response_message = response.content history[-1][1] = "" for character in response_message: history[-1][1] += character time.sleep(0.0025) yield history with gr.Blocks(theme=gr.themes.Soft()) as demo: with gr.Row(): gr.Markdown("""Choose Your Mode""") gr.Markdown("""
IT ASSISTANT
""") with gr.Tab("Chat-Message"): chatbot = gr.Chatbot( [], elem_id="chatbot", bubble_full_width=False, height=500, placeholder="Feel Free To Ask Me Anything Or Start A Conversation On Any Topic..." ) 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]) msg.submit(user, [msg, chatbot], [msg, chatbot], queue=True).then(Chat_Message, chatbot, chatbot) submit.click(user, [msg, chatbot], [msg, chatbot], queue=True).then(Chat_Message, chatbot, chatbot) chatbot.like(print_like_dislike, None, None) with gr.Tab("Web-Search"): chatbot = gr.Chatbot( [], elem_id="chatbot", bubble_full_width=False, height=500, placeholder="Demand What You Seek, And I'll Search The Web For The Most Relevant Information..." ) 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]) msg.submit(user, [msg, chatbot], [msg, chatbot], queue=True).then(Web_Search, chatbot, chatbot) submit.click(user, [msg, chatbot], [msg, chatbot], queue=True).then(Web_Search, chatbot, chatbot) chatbot.like(print_like_dislike, None, None) with gr.Tab("Chart-Generator"): chatbot = gr.Chatbot( [], elem_id="chatbot", bubble_full_width=False, height=500, placeholder="Request Any Chart Or Graph By Giving The Data Or A Description, And I'll Create It..." ) 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]) msg.submit(user, [msg, chatbot], [msg, chatbot], queue=True).then(Chart_Generator, chatbot, chatbot) submit.click(user, [msg, chatbot], [msg, chatbot], queue=True).then(Chart_Generator, chatbot, chatbot) chatbot.like(print_like_dislike, None, None) with gr.Tab("Link-Scratch"): chatbot = gr.Chatbot( [], elem_id="chatbot", bubble_full_width=False, height=500, placeholder="Provide A Link Of Web page Or YouTube Video And Inquire About Its Details..." ) 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]) msg1.submit(user2, [msg2, chatbot, msg1], [msg2, chatbot, msg1], queue=True).then(Link_Scratch, chatbot, chatbot) msg2.submit(user2, [msg2, chatbot, msg1], [msg2, chatbot, msg1], queue=True).then(Link_Scratch, chatbot, chatbot) submit.click(user2, [msg2, chatbot, msg1], [msg2, chatbot, msg1], queue=True).then(Link_Scratch, chatbot, chatbot) chatbot.like(print_like_dislike, None, None) with gr.Tab("File-Interact"): chatbot = gr.Chatbot( [], elem_id="chatbot", bubble_full_width=False, height=500, placeholder="Upload A File And Explore Questions Related To Its Content...
( 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="Type a message...", scale=7, container=False) submit = gr.Button("➡️Send", scale=1) with gr.Row(): file_output = gr.HTML("
") clear = gr.ClearButton([msg, filepath, chatbot,file_output],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],chatbot) submit.click(user3, [msg, chatbot, file_output], [msg, chatbot, file_output], queue=True).then(File_Interact, [chatbot, filepath],chatbot) chatbot.like(print_like_dislike, None, None) demo.queue(max_size=5) demo.launch(max_file_size="5mb",show_api=False)