import streamlit as st from langchain_groq import ChatGroq from langchain_core.output_parsers import StrOutputParser from langchain_core.prompts import ChatPromptTemplate from dotenv import load_dotenv import pytesseract from PIL import Image, ImageEnhance import pdfplumber import docx from io import BytesIO import logging import os from concurrent.futures import ThreadPoolExecutor import requests from bs4 import BeautifulSoup # Load environment variables load_dotenv() # Initialize logging logging.basicConfig(level=logging.INFO, format="%(asctime)s - %(levelname)s - %(message)s") # Initialize LLM llm = ChatGroq(temperature=0.5, groq_api_key="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", model_name="llama3-8b-8192") # OCR Configuration pytesseract.pytesseract.tesseract_cmd = r"/usr/bin/tesseract" # Adjust based on your system's path # Function to enhance image for OCR processing def enhance_image_for_ocr(image): # Convert to grayscale for better processing gray_image = image.convert("L") # Increase contrast for better text clarity enhancer = ImageEnhance.Contrast(gray_image) enhanced_image = enhancer.enhance(2.0) # Increase contrast return enhanced_image # Function to extract text from images using OCR def extract_text_from_images(images, lang="eng"): ocr_text = "" for image in images: try: enhanced_image = enhance_image_for_ocr(image) ocr_text += pytesseract.image_to_string(enhanced_image, lang=lang).strip() + "\n" except Exception as e: logging.error(f"Error in OCR: {e}") return ocr_text.strip() # Function to extract content from PDFs @st.cache_data def extract_pdf_data(pdf_file): data = {"text": "", "images": []} try: with pdfplumber.open(pdf_file) as pdf: for page in pdf.pages: data["text"] += page.extract_text() or "" for img in page.images: base_image = pdf.extract_image(img["object_number"]) image = Image.open(BytesIO(base_image["image"])) data["images"].append(image) except Exception as e: logging.error(f"Error processing PDF: {e}") return data # Function to extract content from DOCX files @st.cache_data def extract_docx_data(docx_file): try: doc = docx.Document(docx_file) text = "\n".join([para.text.strip() for para in doc.paragraphs if para.text.strip()]) return text except Exception as e: logging.error(f"Error processing DOCX: {e}") return "" # Function to extract plain text from TXT files @st.cache_data def extract_txt_data(txt_file): try: return txt_file.read().decode("utf-8").strip() except Exception as e: logging.error(f"Error processing TXT: {e}") return "" # Process uploaded files in parallel and extract text and images def process_files(uploaded_files, lang="eng"): combined_text = "" images = [] def process_file(file): file_type = file.type.split("/")[-1] if file_type == "pdf": pdf_data = extract_pdf_data(file) return pdf_data["text"], pdf_data["images"] elif file_type == "docx": return extract_docx_data(file), [] elif file_type == "txt": return extract_txt_data(file), [] elif file_type in ["png", "jpg", "jpeg"]: return "", [Image.open(file)] else: logging.error(f"Unsupported file type: {file_type}") return "", [] with ThreadPoolExecutor() as executor: results = list(executor.map(process_file, uploaded_files)) for text, img_list in results: combined_text += text images.extend(img_list) ocr_text = extract_text_from_images(images, lang) return combined_text + "\n" + ocr_text # Function to generate questions def generate_questions(question_type, syllabus_text, num_questions, difficulty, prompt_template): # Create a prompt based on user inputs prompt = prompt_template.format( num_questions=num_questions, question_type=question_type, syllabus_text=syllabus_text, **difficulty ) # Pass the prompt to the LLM chain = (ChatPromptTemplate.from_template(prompt) | llm | StrOutputParser()) try: questions = chain.invoke({}) return questions except Exception as e: logging.error(f"Error generating questions: {e}") return "" # Refined function to generate answers def generate_answers(questions, syllabus_text): answers = {} for i, question in enumerate(questions.split("\n")): if question.strip(): prompt = f""" Below is a syllabus excerpt. Please answer the following question based on the content provided. Ensure the answer is directly related to the question and specific to the syllabus. If necessary, explain key concepts clearly. Answer the question in a concise and detailed manner. Syllabus Content: {syllabus_text} Question: {question} Answer: """ chain = (ChatPromptTemplate.from_template(prompt) | llm | StrOutputParser()) try: answer = chain.invoke({}) answers[f"Answer {i+1}"] = answer.strip() except Exception as e: # Fall back to web search if LLM fails answers[f"Answer {i+1}"] = search_answers_online(question) return "\n".join([f"{k}: {v}" for k, v in answers.items()]) # Function to search answers online def search_answers_online(question): search_url = f"https://www.google.com/search?q={question}" headers = {"User-Agent": "Mozilla/5.0"} try: response = requests.get(search_url, headers=headers) soup = BeautifulSoup(response.text, "html.parser") snippets = soup.find_all("div", class_="BNeawe") return "\n".join([snippet.get_text() for snippet in snippets[:3]]) except Exception as e: logging.error(f"Error fetching online answers: {e}") return "No online answer found." # Streamlit UI st.title("AI-Powered Exam Generator") # Tabs for navigation tab1, tab2, tab3, tab4 = st.tabs(["📁 Upload Files", "📄 Preview Content", "📝 Generate Questions", "💡 Generate Answers"]) # Upload files with tab1: st.header("Upload Files") uploaded_files = st.file_uploader( "Upload your syllabus (PDF, DOCX, TXT, Images)", type=["pdf", "docx", "txt", "png", "jpg", "jpeg"], accept_multiple_files=True ) ocr_lang = st.selectbox("Select OCR Language", ["eng", "spa", "fra", "deu", "ita"]) if uploaded_files: syllabus_text = process_files(uploaded_files, lang=ocr_lang) st.session_state["syllabus_text"] = syllabus_text st.success("Files processed successfully!") # Preview content with tab2: st.header("Preview Syllabus Content") if "syllabus_text" in st.session_state: st.text_area("Extracted Content", st.session_state["syllabus_text"], height=300) if st.session_state.get("images"): for img in st.session_state["images"]: st.image(img, caption="Uploaded Image") else: st.warning("No content available. Upload files first.") # Generate questions and answers with tab3: st.header("Generate Questions and Answers") question_type = st.selectbox("Select Question Type", ["MCQs", "Short Questions", "Long Questions", "Fill-in-the-Blank", "Case Study"]) num_questions = st.text_input("Total Number of Questions") difficulty_levels = ["Remember", "Understand", "Apply", "Analyze", "Evaluate", "Create"] difficulty = {level: st.slider(level, 0, 5, 1) for level in difficulty_levels} prompt_template = st.text_area( "Edit Prompt Template", """ Generate {num_questions} {question_type} questions from the syllabus content below. Syllabus Content: {syllabus_text} Difficulty Levels: - Remember: {Remember} - Understand: {Understand} - Apply: {Apply} - Analyze: {Analyze} - Evaluate: {Evaluate} - Create: {Create} """, height=200 ) if num_questions.isdigit() and st.button("Generate Questions and Answers"): num_questions = int(num_questions) # Generate questions questions = generate_questions(question_type, st.session_state.get("syllabus_text", ""), num_questions, difficulty, prompt_template) st.session_state["questions"] = questions st.text_area("Generated Questions", questions, height=300) # Generate answers answers = generate_answers(questions, st.session_state.get("syllabus_text", "")) st.session_state["answers"] = answers st.text_area("Generated Answers", answers, height=300) # Download questions and answers st.download_button("Download Questions", questions, file_name="questions.txt") st.download_button("Download Answers", answers, file_name="answers.txt") # Generate answers with tab4: st.header("Generate Answers (Optional)") if "questions" in st.session_state: if st.button("Generate Answers"): answers = generate_answers(st.session_state["questions"], st.session_state.get("syllabus_text", "")) st.session_state["answers"] = answers st.text_area("Generated Answers", answers, height=300) # Download answers st.download_button("Download Answers", answers, file_name="answers.txt")