app.py

import streamlit as st
import pdfplumber
import re
from transformers import AutoTokenizer, AutoModelForCausalLM, pipeline
from gtts import gTTS
from sklearn.feature_extraction.text import CountVectorizer
from nltk.sentiment import SentimentIntensityAnalyzer

# Initialize necessary components
tokenizer = AutoTokenizer.from_pretrained("Qwen/Qwen-7B", trust_remote_code=True)
model = AutoModelForCausalLM.from_pretrained("Qwen/Qwen-7B", trust_remote_code=True)
qa_pipeline = pipeline("question-answering")
sia = SentimentIntensityAnalyzer()

# Helper functions
def extract_text_from_pdf(file):
    with pdfplumber.open(file) as pdf:
        text = ''.join([page.extract_text() for page in pdf.pages])
    return text

def clean_text(text):
    text = re.sub(r'\s*Page \d+\s*', '', text)  # Remove page numbers
    return text.strip()

def chunk_text(text, max_tokens=1024):
    words = text.split()
    chunks, current_chunk, current_token_count = [], [], 0
    for word in words:
        token_count = len(tokenizer(word)['input_ids'])
        if current_token_count + token_count > max_tokens:
            chunks.append(" ".join(current_chunk))
            current_chunk, current_token_count = [], 0
        current_chunk.append(word)
        current_token_count += token_count
    if current_chunk:
        chunks.append(" ".join(current_chunk))
    return chunks

def summarize_text_qwen(text, max_length=800):
    input_text = f"summarize: {text}"
    tokens = tokenizer(input_text, return_tensors="pt", truncation=True, max_length=1024)
    summary_ids = model.generate(
        tokens["input_ids"], max_length=max_length, min_length=200, 
        length_penalty=2.0, num_beams=4, early_stopping=True
    )
    return tokenizer.decode(summary_ids[0], skip_special_tokens=True)

def summarize_large_document(text, max_length=800):
    chunks = chunk_text(text)
    summaries = [summarize_text_qwen(chunk, max_length=max_length) for chunk in chunks]
    return " ".join(summaries)

def answer_question_with_context(question, context, chunk_size=500):
    chunks = chunk_text(context, max_tokens=chunk_size)
    answers = []
    for chunk in chunks:
        try:
            answers.append(qa_pipeline({'question': question, 'context': chunk})['answer'])
        except:
            continue
    return " ".join(answers)

# Replace Tortoise-TTS with gTTS for text-to-speech functionality
def text_to_speech(text, language="en"):
    tts = gTTS(text=text, lang=language, slow=False)
    file_name = "output_audio.mp3"
    tts.save(file_name)
    return file_name

def extract_keywords(text, top_n=10):
    vectorizer = CountVectorizer(stop_words="english")
    word_counts = vectorizer.fit_transform([text])
    keywords = sorted(
        zip(vectorizer.get_feature_names_out(), word_counts.toarray()[0]), 
        key=lambda x: x[1], reverse=True
    )[:top_n]
    return [word for word, count in keywords]

def analyze_sentiment(text):
    return sia.polarity_scores(text)

# Streamlit App Interface
st.title("Enhanced PDF to Audiobook App")
st.markdown("### Turn documents into interactive audiobooks with advanced features.")

uploaded_file = st.file_uploader("Upload a PDF", type="pdf")

if uploaded_file:
    with st.spinner("Extracting and cleaning PDF content..."):
        raw_text = extract_text_from_pdf(uploaded_file)
        cleaned_text = clean_text(raw_text)
    st.text_area("Extracted Text", cleaned_text[:5000], height=300, help="Displaying first 5000 characters.")

    if st.button("Summarize Document"):
        with st.spinner("Summarizing document..."):
            summary = summarize_large_document(cleaned_text, max_length=800)
        st.text_area("Summary", summary, height=300)

        if st.button("Convert Summary to Audiobook"):
            with st.spinner("Generating audio..."):
                audio_file = text_to_speech(summary)
            st.audio(audio_file, format="audio/mp3", start_time=0)

    st.markdown("### Ask Questions About the Document")
    question = st.text_input("Your Question:")
    if question:
        with st.spinner("Answering your question..."):
            answer = answer_question_with_context(question, cleaned_text)
        st.write(f"**Answer:** {answer}")
        if st.button("Convert Answer to Audio"):
            with st.spinner("Generating answer audio..."):
                answer_audio_file = text_to_speech(answer)
            st.audio(answer_audio_file, format="audio/mp3", start_time=0)

    st.markdown("### Document Insights")
    if st.checkbox("Extract Keywords"):
        with st.spinner("Extracting keywords..."):
            keywords = extract_keywords(cleaned_text)
        st.write("Keywords:", ", ".join(keywords))

    if st.checkbox("Analyze Sentiment"):
        with st.spinner("Analyzing sentiment..."):
            sentiment = analyze_sentiment(cleaned_text)
        st.write("Sentiment Analysis:", sentiment)