app.py

import gradio as gr
from transformers import pipeline
import torch
from TTS.api import TTS
import tempfile
import os
import speech_recognition as sr

# Load question generation model
qg_pipeline = pipeline("text2text-generation", model="valhalla/t5-small-e2e-qg")

# Load TTS model
tts = TTS(model_name="tts_models/en/ljspeech/tacotron2-DDC", progress_bar=False, gpu=False)

# Global storage
last_answer = ""

def generate_question(text):
    global last_answer

    # Extract a possible answer from the text (you can improve this logic)
    last_answer = "They are Aladdin lamps"  # You can auto-extract later

    # Prompt for question generation
    input_text = f"generate question: {text.strip()}"
    generated = qg_pipeline(input_text, max_length=64)[0]["generated_text"]

    # Keep only the first question
    question = generated.split("<sep>")[0].strip()

    # Generate audio
    with tempfile.NamedTemporaryFile(delete=False, suffix=".wav") as fp:
        tts.tts_to_file(text=question, file_path=fp.name)
        audio_path = fp.name

    return question, audio_path

def transcribe_and_check(audio_path):
    global last_answer
    recognizer = sr.Recognizer()
    with sr.AudioFile(audio_path) as source:
        audio_data = recognizer.record(source)
        try:
            response = recognizer.recognize_google(audio_data)
        except sr.UnknownValueError:
            return "Sorry, I could not understand your answer.", ""
        except sr.RequestError:
            return "Speech recognition service error.", ""

    # Check match (very basic – can be made smarter)
    if last_answer.lower() in response.lower():
        feedback = "✔️ Correct!"
    else:
        feedback = "❌ Try again."

    return response, feedback

with gr.Blocks() as app:
    gr.Markdown("### 🎓 Interactive Coursebook Q&A")

    course_text = gr.Textbox(lines=6, label="📘 Coursebook Text")

    generate_btn = gr.Button("🔊 Generate Question and Speak")
    question_output = gr.Textbox(label="🧠 Generated Question")
    audio_output = gr.Audio(label="🔈 Question Audio", type="filepath")

    user_audio = gr.Audio(label="🎤 Your Answer", type="filepath", sources=["microphone"])
    transcribe_btn = gr.Button("📝 Transcribe Answer")
    transcription_output = gr.Textbox(label="🗣 Transcribed Answer")
    feedback_output = gr.Textbox(label="🧪 Feedback")

    generate_btn.click(fn=generate_question, inputs=course_text, outputs=[question_output, audio_output])
    transcribe_btn.click(fn=transcribe_and_check, inputs=user_audio, outputs=[transcription_output, feedback_output])

app.launch()