app.py

import gradio as gr
from transformers import pipeline
import torch
import re
import os
from PyPDF2 import PdfReader
import gtts
import tempfile
import warnings
import threading
import time
import speech_recognition as sr
import cv2
import numpy as np
import moviepy.editor as mp


# Fixed moviepy import
try:
    import moviepy.editor as mp
except ModuleNotFoundError:
    raise ImportError("The 'moviepy' module is not installed. Please add 'moviepy' to your requirements.txt and restart your Hugging Face Space.")

# Suppress warnings
warnings.filterwarnings("ignore", category=UserWarning, module="gtts")

# Initialize NLP models
nlp = pipeline("text-generation", model="distilgpt2", tokenizer="distilgpt2", device=0 if torch.cuda.is_available() else -1)
sentiment_analyzer = pipeline("sentiment-analysis", model="distilbert-base-uncased-finetuned-sst-2-english")

# Speech recognizer setup
r = sr.Recognizer()

# Extract text from PDF resume
def extract_text_from_pdf(pdf_file):
    try:
        reader = PdfReader(pdf_file.name)
        text = ""
        for page in reader.pages:
            text += page.extract_text() or ""
        return text if text else "No text found in the PDF."
    except Exception as e:
        return f"Error reading PDF: {str(e)}"

# Analyze resume and generate questions
def analyze_resume(resume_text, difficulty=1):
    generic_questions = [
        "What’s your greatest strength?",
        "Describe a challenge you overcame.",
        "Why do you want this role?"
    ]
    if not resume_text:
        return generic_questions[:difficulty]

    questions = []
    skills = re.findall(r"Skills:\s*(.*?)(?:\n|$)", resume_text, re.DOTALL | re.IGNORECASE)
    
    if skills:
        first_skill = skills[0].split(',')[0].strip()
        questions.append(f"Tell me about a time you used {first_skill} in a project.")

    return (questions + generic_questions)[:max(1, difficulty)]

# Analyze user response with sentiment analysis
def provide_feedback(response):
    if not response:
        return "Please provide an answer."
    sentiment = sentiment_analyzer(response)[0]
    feedback = []

    if len(response.split()) < 20:
        feedback.append("Your answer is short. Please elaborate.")
    if "I don’t know" in response.lower():
        feedback.append("Try sharing a related experience instead.")
    if sentiment["label"] == "NEGATIVE":
        feedback.append("Try to sound more positive and confident!")

    return " ".join(feedback) or "Great answer!"

# Analyze code syntax
def analyze_code(code):
    if not code:
        return "No code provided."
    try:
        ast.parse(code)
        return "Code syntax is valid! Consider adding comments for clarity."
    except SyntaxError as e:
        return f"Code error: {str(e)}"

# Transcribe audio from video or audio file
def transcribe_audio(file_path):
    try:
        if file_path.endswith(".mp4"):  # Handle video input
            video = mp.VideoFileClip(file_path)
            audio_path = tempfile.NamedTemporaryFile(suffix=".wav").name
            video.audio.write_audiofile(audio_path)
        else:
            audio_path = file_path

        with sr.AudioFile(audio_path) as source:
            audio = r.record(source)
        return r.recognize_google(audio)
    except Exception as e:
        return f"Error transcribing: {str(e)}"

# Gradio interface
with gr.Blocks(title="Nancy AI - Advanced Interview Simulator") as demo:
    gr.Markdown("# 🎤 Nancy AI - Advanced Interview Simulator")
    gr.Markdown("Upload your resume and a video response to get interview questions and feedback.")

    question_state = gr.State(value=0)
    questions_state = gr.State(value=[])
    responses_state = gr.State(value=[])
    timer_state = gr.State(value=60)

    with gr.Row():
        pdf_input = gr.File(label="📄 Upload PDF Resume", file_types=[".pdf"])
        difficulty = gr.Slider(1, 5, step=1, label="🔹 Difficulty Level", value=1)

    with gr.Row():
        video_input = gr.Video(label="🎥 Upload or Record Video Response", interactive=True)
        code_input = gr.Code(language="python", label="📝 Write Your Code (if applicable)")
        text_input = gr.Textbox(label="🗣️ Your Response", placeholder="Type your answer here...")

    with gr.Row():
        question_output = gr.Textbox(label="❓ Current Question", interactive=False)
        feedback_output = gr.Textbox(label="💡 Feedback", interactive=False)
        timer_display = gr.Textbox(label="⏳ Time Left (seconds)", interactive=False, value="60")

    submit_btn = gr.Button("✅ Submit Response")

    # Define click action
    def process_response(pdf_file, video_file, code_input, text_input, question_index, questions_state, responses_state, timer_state, difficulty):
        try:
            # Load or generate interview questions
            if not questions_state:
                resume_text = extract_text_from_pdf(pdf_file) if pdf_file else ""
                questions_state = analyze_resume(resume_text, difficulty)
                responses_state = [""] * len(questions_state)
                timer_state = 60  # Reset timer

            # Transcribe video/audio response
            user_response = transcribe_audio(video_file) if video_file else text_input
            if code_input:
                user_response = code_input
                code_feedback = analyze_code(code_input)
            else:
                code_feedback = ""

            # Save response
            if user_response and 0 <= question_index < len(questions_state):
                responses_state[question_index] = user_response

            # Provide feedback
            feedback = provide_feedback(user_response) + (f" {code_feedback}" if code_feedback else "")

            # Move to the next question
            if question_index >= len(questions_state) - 1:
                return "Interview complete!", "Thank you!", questions_state, responses_state, 0, None

            return questions_state[question_index], feedback, questions_state, responses_state, question_index + 1, str(max(0, timer_state - 10))
        except Exception as e:
            return f"Error: {str(e)}", "Something went wrong.", [], [], 0, "60"

    submit_btn.click(
        fn=process_response,
        inputs=[pdf_input, video_input, code_input, text_input, question_state, questions_state, responses_state, timer_state, difficulty],
        outputs=[question_output, feedback_output, questions_state, responses_state, question_state, timer_display]
    )

demo.launch()