app.py

# app_gradio.py
import gradio as gr
import numpy as np
import os
import yaml
from dotenv import load_dotenv
import io
from scipy.io.wavfile import read as read_wav

# Correctly import from the drive_paddy package structure
from src.detection.factory import get_detector
from src.alerting.alert_system import get_alerter

# --- Load Configuration and Environment Variables ---
# This part is the same as our Streamlit app
load_dotenv()
config_path = 'config.yaml'
with open(config_path, 'r') as f:
    config = yaml.safe_load(f)
secrets = {
    "gemini_api_key": os.getenv("GEMINI_API_KEY"),
}

# --- Initialize Backend Components ---
# We create these once and reuse them.
detector = get_detector(config)
alerter = get_alerter(config, secrets["gemini_api_key"])

# --- Audio Processing for Gradio ---
# Gradio's gr.Audio component needs a specific format: (sample_rate, numpy_array)
def process_audio_for_gradio(audio_bytes):
    """Converts in-memory audio bytes to a format Gradio can play."""
    # gTTS creates MP3, so we read it as such
    byte_io = io.BytesIO(audio_bytes)
    # The 'read' function from scipy.io.wavfile expects a WAV file.
    # We need to first convert the MP3 bytes from gTTS to WAV bytes.
    # This requires pydub.
    try:
        from pydub import AudioSegment
        audio = AudioSegment.from_mp3(byte_io)
        wav_byte_io = io.BytesIO()
        audio.export(wav_byte_io, format="wav")
        wav_byte_io.seek(0)
        
        sample_rate, data = read_wav(wav_byte_io)
        return (sample_rate, data)
    except Exception as e:
        print(f"Could not process audio for Gradio: {e}")
        return None

# --- Main Processing Function for Gradio ---
# This function is the core of the app. It takes a webcam frame and returns
# updates for all the output components.
def process_live_frame(frame):
    """
    Takes a single frame from the Gradio webcam input, processes it,
    and returns the processed frame, status text, and any audio alerts.
    """
    if frame is None:
        # Return default values if frame is None
        blank_image = np.zeros((480, 640, 3), dtype=np.uint8)
        return blank_image, "Status: Inactive", None

    # Process the frame using our existing detector
    processed_frame, indicators, _ = detector.process_frame(frame)
    drowsiness_level = indicators.get("drowsiness_level", "Awake")
    lighting = indicators.get("lighting", "Good")
    score = indicators.get("details", {}).get("Score", 0)

    # Build the status text
    status_text = f"Lighting: {lighting}\n"
    if lighting == "Low":
        status_text += "Detection paused due to low light."
    else:
        status_text += f"Status: {drowsiness_level}\nScore: {score:.2f}"

    # Handle alerts
    audio_output = None
    if drowsiness_level != "Awake":
        audio_data = alerter.trigger_alert(level=drowsiness_level)
        if audio_data:
            audio_output = process_audio_for_gradio(audio_data)
    else:
        alerter.reset_alert()

    return processed_frame, status_text, audio_output

# --- UI Definition for the Live Detection Page ---
def create_live_detection_page():
    """Builds the Gradio UI components for the live detection tab."""
    with gr.Blocks(theme=gr.themes.Default(primary_hue="blue", secondary_hue="blue")) as live_detection_page:
        gr.Markdown("A live test using Gradio's webcam component.")
        with gr.Row():
            with gr.Column():
                webcam_input = gr.Image(sources=["webcam"], streaming=True, label="Live Camera Feed")
            with gr.Column():
                processed_output = gr.Image(label="Processed Feed")
                status_output = gr.Textbox(label="Live Status", lines=3, interactive=False)
                # Audio player is now visible for debugging and user feedback.
                audio_alert_output = gr.Audio(autoplay=True, visible=True, label="Alert Sound")

        webcam_input.stream(
            fn=process_live_frame,
            inputs=[webcam_input],
            outputs=[processed_output, status_output, audio_alert_output]
        )
    return live_detection_page

# --- UI Definition for the Home Page ---
def create_home_page():
    """Builds the Gradio UI components for the home/welcome tab."""
    with gr.Blocks(theme=gr.themes.Default(primary_hue="blue", secondary_hue="blue")) as home_page:
        gr.Markdown(
        """
        <div align="center">
          <img src="https://em-content.zobj.net/source/samsung/380/automobile_1f697.png" alt="Car Emoji" width="100"/>
          <h1>Welcome to Drive Paddy!</h1>
          <p><strong>Your Drowsiness Detection Assistant</strong></p>
        </div>
        
        ---
        
        ### How It Works
        This application uses your webcam to monitor for signs of drowsiness in real-time. Navigate to the **Live Detection** tab to begin.
        
        - **Multi-Signal Analysis**: Detects eye closure, yawning, and head position.
        - **AI-Powered Alerts**: Uses Gemini to generate dynamic audio warnings.
        - **Live Feedback**: Provides instant visual feedback on the video stream and status panel.
        """
        )
    return home_page

# --- Combine Pages into a Tabbed Interface ---
app = gr.TabbedInterface(
    [create_home_page(), create_live_detection_page()],
    ["Home", "Live Detection"]
)

# --- Launch the App ---
app.launch(debug=True)