services/tts_service.py

import logging
import io
import wave
import numpy as np

logger = logging.getLogger(__name__)

# Try to import gTTS, but provide fallback if not available
try:
    from gtts import gTTS
    GTTS_AVAILABLE = True
except ImportError:
    GTTS_AVAILABLE = False
    logger.warning("gTTS not available. Using fallback audio generation.")


async def generate_tts(text: str) -> bytes:
    """
    Convert text to speech using free gTTS (Google Text-to-Speech).
    
    Args:
        text: Text to convert to speech
    
    Returns:
        Audio bytes in MP3 format
    
    Raises:
        Exception: If TTS generation fails
    """
    try:
        logger.info(f"Generating speech for: '{text}'")
        
        # Use gTTS if available
        if GTTS_AVAILABLE:
            tts = gTTS(text=text, lang='en', slow=False)
            audio_buffer = io.BytesIO()
            tts.write_to_fp(audio_buffer)
            audio_bytes = audio_buffer.getvalue()
        else:
            # Fallback to simple tone generation
            audio_bytes = generate_fallback_audio(text)
        
        logger.info(f"✓ TTS successful: {len(audio_bytes)} bytes generated")
        return audio_bytes
        
    except Exception as e:
        logger.error(f"✗ TTS failed: {str(e)}")
        # Ultimate fallback
        return generate_silent_audio()


def generate_fallback_audio(text: str) -> bytes:
    """
    Generate a simple tone-based audio file as fallback.
    """
    try:
        # Create a simple sine wave
        sample_rate = 22050
        duration = max(1.0, min(3.0, len(text) * 0.1))
        
        t = np.linspace(0, duration, int(sample_rate * duration), False)
        
        # Generate tones that vary with the text length
        base_freq = 440  # A4 note
        # Add some variation based on text
        freq_variation = min(200, len(text) * 5)
        tone = 0.3 * np.sin(2 * np.pi * (base_freq + freq_variation) * t)
        
        # Convert to 16-bit PCM
        audio_data = (tone * 32767).astype(np.int16)
        
        # Create WAV file in memory
        buffer = io.BytesIO()
        with wave.open(buffer, 'wb') as wav_file:
            wav_file.setnchannels(1)  # Mono
            wav_file.setsampwidth(2)  # 2 bytes = 16-bit
            wav_file.setframerate(sample_rate)
            wav_file.writeframes(audio_data.tobytes())
        
        return buffer.getvalue()
        
    except Exception as e:
        logger.error(f"Fallback audio generation failed: {str(e)}")
        return generate_silent_audio()


def generate_silent_audio() -> bytes:
    """
    Generate a short silent audio file as ultimate fallback.
    """
    try:
        sample_rate = 22050
        duration = 1.0
        
        # Generate silence
        silent_data = np.zeros(int(sample_rate * duration), dtype=np.int16)
        
        # Create WAV file in memory
        buffer = io.BytesIO()
        with wave.open(buffer, 'wb') as wav_file:
            wav_file.setnchannels(1)  # Mono
            wav_file.setsampwidth(2)  # 2 bytes = 16-bit
            wav_file.setframerate(sample_rate)
            wav_file.writeframes(silent_data.tobytes())
        
        return buffer.getvalue()
        
    except Exception as e:
        logger.error(f"Silent audio generation failed: {str(e)}")
        # Return empty bytes as last resort
        return b""