backend/audio/tts_service.py

"""Text-to-Speech service using ElevenLabs API for on-demand audio generation."""

import os
import logging
import tempfile
from typing import Optional, List, Dict, Any
from elevenlabs.client import AsyncElevenLabs
from elevenlabs import VoiceSettings

logger = logging.getLogger(__name__)


class TTSService:
    """Text-to-Speech service for generating audio narration on-demand."""

    def __init__(self, api_key: Optional[str] = None):
        """Initialise TTS service with ElevenLabs API.

        Args:
            api_key: ElevenLabs API key (uses env var if not provided)
        """
        self.api_key = api_key or os.getenv("ELEVENLABS_API_KEY")
        if not self.api_key:
            logger.warning("ELEVENLABS_API_KEY not set - audio generation will fail")
            self.client = None
        else:
            self.client = AsyncElevenLabs(api_key=self.api_key)

        # Default voice: George - professional, neutral male voice
        self.default_voice_id = "JBFqnCBsd6RMkjVDRZzb"

    def is_available(self) -> bool:
        """Check if TTS service is available."""
        return self.client is not None

    async def generate_audio(
        self,
        text: str,
        voice_id: Optional[str] = None,
        model: str = "eleven_multilingual_v2",
        voice_settings: Optional[VoiceSettings] = None
    ) -> bytes:
        """Generate audio from text.

        Args:
            text: Text to convert to speech
            voice_id: ElevenLabs voice ID (uses default if not provided)
            model: ElevenLabs model ID
            voice_settings: Optional voice customisation

        Returns:
            Audio data as bytes (MP3 format)

        Raises:
            RuntimeError: If TTS service not available
        """
        if not self.is_available():
            raise RuntimeError("TTS service not available - check ELEVENLABS_API_KEY")

        if not text or not text.strip():
            raise ValueError("Text cannot be empty")

        logger.info(f"Generating audio: {len(text)} characters")

        try:
            audio_generator = self.client.text_to_speech.convert(
                text=text,
                voice_id=voice_id or self.default_voice_id,
                model_id=model,
                voice_settings=voice_settings,
                output_format="mp3_44100_128"
            )

            # Collect audio chunks
            audio_chunks = []
            async for chunk in audio_generator:
                audio_chunks.append(chunk)

            audio_data = b"".join(audio_chunks)
            logger.info(f"Audio generated: {len(audio_data)} bytes")
            return audio_data

        except Exception as e:
            logger.error(f"Audio generation failed: {e}")
            raise

    async def generate_analysis_narration(
        self,
        analysis_text: str,
        recommendations: Optional[List[str]] = None
    ) -> str:
        """Generate audio narration for portfolio analysis.

        Args:
            analysis_text: Main analysis text/summary
            recommendations: Optional list of recommendations

        Returns:
            Path to generated MP3 file
        """
        if not self.is_available():
            raise RuntimeError("TTS service not available")

        # Build narrative script
        script = "Portfolio Analysis Summary.\n\n"
        script += analysis_text

        if recommendations:
            script += "\n\nRecommendations:\n"
            for i, rec in enumerate(recommendations, 1):
                script += f"\n{i}. {rec}\n"

        script += "\n\nThis analysis is for informational purposes only and does not constitute financial advice."

        # Generate audio
        audio_data = await self.generate_audio(script)

        # Save to temporary file
        temp_file = tempfile.NamedTemporaryFile(
            delete=False,
            suffix=".mp3",
            prefix="analysis_"
        )
        temp_file.write(audio_data)
        temp_file.close()

        logger.info(f"Analysis narration saved to: {temp_file.name}")
        return temp_file.name

    async def generate_portfolio_narration(
        self,
        portfolio_summary: str,
        holdings: Optional[List[Dict[str, Any]]] = None
    ) -> str:
        """Generate audio narration for built portfolio.

        Args:
            portfolio_summary: Portfolio summary text
            holdings: Optional list of holdings with ticker and allocation

        Returns:
            Path to generated MP3 file
        """
        if not self.is_available():
            raise RuntimeError("TTS service not available")

        script = "Portfolio Construction Complete.\n\n"
        script += portfolio_summary

        if holdings:
            script += "\n\nPortfolio Holdings:\n"
            for holding in holdings[:10]:  # Limit to top 10
                ticker = holding.get("ticker", "Unknown")
                weight = holding.get("weight", 0)
                script += f"{ticker}: {weight:.1f}% allocation. "

        script += "\n\nRemember to conduct your own research before making investment decisions."

        # Generate audio
        audio_data = await self.generate_audio(script)

        # Save to temporary file
        temp_file = tempfile.NamedTemporaryFile(
            delete=False,
            suffix=".mp3",
            prefix="portfolio_"
        )
        temp_file.write(audio_data)
        temp_file.close()

        logger.info(f"Portfolio narration saved to: {temp_file.name}")
        return temp_file.name


class DebateAudioGenerator:
    """Generate multi-speaker audio for debate simulation."""

    def __init__(self, api_key: Optional[str] = None):
        """Initialise debate audio generator.

        Args:
            api_key: ElevenLabs API key (uses env var if not provided)
        """
        self.api_key = api_key or os.getenv("ELEVENLABS_API_KEY")
        if not self.api_key:
            logger.warning("ELEVENLABS_API_KEY not set - audio generation will fail")
            self.client = None
        else:
            self.client = AsyncElevenLabs(api_key=self.api_key)

        # Voice assignments for debate roles
        self.voices = {
            "bull": "pNInz6obpgDQGcFmaJgB",     # Adam - optimistic, energetic
            "bear": "XB0fDUnXU5powFXDhCwa",     # Charlotte - cautious, analytical
            "consensus": "JBFqnCBsd6RMkjVDRZzb", # George - neutral, professional
            "moderator": "EXAVITQu4vr4xnSDxMaL"  # Bella - clear, articulate
        }

    def is_available(self) -> bool:
        """Check if debate audio generator is available."""
        return self.client is not None

    async def generate_debate_audio(
        self,
        bull_case: str,
        bear_case: str,
        consensus: str,
        bull_confidence: Optional[float] = None,
        bear_confidence: Optional[float] = None,
        stance: Optional[str] = None
    ) -> str:
        """Generate multi-speaker debate simulation audio.

        Args:
            bull_case: Bull perspective text
            bear_case: Bear perspective text
            consensus: Consensus recommendation text
            bull_confidence: Bull confidence percentage
            bear_confidence: Bear confidence percentage
            stance: Final stance (bullish/bearish/neutral)

        Returns:
            Path to generated MP3 file with complete debate
        """
        if not self.is_available():
            raise RuntimeError("Debate audio generator not available")

        logger.info("Generating debate simulation audio")

        audio_segments = []

        # Introduction
        intro_text = "Advisory Council Debate. We will hear from the Bull researcher, followed by the Bear researcher, and conclude with a consensus recommendation."
        intro_audio = await self._generate_segment(intro_text, self.voices["moderator"])
        audio_segments.append(intro_audio)
        audio_segments.append(self._generate_pause(1.0))

        # Bull case
        bull_intro = f"Bull Case. Confidence level: {bull_confidence:.0f} percent. " if bull_confidence else "Bull Case. "
        bull_audio = await self._generate_segment(bull_intro + bull_case, self.voices["bull"])
        audio_segments.append(bull_audio)
        audio_segments.append(self._generate_pause(1.5))

        # Bear case
        bear_intro = f"Bear Case. Confidence level: {bear_confidence:.0f} percent. " if bear_confidence else "Bear Case. "
        bear_audio = await self._generate_segment(bear_intro + bear_case, self.voices["bear"])
        audio_segments.append(bear_audio)
        audio_segments.append(self._generate_pause(1.5))

        # Consensus
        consensus_intro = f"Consensus Recommendation. Final stance: {stance}. " if stance else "Consensus Recommendation. "
        consensus_audio = await self._generate_segment(consensus_intro + consensus, self.voices["consensus"])
        audio_segments.append(consensus_audio)

        # Combine all segments
        final_audio = b"".join(audio_segments)

        # Save to temporary file
        temp_file = tempfile.NamedTemporaryFile(
            delete=False,
            suffix=".mp3",
            prefix="debate_"
        )
        temp_file.write(final_audio)
        temp_file.close()

        logger.info(f"Debate audio saved to: {temp_file.name}")
        return temp_file.name

    async def _generate_segment(self, text: str, voice_id: str) -> bytes:
        """Generate audio segment with specific voice.

        Args:
            text: Text to convert
            voice_id: ElevenLabs voice ID

        Returns:
            Audio data as bytes
        """
        audio_generator = self.client.text_to_speech.convert(
            text=text,
            voice_id=voice_id,
            model_id="eleven_multilingual_v2",
            output_format="mp3_44100_128"
        )

        chunks = []
        async for chunk in audio_generator:
            chunks.append(chunk)

        return b"".join(chunks)

    def _generate_pause(self, duration: float) -> bytes:
        """Generate silence pause between segments.

        Args:
            duration: Pause duration in seconds

        Returns:
            Silence audio data
        """
        # Simple silence: MP3 frame with minimal data
        # For production, use proper silent MP3 frames
        sample_rate = 44100
        silence_samples = int(sample_rate * duration * 0.1)  # Simplified
        return b'\x00' * silence_samples