pip install livekit-plugins-faseeh

git clone https://github.com/actualize-ae/livekit-plugins-faseeh.git
cd livekit-plugins-faseeh
pip install -e .

# Munsit Configuration
MUNSIT_API_KEY=your_MUNSIT_API_KEY_here

# LiveKit Configuration
LIVEKIT_URL=wss://your-livekit-server.com
LIVEKIT_API_KEY=your_livekit_api_key
LIVEKIT_API_SECRET=your_livekit_api_secret

# LLM Configuration (for the agent's brain)
OPENAI_API_KEY=your_openai_api_key

# Optional: Deepgram for Arabic STT
DEEPGRAM_API_KEY=your_deepgram_api_key

"""
Simple Arabic Voice Assistant using Munsit TTS
"""
from dotenv import load_dotenv
from livekit import agents
from livekit.agents import AgentServer, AgentSession, Agent
from livekit.plugins import silero, deepgram, openai, faseeh

# Load environment variables
load_dotenv(".env.local")


class ArabicAssistant(Agent):
    """Arabic-speaking voice assistant"""

    def __init__(self) -> None:
        super().__init__(
            instructions="""أنت مساعد صوتي ذكي يتحدث العربية بطلاقة.
            مهمتك مساعدة المستخدمين بالإجابة على أسئلتهم بطريقة واضحة ومفيدة.
            كن ودوداً ومحترماً في تعاملك."""
        )


# Create the agent server
server = AgentServer()


@server.rtc_session()
async def my_agent(ctx: agents.JobContext):
    """Entry point for each agent session"""

    # Configure Munsit TTS
    faseeh_tts = faseeh.TTS(
        voice_id="ar-uae-male-1",           # Choose your voice
        model="faseeh-v1-preview",          # Use full model for best quality
        stability=0.75,                     # Balanced stability
        speed=1.0,                          # Normal speech speed (0.7-1.2)
    )

    # Create agent session with all components
    session = AgentSession(
        stt=openai.STT(),
        llm=openai.LLM(                     # Language model
            model="gpt-4o",
            temperature=0.7
        ),
        tts=faseeh_tts,                     # Munsit for Arabic speech
        vad=silero.VAD.load(),              # Voice activity detection
    )

    # Start the session
    await session.start(
        room=ctx.room,
        agent=ArabicAssistant()
    )

    # Generate initial greeting
    await session.generate_reply(
        instructions="رحب بالمستخدم باللغة العربية وقدم نفسك كمساعد ذكي جاهز للمساعدة."
    )


if __name__ == "__main__":
    agents.cli.run_app(server)

python arabic_agent.py dev

python arabic_agent.py start

faseeh_tts = faseeh.TTS(
    voice_id="ar-uae-male-1",  # Paste the copied voice ID here
    model="faseeh-v1-preview",
    speed=1.0                  # Speech speed (0.7-1.2)
)

# Full model - Best quality, slightly higher latency
faseeh_tts = faseeh.TTS(
    model="faseeh-v1-preview",
    stability=0.75,
    speed=1.0
)

# Mini model - Faster, lower latency, good quality
faseeh_tts = faseeh.TTS(
    model="faseeh-mini-v1-preview",
    stability=0.75,
    speed=1.0
)

# Dynamic voice - More expressive
faseeh_tts = faseeh.TTS(stability=0.4, speed=1.0)

# Consistent voice - Professional applications
faseeh_tts = faseeh.TTS(stability=0.9, speed=1.0)

# Slower speech - For clear enunciation
faseeh_tts = faseeh.TTS(stability=0.75, speed=0.8)

# Faster speech - For rapid delivery
faseeh_tts = faseeh.TTS(stability=0.75, speed=1.2)

# Start with one voice
faseeh_tts = faseeh.TTS(voice_id="ar-uae-male-1", speed=1.0)

# Switch to a different voice based on context
faseeh_tts.update_options(
    voice_id="ar-hijazi-female-2",
    stability=0.8,
    speed=1.0
)

class BilingualAssistant(Agent):
    def __init__(self) -> None:
        super().__init__(
            instructions="""You are a bilingual AI assistant fluent in both Arabic and English.
            Automatically detect the user's language and respond in the same language.
            If the user speaks Arabic, respond in Arabic using natural expressions.
            If the user speaks English, respond in English."""
        )

@server.rtc_session()
async def bilingual_agent(ctx: agents.JobContext):
    session = AgentSession(
        stt=deepgram.STT(
            language="multi",  # Multi-language support
            model="nova-2-general"
        ),
        llm=openai.LLM(model="gpt-4o-mini"),
        tts=faseeh.TTS(
            voice_id="ar-uae-male-1",
            model="faseeh-v1-preview",
            speed=1.0
        ),
        vad=silero.VAD.load(),
    )

    await session.start(room=ctx.room, agent=BilingualAssistant())

# Streaming is automatically enabled
# The agent will start speaking as soon as the first audio chunk is ready
faseeh_tts = faseeh.TTS(
    model="faseeh-mini-v1-preview",  # Mini model for even lower latency
    stability=0.6,
    speed=1.0
)

from livekit.agents.metrics import TTSMetrics
import logging

logger = logging.getLogger("faseeh_agent")

@session.on("metrics_collected")
def _on_metrics_collected(event):
    metrics = event.metrics

    if isinstance(metrics, TTSMetrics):
        logger.info(
            f"TTS Performance - "
            f"TTFB: {metrics.ttfb:.2f}s, "
            f"Audio Duration: {metrics.audio_duration:.2f}s"
        )

# Verify the key is set correctly
echo $MUNSIT_API_KEY