feat: initial commit — VoiceCoach real-time voice coaching pipeline

Services: Pipecat agent (LiveKit + Claude + Whisper STT + Piper TTS),
SpeechBrain analysis service, Piper TTS service, web client (nginx).
Includes LiveKit server config and docker-compose orchestration.

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

piper-service/Dockerfile

@@ -0,0 +1,27 @@
+FROM python:3.11-slim
+
+RUN apt-get update && apt-get install -y \
+    libsndfile1 \
+    wget \
+    && rm -rf /var/lib/apt/lists/*
+
+WORKDIR /app
+
+COPY requirements.txt .
+RUN pip install --no-cache-dir -r requirements.txt
+
+# Download Piper voice model at build time from HuggingFace (lessac medium - good quality, fast on CPU)
+RUN mkdir -p /app/models && \
+    wget -q "https://huggingface.co/rhasspy/piper-voices/resolve/main/en/en_US/lessac/medium/en_US-lessac-medium.onnx?download=true" \
+        -O /app/models/en_US-lessac-medium.onnx && \
+    wget -q "https://huggingface.co/rhasspy/piper-voices/resolve/main/en/en_US/lessac/medium/en_US-lessac-medium.onnx.json?download=true" \
+        -O /app/models/en_US-lessac-medium.onnx.json
+
+COPY . .
+
+EXPOSE 8003
+
+HEALTHCHECK --interval=30s --timeout=10s --start-period=30s \
+    CMD python -c "import httpx; httpx.get('http://localhost:8003/health').raise_for_status()"
+
+CMD ["uvicorn", "main:app", "--host", "0.0.0.0", "--port", "8003", "--workers", "1"]

piper-service/main.py

@@ -0,0 +1,72 @@
+"""
+Piper TTS HTTP service.
+Self-hosted text-to-speech, internal network only.
+"""
+import io
+import logging
+import os
+import wave
+from contextlib import asynccontextmanager
+
+from fastapi import FastAPI, HTTPException
+from fastapi.responses import Response
+from pydantic import BaseModel
+import piper
+
+logging.basicConfig(level=logging.INFO)
+logger = logging.getLogger(__name__)
+
+MODEL_PATH = os.getenv("PIPER_MODEL", "/app/models/en_US-lessac-medium.onnx")
+voice = None
+
+
+@asynccontextmanager
+async def lifespan(app: FastAPI):
+    global voice
+    logger.info(f"Loading Piper model: {MODEL_PATH}")
+    voice = piper.PiperVoice.load(MODEL_PATH)
+    logger.info("Piper TTS ready.")
+    yield
+    logger.info("Shutting down.")
+
+
+app = FastAPI(title="Piper TTS Service", lifespan=lifespan)
+
+
+class TTSRequest(BaseModel):
+    text: str
+    speaker_id: int = 0
+    length_scale: float = 1.0
+    sentence_silence: float = 0.2
+
+
+@app.get("/health")
+async def health():
+    if voice is None:
+        raise HTTPException(status_code=503, detail="Model not loaded")
+    return {"status": "ok", "model": os.path.basename(MODEL_PATH)}
+
+
+@app.post("/synthesize")
+async def synthesize(req: TTSRequest):
+    if voice is None:
+        raise HTTPException(status_code=503, detail="Model not loaded")
+
+    if not req.text.strip():
+        raise HTTPException(status_code=400, detail="Empty text")
+
+    buf = io.BytesIO()
+    with wave.open(buf, "wb") as wf:
+        voice.synthesize(
+            req.text,
+            wf,
+            speaker_id=req.speaker_id,
+            length_scale=req.length_scale,
+            sentence_silence=req.sentence_silence,
+        )
+
+    return Response(
+        content=buf.getvalue(),
+        media_type="audio/wav",
+        headers={"Content-Disposition": "inline; filename=speech.wav"},
+    )

piper-service/requirements.txt

@@ -0,0 +1,5 @@
+fastapi==0.115.0
+uvicorn[standard]==0.30.0
+piper-tts==1.2.0
+httpx==0.27.0
+numpy==1.26.4