Multi-pass warmup and smaller chunk_size to fix HA timeout

torch.compile with dynamic=True still specializes per shape family on
first call. The warmup was running one text length, leaving real requests
to JIT-compile their own shapes (15-22s for first chunk). HA freezes
because it gets no AudioChunk for 22 seconds.

Fix:
- Run 3 warmup passes (short/medium/long text) so torch.compile builds
  a dynamic shape graph covering the range HA actually sends. Real
  requests then hit a cached compilation and synthesize in 3-8s.
- Reduce default chunk_size from 300 to 120 chars so the first text
  chunk is shorter, producing faster synthesis and earlier first audio.

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

config.py

@@ -29,7 +29,7 @@ DEFAULTS = {
     "wyoming": {
         "host": "0.0.0.0",
         "port": 10200,
-        "chunk_size": 300,
+        "chunk_size": 120,
     },
     "paths": {
         "model_cache": "/app/hf_cache",

config.yaml

@@ -23,7 +23,7 @@ wyoming:
   host: "0.0.0.0"
   port: 10200
   # Max characters per synthesis chunk (split at sentence boundaries)
-  chunk_size: 300
+  chunk_size: 120
 
 paths:
   model_cache: /app/hf_cache

main.py

@@ -18,21 +18,33 @@ logging.basicConfig(
 logger = logging.getLogger(__name__)
 
 
+_WARMUP_TEXTS = [
+    # Short: covers brief HA notifications (lights on/off, etc.)
+    "Okay.",
+    # Medium: covers typical HA announcements
+    "The front door is open. Please close it.",
+    # Long: covers longer TTS requests and pre-compiles dynamic shape graph
+    (
+        "This is a warmup synthesis to pre-compile neural network kernels "
+        "for longer text lengths used in Home Assistant announcements and notifications."
+    ),
+]
+
+
 def _warmup(voices: dict) -> None:
     from wyoming_voices import resolve_voice
     audio_prompt = resolve_voice(None, voices) if voices else None
-    logger.info("Running warmup synthesis to populate MIOpen kernel cache...")
+    logger.info(
-    try:
+        f"Running {len(_WARMUP_TEXTS)}-pass warmup to pre-compile torch kernels "
-        engine.synthesize(
+        "for short, medium, and long text lengths..."
-            text=(
+    )
-                "This is a warmup synthesis request used to pre-compile neural network kernels "
+    for i, text in enumerate(_WARMUP_TEXTS, 1):
-                "for typical text lengths, so that the first real request runs at full speed."
+        try:
-            ),
+            engine.synthesize(text=text, audio_prompt_path=audio_prompt)
-            audio_prompt_path=audio_prompt,
+            logger.info(f"Warmup pass {i}/{len(_WARMUP_TEXTS)} complete")
-        )
+        except Exception:
-        logger.info("Warmup complete — MIOpen cache populated")
+            logger.warning(f"Warmup pass {i} failed (non-fatal)", exc_info=True)
-    except Exception:
+    logger.info("Warmup complete")
-        logger.warning("Warmup synthesis failed (non-fatal)", exc_info=True)
 
 
 async def main() -> None: