Initial import with GitLab CI/CD and registry deploy flow

scripts/materialize_tts_prompt.py

@@ -0,0 +1,72 @@
+from __future__ import annotations
+
+import argparse
+import os
+from pathlib import Path
+import sys
+import tempfile
+
+
+ROOT = Path(__file__).resolve().parents[1]
+if str(ROOT) not in sys.path:
+    sys.path.insert(0, str(ROOT))
+
+from services.ai_voice_runtime_service.audiosocket import pcm16le_to_wav_bytes, resample_pcm16le
+from services.ai_voice_runtime_service.providers.tts import build_tts_provider
+
+
+def _parse_args() -> argparse.Namespace:
+    default_provider = os.getenv("AI_VOICE_TTS_PROVIDER", "openai").strip() or "openai"
+    parser = argparse.ArgumentParser(
+        description="Generate TTS once, cache it on the server, and materialize a WAV prompt file.",
+    )
+    parser.add_argument("--text", required=True, help="Text to synthesize.")
+    parser.add_argument("--output", required=True, help="Target WAV path.")
+    parser.add_argument("--language", default=None, help="Optional language hint.")
+    parser.add_argument(
+        "--provider",
+        default=default_provider,
+        help="TTS provider name. Defaults to AI_VOICE_TTS_PROVIDER or openai.",
+    )
+    parser.add_argument("--sample-rate", type=int, default=8000, help="Target WAV sample rate. Defaults to 8000.")
+    return parser.parse_args()
+
+
+def _write_atomic(target: Path, payload: bytes) -> None:
+    target.parent.mkdir(parents=True, exist_ok=True)
+    temp_path: str | None = None
+    try:
+        with tempfile.NamedTemporaryFile(dir=target.parent, delete=False, suffix=".tmp") as handle:
+            handle.write(payload)
+            temp_path = handle.name
+        Path(temp_path).replace(target)
+        target.chmod(0o644)
+    finally:
+        if temp_path:
+            try:
+                Path(temp_path).unlink(missing_ok=True)
+            except OSError:
+                pass
+
+
+def main() -> int:
+    args = _parse_args()
+    provider = build_tts_provider(args.provider)
+    synthesis = provider.synthesize(args.text, language=args.language)
+    pcm_bytes = resample_pcm16le(
+        synthesis.audio_bytes,
+        input_rate_hz=synthesis.sample_rate_hz,
+        output_rate_hz=max(args.sample_rate, 1),
+    )
+    wav_bytes = pcm16le_to_wav_bytes(
+        pcm_bytes,
+        sample_rate_hz=max(args.sample_rate, 1),
+    )
+    output_path = Path(args.output).expanduser()
+    _write_atomic(output_path, wav_bytes)
+    print(output_path)
+    return 0
+
+
+if __name__ == "__main__":
+    raise SystemExit(main())