음성 재생과 운영 배포 정리

apps/api/app/services/voice.py

@@ -21,11 +21,12 @@ from __future__ import annotations
 import re
 from dataclasses import dataclass
 from pathlib import Path
-from typing import AsyncIterator, Optional
+from typing import Any, AsyncIterator, Mapping, Optional
 
 import httpx
 
 from ..config import settings
+from ..paths import repo_root, repo_path
 
 # ════════════════════════════════════════════════════════════════════════════
 # OpenAI 음성 엔드포인트/모델 상수
@@ -50,10 +51,9 @@ TTS_RESPONSE_FORMAT = "mp3"
 # End-of-turn readiness default for cascaded STT providers.
 EOT_SILENCE_THRESHOLD_MS = 1200
 
-_REPO_ROOT = Path(__file__).resolve().parents[4]
 POC_SAMPLE_TTS_PRESET = "soft-young-fem"
 POC_SAMPLE_TTS_DEFAULT_DIR = (
-    _REPO_ROOT / "docs" / "voice-art" / "p1-seoyeon-higgs-v3-20260627"
+    repo_path("docs", "voice-art", "p1-seoyeon-higgs-v3-20260627")
 )
 POC_SAMPLE_TTS_CHUNK_SIZE = 4096
 _POC_SAMPLE_TTS_DEFAULT_SAMPLE = "p1_seoyeon_01_depressed_slow"
@@ -190,6 +190,51 @@ def resolve_voice(
     )
 
 
+def resolve_voice_from_map(
+    *,
+    provider: str,
+    voice_id: str,
+    base_params: Mapping[str, Any] | None,
+    persona_code: Optional[str] = None,
+) -> VoicePreset | None:
+    """DB persona_voice_map row -> live OpenAI VoicePreset.
+
+    provider-agnostic rows are allowed in the catalog, but this service only
+    knows how to send OpenAI TTS. Unsupported providers return None so callers
+    can fall back to the existing preset resolver.
+    """
+    if provider.strip().lower() != "openai":
+        return None
+
+    fallback = resolve_voice(persona_code=persona_code)
+    params = dict(base_params or {})
+    preset = _clean_optional_text(params.get("preset")) or fallback.preset
+    mapped_voice = _clean_optional_text(params.get("openai_voice"))
+    voice_id_value = _clean_optional_text(voice_id)
+    if not mapped_voice and voice_id_value in _OPENAI_VOICES:
+        mapped_voice = voice_id_value
+    if not mapped_voice:
+        mapped_voice = PRESET_TO_OPENAI_VOICE.get(preset, fallback.openai_voice)
+    if mapped_voice not in _OPENAI_VOICES:
+        mapped_voice = fallback.openai_voice
+    if mapped_voice not in _OPENAI_VOICES:
+        mapped_voice = DEFAULT_OPENAI_VOICE
+
+    rate = PRESET_RATE.get(preset, fallback.rate)
+    if "rate" in params:
+        try:
+            rate = float(params["rate"])
+        except (TypeError, ValueError):
+            rate = fallback.rate
+
+    return VoicePreset(
+        preset=preset,
+        openai_voice=mapped_voice,
+        rate=rate,
+        instructions=_clean_optional_text(params.get("instructions")),
+    )
+
+
 # 비언어 지문 패턴: (…)·（…）·[…]·【…】. 내담자 발화의 무대지시(고개 끄덕/한숨/침묵 등).
 _STAGE_DIRECTION_RE = re.compile(r"[\(（\[【][^\)）\]】]*[\)）\]】]")
 
@@ -291,7 +336,7 @@ class VoiceService:
         )
         sample_dir = Path(sample_dir_value)
         if not sample_dir.is_absolute():
-            sample_dir = _REPO_ROOT / sample_dir
+            sample_dir = repo_root() / sample_dir
         self._poc_sample_tts_dir = sample_dir
         self._client: Optional[httpx.AsyncClient] = None
 
@@ -507,6 +552,13 @@ def _nonnegative_int(value: object) -> int:
         return 0
 
 
+def _clean_optional_text(value: object) -> str | None:
+    if value is None:
+        return None
+    text = str(value).strip()
+    return text or None
+
+
 # 앱 전역 싱글톤 (main lifespan 이 startup/shutdown — Foundation 이 관리하거나
 # 라우트가 lazy 사용). engine_client 패턴과 동일.
 voice_service = VoiceService()
@@ -521,6 +573,7 @@ __all__ = [
     "VoiceService",
     "voice_service",
     "resolve_voice",
+    "resolve_voice_from_map",
     "build_tts_payload",
     "assess_end_of_turn",
     "EOT_SILENCE_THRESHOLD_MS",