231 lines
8.2 KiB
Python
231 lines
8.2 KiB
Python
from __future__ import annotations
|
|
|
|
import json
|
|
import os
|
|
from typing import Any
|
|
|
|
from sqlalchemy import select
|
|
|
|
from services.shared.core import utc_now_iso
|
|
from services.shared.models import (
|
|
VoiceTTSConfig,
|
|
VoiceTTSConfigOut,
|
|
VoiceTTSLanguageConfig,
|
|
VoiceTTSProviderConfig,
|
|
VoiceTTSVoiceOption,
|
|
)
|
|
from services.shared.sql_models import VoiceTTSSettingsRow
|
|
|
|
|
|
VOICE_TTS_SETTINGS_KEY = "global"
|
|
|
|
|
|
def _normalize_provider_name(value: str | None) -> str:
|
|
normalized = str(value or "yandex").strip().lower()
|
|
if normalized in {"speechkit", "yandex_speechkit"}:
|
|
return "yandex"
|
|
if normalized in {"11labs"}:
|
|
return "elevenlabs"
|
|
if normalized not in {"yandex", "elevenlabs", "openai"}:
|
|
return "yandex"
|
|
return normalized
|
|
|
|
|
|
def _openai_default_model() -> str:
|
|
return str(os.getenv("AI_VOICE_TTS_MODEL", "gpt-4o-mini-tts") or "gpt-4o-mini-tts").strip() or "gpt-4o-mini-tts"
|
|
|
|
|
|
def _openai_default_voice() -> str:
|
|
return str(os.getenv("AI_VOICE_TTS_VOICE", "alloy") or "alloy").strip() or "alloy"
|
|
|
|
|
|
def _yandex_default_voice(language: str) -> str:
|
|
normalized = str(language or "ru").strip().lower()
|
|
if normalized == "kz":
|
|
return str(os.getenv("AI_VOICE_TTS_YANDEX_KK_VOICE", "amira") or "amira").strip() or "amira"
|
|
return str(os.getenv("AI_VOICE_TTS_YANDEX_VOICE", "jane") or "jane").strip() or "jane"
|
|
|
|
|
|
def _elevenlabs_default_model() -> str:
|
|
return str(os.getenv("AI_VOICE_TTS_ELEVENLABS_MODEL_ID", "eleven_v3") or "eleven_v3").strip() or "eleven_v3"
|
|
|
|
|
|
def _elevenlabs_default_voice(language: str) -> str:
|
|
normalized = str(language or "ru").strip().lower()
|
|
if normalized == "kz":
|
|
return (
|
|
str(os.getenv("AI_VOICE_TTS_ELEVENLABS_KK_VOICE_ID", "nPczCjzI2devNBz1zQrb") or "nPczCjzI2devNBz1zQrb").strip()
|
|
or "nPczCjzI2devNBz1zQrb"
|
|
)
|
|
return (
|
|
str(os.getenv("AI_VOICE_TTS_ELEVENLABS_RU_VOICE_ID", "nPczCjzI2devNBz1zQrb") or "nPczCjzI2devNBz1zQrb").strip()
|
|
or "nPczCjzI2devNBz1zQrb"
|
|
)
|
|
|
|
|
|
def voice_tts_default_config() -> VoiceTTSConfig:
|
|
return VoiceTTSConfig(
|
|
provider=_normalize_provider_name(os.getenv("AI_VOICE_TTS_PROVIDER", "yandex")),
|
|
yandex=VoiceTTSProviderConfig(
|
|
ru=VoiceTTSLanguageConfig(voice=_yandex_default_voice("ru")),
|
|
kz=VoiceTTSLanguageConfig(voice=_yandex_default_voice("kz")),
|
|
),
|
|
elevenlabs=VoiceTTSProviderConfig(
|
|
ru=VoiceTTSLanguageConfig(
|
|
voice=_elevenlabs_default_voice("ru"),
|
|
model_id=_elevenlabs_default_model(),
|
|
language_code="ru",
|
|
),
|
|
kz=VoiceTTSLanguageConfig(
|
|
voice=_elevenlabs_default_voice("kz"),
|
|
model_id=_elevenlabs_default_model(),
|
|
language_code="kk",
|
|
),
|
|
),
|
|
openai=VoiceTTSProviderConfig(
|
|
ru=VoiceTTSLanguageConfig(
|
|
voice=_openai_default_voice(),
|
|
model_id=_openai_default_model(),
|
|
language_code="ru",
|
|
),
|
|
kz=VoiceTTSLanguageConfig(
|
|
voice=_openai_default_voice(),
|
|
model_id=_openai_default_model(),
|
|
language_code="kz",
|
|
),
|
|
),
|
|
)
|
|
|
|
|
|
def voice_tts_provider_options() -> list[str]:
|
|
return ["yandex", "elevenlabs", "openai"]
|
|
|
|
|
|
def voice_tts_voice_options() -> dict[str, dict[str, list[VoiceTTSVoiceOption]]]:
|
|
return {
|
|
"yandex": {
|
|
"ru": [
|
|
VoiceTTSVoiceOption(value="jane", label="Jane"),
|
|
VoiceTTSVoiceOption(value="oksana", label="Oksana"),
|
|
VoiceTTSVoiceOption(value="ermil", label="Ermil"),
|
|
],
|
|
"kz": [
|
|
VoiceTTSVoiceOption(value="amira", label="Amira"),
|
|
VoiceTTSVoiceOption(value="madi", label="Madi"),
|
|
VoiceTTSVoiceOption(value="saule", label="Saule"),
|
|
VoiceTTSVoiceOption(value="zhanar", label="Zhanar"),
|
|
],
|
|
},
|
|
"elevenlabs": {
|
|
"ru": [
|
|
VoiceTTSVoiceOption(value="nPczCjzI2devNBz1zQrb", label="Brian"),
|
|
],
|
|
"kz": [
|
|
VoiceTTSVoiceOption(value="nPczCjzI2devNBz1zQrb", label="Brian"),
|
|
],
|
|
},
|
|
"openai": {
|
|
"ru": [
|
|
VoiceTTSVoiceOption(value="alloy", label="Alloy"),
|
|
VoiceTTSVoiceOption(value="echo", label="Echo"),
|
|
VoiceTTSVoiceOption(value="fable", label="Fable"),
|
|
VoiceTTSVoiceOption(value="onyx", label="Onyx"),
|
|
VoiceTTSVoiceOption(value="nova", label="Nova"),
|
|
VoiceTTSVoiceOption(value="shimmer", label="Shimmer"),
|
|
],
|
|
"kz": [
|
|
VoiceTTSVoiceOption(value="alloy", label="Alloy"),
|
|
VoiceTTSVoiceOption(value="echo", label="Echo"),
|
|
VoiceTTSVoiceOption(value="fable", label="Fable"),
|
|
VoiceTTSVoiceOption(value="onyx", label="Onyx"),
|
|
VoiceTTSVoiceOption(value="nova", label="Nova"),
|
|
VoiceTTSVoiceOption(value="shimmer", label="Shimmer"),
|
|
],
|
|
},
|
|
}
|
|
|
|
|
|
def _settings_row(session) -> VoiceTTSSettingsRow | None:
|
|
return session.execute(
|
|
select(VoiceTTSSettingsRow).where(VoiceTTSSettingsRow.settings_key == VOICE_TTS_SETTINGS_KEY)
|
|
).scalar_one_or_none()
|
|
|
|
|
|
def _normalize_config_payload(raw: Any) -> VoiceTTSConfig:
|
|
if isinstance(raw, VoiceTTSConfig):
|
|
return raw
|
|
if not isinstance(raw, dict):
|
|
raise ValueError("Voice TTS config payload must be an object")
|
|
return VoiceTTSConfig.model_validate(raw)
|
|
|
|
|
|
def load_voice_tts_config(session) -> VoiceTTSConfigOut:
|
|
default_config = voice_tts_default_config()
|
|
row = _settings_row(session)
|
|
if row is None:
|
|
return VoiceTTSConfigOut(
|
|
config=default_config,
|
|
updated_at=None,
|
|
source="defaults",
|
|
provider_options=voice_tts_provider_options(),
|
|
voice_options=voice_tts_voice_options(),
|
|
)
|
|
try:
|
|
parsed = json.loads(row.config_json or "{}")
|
|
config = _normalize_config_payload(parsed)
|
|
return VoiceTTSConfigOut(
|
|
config=config,
|
|
updated_at=row.updated_at,
|
|
source="database",
|
|
provider_options=voice_tts_provider_options(),
|
|
voice_options=voice_tts_voice_options(),
|
|
)
|
|
except Exception:
|
|
return VoiceTTSConfigOut(
|
|
config=default_config,
|
|
updated_at=row.updated_at,
|
|
source="defaults",
|
|
provider_options=voice_tts_provider_options(),
|
|
voice_options=voice_tts_voice_options(),
|
|
)
|
|
|
|
|
|
def load_effective_voice_tts_config(session) -> VoiceTTSConfig:
|
|
config = load_voice_tts_config(session).config
|
|
config.provider = _normalize_provider_name(config.provider)
|
|
return config
|
|
|
|
|
|
def save_voice_tts_config(session, config_payload: VoiceTTSConfig | dict[str, Any]) -> VoiceTTSConfigOut:
|
|
config = _normalize_config_payload(config_payload)
|
|
config.provider = _normalize_provider_name(config.provider)
|
|
row = _settings_row(session)
|
|
now = utc_now_iso()
|
|
if row is None:
|
|
row = VoiceTTSSettingsRow(
|
|
settings_key=VOICE_TTS_SETTINGS_KEY,
|
|
config_json="{}",
|
|
updated_at=now,
|
|
)
|
|
session.add(row)
|
|
row.config_json = json.dumps(config.model_dump(), ensure_ascii=False)
|
|
row.updated_at = now
|
|
session.commit()
|
|
session.refresh(row)
|
|
return VoiceTTSConfigOut(
|
|
config=config,
|
|
updated_at=row.updated_at,
|
|
source="database",
|
|
provider_options=voice_tts_provider_options(),
|
|
voice_options=voice_tts_voice_options(),
|
|
)
|
|
|
|
|
|
def voice_tts_language_config(config: VoiceTTSConfig, provider: str, language: str) -> VoiceTTSLanguageConfig:
|
|
normalized_provider = str(provider or config.provider or "yandex").strip().lower()
|
|
provider_config = getattr(config, normalized_provider, None)
|
|
if not isinstance(provider_config, VoiceTTSProviderConfig):
|
|
provider_config = getattr(voice_tts_default_config(), normalized_provider)
|
|
normalized_language = "kz" if str(language or "").strip().lower() == "kz" else "ru"
|
|
return provider_config.kz if normalized_language == "kz" else provider_config.ru
|