wx44wx/AI-VideoAssistant-Engine-V2
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Organize config

Browse Source
This commit is contained in:
Xin Wang
2026-02-25 15:52:55 +08:00
parent 2b2193557d
commit 8b9064f6e6
12 changed files with 1248 additions and 92 deletions
Download Patch File Download Diff File Expand all files Collapse all files

59
core/duplex_pipeline.py
View File

@@ -310,7 +310,12 @@ class DuplexPipeline:
def resolved_runtime_config(self) -> Dict[str, Any]:
"""Return current effective runtime configuration without secrets."""
llm_provider = str(self._runtime_llm.get("provider") or "openai").lower()
llm_provider = str(self._runtime_llm.get("provider") or settings.llm_provider).lower()
llm_base_url = (
self._runtime_llm.get("baseUrl")
or settings.llm_api_url
or self._default_llm_base_url(llm_provider)
)
tts_provider = str(self._runtime_tts.get("provider") or settings.tts_provider).lower()
asr_provider = str(self._runtime_asr.get("provider") or settings.asr_provider).lower()
output_mode = str(self._runtime_output.get("mode") or "").strip().lower()
@@ -323,18 +328,18 @@ class DuplexPipeline:
"llm": {
"provider": llm_provider,
"model": str(self._runtime_llm.get("model") or settings.llm_model),
"baseUrl": self._runtime_llm.get("baseUrl") or settings.openai_api_url,
"baseUrl": llm_base_url,
},
"asr": {
"provider": asr_provider,
"model": str(self._runtime_asr.get("model") or settings.siliconflow_asr_model),
"model": str(self._runtime_asr.get("model") or settings.asr_model or ""),
"interimIntervalMs": int(self._runtime_asr.get("interimIntervalMs") or settings.asr_interim_interval_ms),
"minAudioMs": int(self._runtime_asr.get("minAudioMs") or settings.asr_min_audio_ms),
},
"tts": {
"enabled": self._tts_output_enabled(),
"provider": tts_provider,
"model": str(self._runtime_tts.get("model") or settings.siliconflow_tts_model),
"model": str(self._runtime_tts.get("model") or settings.tts_model or ""),
"voice": str(self._runtime_tts.get("voice") or settings.tts_voice),
"speed": float(self._runtime_tts.get("speed") or settings.tts_speed),
},
@@ -452,6 +457,18 @@ class DuplexPipeline:
normalized = str(provider or "").strip().lower()
return normalized in {"openai_compatible", "openai-compatible", "siliconflow"}
@staticmethod
def _is_llm_provider_supported(provider: Any) -> bool:
normalized = str(provider or "").strip().lower()
return normalized in {"openai", "openai_compatible", "openai-compatible", "siliconflow"}
@staticmethod
def _default_llm_base_url(provider: Any) -> Optional[str]:
normalized = str(provider or "").strip().lower()
if normalized == "siliconflow":
return "https://api.siliconflow.cn/v1"
return None
def _tts_output_enabled(self) -> bool:
enabled = self._coerce_bool(self._runtime_tts.get("enabled"))
if enabled is not None:
@@ -527,12 +544,16 @@ class DuplexPipeline:
try:
# Connect LLM service
if not self.llm_service:
llm_api_key = self._runtime_llm.get("apiKey") or settings.openai_api_key
llm_base_url = self._runtime_llm.get("baseUrl") or settings.openai_api_url
llm_provider = (self._runtime_llm.get("provider") or settings.llm_provider).lower()
llm_api_key = self._runtime_llm.get("apiKey") or settings.llm_api_key
llm_base_url = (
self._runtime_llm.get("baseUrl")
or settings.llm_api_url
or self._default_llm_base_url(llm_provider)
)
llm_model = self._runtime_llm.get("model") or settings.llm_model
llm_provider = (self._runtime_llm.get("provider") or "openai").lower()
if llm_provider == "openai" and llm_api_key:
if self._is_llm_provider_supported(llm_provider) and llm_api_key:
self.llm_service = OpenAILLMService(
api_key=llm_api_key,
base_url=llm_base_url,
@@ -540,7 +561,7 @@ class DuplexPipeline:
knowledge_config=self._resolved_knowledge_config(),
)
else:
logger.warning("No OpenAI API key - using mock LLM")
logger.warning("LLM provider unsupported or API key missing - using mock LLM")
self.llm_service = MockLLMService()
if hasattr(self.llm_service, "set_knowledge_config"):
@@ -556,20 +577,22 @@ class DuplexPipeline:
if tts_output_enabled:
if not self.tts_service:
tts_provider = (self._runtime_tts.get("provider") or settings.tts_provider).lower()
tts_api_key = self._runtime_tts.get("apiKey") or settings.siliconflow_api_key
tts_api_key = self._runtime_tts.get("apiKey") or settings.tts_api_key
tts_api_url = self._runtime_tts.get("baseUrl") or settings.tts_api_url
tts_voice = self._runtime_tts.get("voice") or settings.tts_voice
tts_model = self._runtime_tts.get("model") or settings.siliconflow_tts_model
tts_model = self._runtime_tts.get("model") or settings.tts_model
tts_speed = float(self._runtime_tts.get("speed") or settings.tts_speed)
if self._is_openai_compatible_provider(tts_provider) and tts_api_key:
self.tts_service = OpenAICompatibleTTSService(
api_key=tts_api_key,
api_url=tts_api_url,
voice=tts_voice,
model=tts_model,
model=tts_model or "FunAudioLLM/CosyVoice2-0.5B",
sample_rate=settings.sample_rate,
speed=tts_speed
)
logger.info("Using OpenAI-compatible TTS service (SiliconFlow implementation)")
logger.info(f"Using OpenAI-compatible TTS service (provider={tts_provider})")
else:
self.tts_service = EdgeTTSService(
voice=tts_voice,
@@ -592,21 +615,23 @@ class DuplexPipeline:
# Connect ASR service
if not self.asr_service:
asr_provider = (self._runtime_asr.get("provider") or settings.asr_provider).lower()
asr_api_key = self._runtime_asr.get("apiKey") or settings.siliconflow_api_key
asr_model = self._runtime_asr.get("model") or settings.siliconflow_asr_model
asr_api_key = self._runtime_asr.get("apiKey") or settings.asr_api_key
asr_api_url = self._runtime_asr.get("baseUrl") or settings.asr_api_url
asr_model = self._runtime_asr.get("model") or settings.asr_model
asr_interim_interval = int(self._runtime_asr.get("interimIntervalMs") or settings.asr_interim_interval_ms)
asr_min_audio_ms = int(self._runtime_asr.get("minAudioMs") or settings.asr_min_audio_ms)
if self._is_openai_compatible_provider(asr_provider) and asr_api_key:
self.asr_service = OpenAICompatibleASRService(
api_key=asr_api_key,
model=asr_model,
api_url=asr_api_url,
model=asr_model or "FunAudioLLM/SenseVoiceSmall",
sample_rate=settings.sample_rate,
interim_interval_ms=asr_interim_interval,
min_audio_for_interim_ms=asr_min_audio_ms,
on_transcript=self._on_transcript_callback
)
logger.info("Using OpenAI-compatible ASR service (SiliconFlow implementation)")
logger.info(f"Using OpenAI-compatible ASR service (provider={asr_provider})")
else:
self.asr_service = BufferedASRService(
sample_rate=settings.sample_rate