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

Add tts/text output schema

Browse Source
This commit is contained in:
Xin Wang
2026-02-11 09:50:46 +08:00
parent 2d7fc2b700
commit 15523d9ec2
13 changed files with 219 additions and 50 deletions
Download Patch File Download Diff File Expand all files Collapse all files

28
api/app/db.py
View File

@@ -1,4 +1,4 @@
from sqlalchemy import create_engine
from sqlalchemy import create_engine, text
from sqlalchemy.orm import sessionmaker, DeclarativeBase
import os
@@ -14,6 +14,32 @@ class Base(DeclarativeBase):
pass
def ensure_schema_compatibility() -> None:
"""Best-effort lightweight migrations for SQLite deployments."""
if engine.dialect.name != "sqlite":
return
with engine.begin() as conn:
columns = {
row[1]
for row in conn.execute(text("PRAGMA table_info(assistants)"))
}
if "voice_output_enabled" not in columns:
conn.execute(
text(
"ALTER TABLE assistants "
"ADD COLUMN voice_output_enabled BOOLEAN DEFAULT 1"
)
)
conn.execute(
text(
"UPDATE assistants "
"SET voice_output_enabled = 1 "
"WHERE voice_output_enabled IS NULL"
)
)
def get_db():
db = SessionLocal()
try:

3
api/app/main.py
View File

@@ -3,7 +3,7 @@ from fastapi.middleware.cors import CORSMiddleware
from contextlib import asynccontextmanager
import os
from .db import Base, engine
from .db import Base, engine, ensure_schema_compatibility
from .routers import assistants, voices, workflows, history, knowledge, llm, asr, tools
@@ -11,6 +11,7 @@ from .routers import assistants, voices, workflows, history, knowledge, llm, asr
async def lifespan(app: FastAPI):
# 启动时创建表
Base.metadata.create_all(bind=engine)
ensure_schema_compatibility()
yield

1
api/app/models.py
View File

@@ -112,6 +112,7 @@ class Assistant(Base):
prompt: Mapped[str] = mapped_column(Text, default="")
knowledge_base_id: Mapped[Optional[str]] = mapped_column(String(64), nullable=True)
language: Mapped[str] = mapped_column(String(16), default="zh")
voice_output_enabled: Mapped[bool] = mapped_column(default=True)
voice: Mapped[Optional[str]] = mapped_column(String(64), nullable=True)
speed: Mapped[float] = mapped_column(Float, default=1.0)
hotwords: Mapped[dict] = mapped_column(JSON, default=list)

10
api/app/routers/assistants.py
View File

@@ -21,6 +21,7 @@ def _resolve_runtime_metadata(db: Session, assistant: Assistant) -> dict:
metadata = {
"systemPrompt": assistant.prompt or "",
"greeting": assistant.opener or "",
"output": {"mode": "audio" if assistant.voice_output_enabled else "text"},
"services": {},
}
warnings = []
@@ -49,11 +50,14 @@ def _resolve_runtime_metadata(db: Session, assistant: Assistant) -> dict:
else:
warnings.append(f"ASR model not found: {assistant.asr_model_id}")
if assistant.voice:
if not assistant.voice_output_enabled:
metadata["services"]["tts"] = {"enabled": False}
elif assistant.voice:
voice = db.query(Voice).filter(Voice.id == assistant.voice).first()
if voice:
tts_provider = "siliconflow" if _is_siliconflow_vendor(voice.vendor) else "edge"
metadata["services"]["tts"] = {
"enabled": True,
"provider": tts_provider,
"model": voice.model,
"apiKey": voice.api_key if tts_provider == "siliconflow" else None,
@@ -63,6 +67,7 @@ def _resolve_runtime_metadata(db: Session, assistant: Assistant) -> dict:
else:
# Keep assistant.voice as direct voice identifier fallback
metadata["services"]["tts"] = {
"enabled": True,
"voice": assistant.voice,
"speed": assistant.speed or 1.0,
}
@@ -98,6 +103,7 @@ def assistant_to_dict(assistant: Assistant) -> dict:
"prompt": assistant.prompt or "",
"knowledgeBaseId": assistant.knowledge_base_id,
"language": assistant.language,
"voiceOutputEnabled": assistant.voice_output_enabled,
"voice": assistant.voice,
"speed": assistant.speed,
"hotwords": assistant.hotwords or [],
@@ -120,6 +126,7 @@ def _apply_assistant_update(assistant: Assistant, update_data: dict) -> None:
"knowledgeBaseId": "knowledge_base_id",
"interruptionSensitivity": "interruption_sensitivity",
"configMode": "config_mode",
"voiceOutputEnabled": "voice_output_enabled",
"apiUrl": "api_url",
"apiKey": "api_key",
"llmModelId": "llm_model_id",
@@ -180,6 +187,7 @@ def create_assistant(data: AssistantCreate, db: Session = Depends(get_db)):
prompt=data.prompt,
knowledge_base_id=data.knowledgeBaseId,
language=data.language,
voice_output_enabled=data.voiceOutputEnabled,
voice=data.voice,
speed=data.speed,
hotwords=data.hotwords,

2
api/app/schemas.py
View File

@@ -268,6 +268,7 @@ class AssistantBase(BaseModel):
prompt: str = ""
knowledgeBaseId: Optional[str] = None
language: str = "zh"
voiceOutputEnabled: bool = True
voice: Optional[str] = None
speed: float = 1.0
hotwords: List[str] = []
@@ -293,6 +294,7 @@ class AssistantUpdate(BaseModel):
prompt: Optional[str] = None
knowledgeBaseId: Optional[str] = None
language: Optional[str] = None
voiceOutputEnabled: Optional[bool] = None
voice: Optional[str] = None
speed: Optional[float] = None
hotwords: Optional[List[str]] = None

1
api/tests/conftest.py
View File

@@ -85,6 +85,7 @@ def sample_assistant_data():
"opener": "Hello, welcome!",
"prompt": "You are a helpful assistant.",
"language": "zh",
"voiceOutputEnabled": True,
"speed": 1.0,
"hotwords": ["test", "hello"],
"tools": [],

17
api/tests/test_assistants.py
View File

@@ -23,6 +23,7 @@ class TestAssistantAPI:
assert data["opener"] == sample_assistant_data["opener"]
assert data["prompt"] == sample_assistant_data["prompt"]
assert data["language"] == sample_assistant_data["language"]
assert data["voiceOutputEnabled"] is True
assert "id" in data
assert data["callCount"] == 0
@@ -61,7 +62,8 @@ class TestAssistantAPI:
update_data = {
"name": "Updated Assistant",
"prompt": "You are an updated assistant.",
"speed": 1.5
"speed": 1.5,
"voiceOutputEnabled": False,
}
response = client.put(f"/api/assistants/{assistant_id}", json=update_data)
assert response.status_code == 200
@@ -69,6 +71,7 @@ class TestAssistantAPI:
assert data["name"] == "Updated Assistant"
assert data["prompt"] == "You are an updated assistant."
assert data["speed"] == 1.5
assert data["voiceOutputEnabled"] is False
def test_delete_assistant(self, client, sample_assistant_data):
"""Test deleting an assistant"""
@@ -210,3 +213,15 @@ class TestAssistantAPI:
assert metadata["services"]["llm"]["model"] == sample_llm_model_data["model_name"]
assert metadata["services"]["asr"]["model"] == sample_asr_model_data["model_name"]
assert metadata["services"]["tts"]["voice"] == sample_voice_data["voice_key"]
def test_runtime_config_text_mode_when_voice_output_disabled(self, client, sample_assistant_data):
sample_assistant_data["voiceOutputEnabled"] = False
assistant_resp = client.post("/api/assistants", json=sample_assistant_data)
assert assistant_resp.status_code == 200
assistant_id = assistant_resp.json()["id"]
runtime_resp = client.get(f"/api/assistants/{assistant_id}/runtime-config")
assert runtime_resp.status_code == 200
metadata = runtime_resp.json()["sessionStartMetadata"]
assert metadata["output"]["mode"] == "text"
assert metadata["services"]["tts"]["enabled"] is False

58
engine/core/duplex_pipeline.py
View File

@@ -211,6 +211,7 @@ class DuplexPipeline:
self._runtime_llm: Dict[str, Any] = {}
self._runtime_asr: Dict[str, Any] = {}
self._runtime_tts: Dict[str, Any] = {}
self._runtime_output: Dict[str, Any] = {}
self._runtime_system_prompt: Optional[str] = None
self._runtime_greeting: Optional[str] = None
self._runtime_knowledge: Dict[str, Any] = {}
@@ -257,6 +258,9 @@ class DuplexPipeline:
self._runtime_asr = services["asr"]
if isinstance(services.get("tts"), dict):
self._runtime_tts = services["tts"]
output = metadata.get("output") or {}
if isinstance(output, dict):
self._runtime_output = output
knowledge_base_id = metadata.get("knowledgeBaseId")
if knowledge_base_id is not None:
@@ -283,6 +287,31 @@ class DuplexPipeline:
if self.llm_service and hasattr(self.llm_service, "set_tool_schemas"):
self.llm_service.set_tool_schemas(self._resolved_tool_schemas())
@staticmethod
def _coerce_bool(value: Any) -> Optional[bool]:
if isinstance(value, bool):
return value
if isinstance(value, (int, float)):
return bool(value)
if isinstance(value, str):
normalized = value.strip().lower()
if normalized in {"1", "true", "yes", "on", "enabled"}:
return True
if normalized in {"0", "false", "no", "off", "disabled"}:
return False
return None
def _tts_output_enabled(self) -> bool:
enabled = self._coerce_bool(self._runtime_tts.get("enabled"))
if enabled is not None:
return enabled
output_mode = str(self._runtime_output.get("mode") or "").strip().lower()
if output_mode in {"text", "text_only", "text-only"}:
return False
return True
async def start(self) -> None:
"""Start the pipeline and connect services."""
try:
@@ -311,7 +340,10 @@ class DuplexPipeline:
await self.llm_service.connect()
# Connect TTS service
tts_output_enabled = self._tts_output_enabled()
# Connect TTS service only when audio output is enabled.
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
@@ -343,6 +375,9 @@ class DuplexPipeline:
sample_rate=settings.sample_rate
)
await self.tts_service.connect()
else:
self.tts_service = None
logger.info("TTS output disabled by runtime metadata")
# Connect ASR service
if not self.asr_service:
@@ -375,7 +410,7 @@ class DuplexPipeline:
self._outbound_task = asyncio.create_task(self._outbound_loop())
# Speak greeting if configured
if self.conversation.greeting:
if self.conversation.greeting and tts_output_enabled:
await self._speak(self.conversation.greeting)
except Exception as e:
@@ -932,7 +967,7 @@ class DuplexPipeline:
pending_punctuation = sentence
continue
if not self._interrupt_event.is_set():
if self._tts_output_enabled() and not self._interrupt_event.is_set():
if not first_audio_sent:
await self._send_event(
{
@@ -952,7 +987,12 @@ class DuplexPipeline:
)
remaining_text = f"{pending_punctuation}{sentence_buffer}".strip()
if remaining_text and has_spoken_content(remaining_text) and not self._interrupt_event.is_set():
if (
self._tts_output_enabled()
and remaining_text
and has_spoken_content(remaining_text)
and not self._interrupt_event.is_set()
):
if not first_audio_sent:
await self._send_event(
{
@@ -1066,7 +1106,10 @@ class DuplexPipeline:
fade_in_ms: Fade-in duration for sentence start chunks
fade_out_ms: Fade-out duration for sentence end chunks
"""
if not text.strip() or self._interrupt_event.is_set():
if not self._tts_output_enabled():
return
if not text.strip() or self._interrupt_event.is_set() or not self.tts_service:
return
logger.info(f"[TTS] split sentence: {text!r}")
@@ -1153,7 +1196,10 @@ class DuplexPipeline:
Args:
text: Text to speak
"""
if not text.strip():
if not self._tts_output_enabled():
return
if not text.strip() or not self.tts_service:
return
try:

8
engine/docs/ws_v1_schema.md
View File

@@ -53,6 +53,9 @@ Rules:
},
"metadata": {
"client": "web-debug",
"output": {
"mode": "audio"
},
"systemPrompt": "You are concise.",
"greeting": "Hi, how can I help?",
"services": {
@@ -70,6 +73,7 @@ Rules:
"minAudioMs": 300
},
"tts": {
"enabled": true,
"provider": "siliconflow",
"model": "FunAudioLLM/CosyVoice2-0.5B",
"apiKey": "sf-...",
@@ -83,6 +87,10 @@ Rules:
`metadata.services` is optional. If omitted, server defaults to environment configuration.
Text-only mode:
- Set `metadata.output.mode = "text"` OR `metadata.services.tts.enabled = false`.
- In this mode server still sends `assistant.response.delta/final`, but will not emit audio frames or `output.audio.start/end`.
### `input.text`
```json

30
engine/tests/test_tool_call_flow.py
View File

@@ -125,6 +125,36 @@ async def test_turn_without_tool_keeps_streaming(monkeypatch):
assert "assistant.tool_call" not in event_types
@pytest.mark.asyncio
@pytest.mark.parametrize(
"metadata",
[
{"output": {"mode": "text"}},
{"services": {"tts": {"enabled": False}}},
],
)
async def test_text_output_mode_skips_audio_events(monkeypatch, metadata):
pipeline, events = _build_pipeline(
monkeypatch,
[
[
LLMStreamEvent(type="text_delta", text="hello "),
LLMStreamEvent(type="text_delta", text="world."),
LLMStreamEvent(type="done"),
]
],
)
pipeline.apply_runtime_overrides(metadata)
await pipeline._handle_turn("hi")
event_types = [e.get("type") for e in events]
assert "assistant.response.delta" in event_types
assert "assistant.response.final" in event_types
assert "output.audio.start" not in event_types
assert "output.audio.end" not in event_types
@pytest.mark.asyncio
async def test_turn_with_tool_call_then_results(monkeypatch):
pipeline, events = _build_pipeline(

51
web/pages/Assistants.tsx
View File

@@ -118,6 +118,7 @@ export const AssistantsPage: React.FC = () => {
prompt: '',
knowledgeBaseId: '',
language: 'zh',
voiceOutputEnabled: true,
voice: voices[0]?.id || '',
speed: 1,
hotwords: [],
@@ -531,6 +532,7 @@ export const AssistantsPage: React.FC = () => {
placeholder="设定小助手的人设、语气、行为规范以及业务逻辑..."
/>
</div>
</div>
)}
@@ -624,15 +626,32 @@ export const AssistantsPage: React.FC = () => {
</p>
</div>
<div className="space-y-2">
<label className="text-sm font-medium text-white flex items-center">
<Volume2 className="w-4 h-4 mr-2 text-primary" /> 启用语音输出
</label>
<label className="flex h-12 items-center justify-between rounded-xl border border-white/10 bg-white/5 px-4 text-sm">
<span className="text-foreground">TTS 输出</span>
<input
type="checkbox"
checked={selectedAssistant.voiceOutputEnabled !== false}
onChange={(e) => updateAssistant('voiceOutputEnabled', e.target.checked)}
className="accent-primary"
/>
</label>
<p className="text-xs text-muted-foreground">关闭后将进入纯文本输出模式，不会产生语音音频。</p>
</div>
<div className="space-y-2">
<label className="text-sm font-medium text-white flex items-center">
<Volume2 className="w-4 h-4 mr-2 text-primary"/> 选择音色 (From Voice Library)
</label>
<div className="relative group">
<select
className="flex h-12 w-full rounded-xl border border-white/10 bg-white/5 px-4 py-1 text-sm shadow-sm transition-all focus-visible:outline-none focus-visible:ring-1 focus-visible:ring-primary/50 [&>option]:bg-card text-foreground appearance-none cursor-pointer"
className="flex h-12 w-full rounded-xl border border-white/10 bg-white/5 px-4 py-1 text-sm shadow-sm transition-all focus-visible:outline-none focus-visible:ring-1 focus-visible:ring-primary/50 [&>option]:bg-card text-foreground appearance-none cursor-pointer disabled:opacity-50 disabled:cursor-not-allowed"
value={selectedAssistant.voice}
onChange={(e) => updateAssistant('voice', e.target.value)}
disabled={selectedAssistant.voiceOutputEnabled === false}
>
<option value="" disabled>请选择声音库中的声音...</option>
{voices.map(voice => (
@@ -645,7 +664,9 @@ export const AssistantsPage: React.FC = () => {
</div>
<p className="text-xs text-muted-foreground flex items-center mt-1">
<Sparkles className="w-3 h-3 mr-1 text-primary opacity-70" />
音色配置同步自声音库。如需添加更多音色，请前往“声音库”模块。
{selectedAssistant.voiceOutputEnabled === false
? '启用语音输出后才可选择音色。'
: '音色配置同步自声音库。如需添加更多音色，请前往“声音库”模块。'}
</p>
</div>
@@ -916,6 +937,7 @@ export const AssistantsPage: React.FC = () => {
llmModels={llmModels}
asrModels={asrModels}
tools={tools}
textTtsEnabled={selectedAssistant.voiceOutputEnabled !== false}
/>
)}
@@ -1030,6 +1052,7 @@ export const DebugDrawer: React.FC<{
llmModels?: LLMModel[];
asrModels?: ASRModel[];
tools?: Tool[];
textTtsEnabled: boolean;
sessionMetadataExtras?: Record<string, any>;
onProtocolEvent?: (event: Record<string, any>) => void;
}> = ({
@@ -1040,6 +1063,7 @@ export const DebugDrawer: React.FC<{
llmModels = [],
asrModels = [],
tools = [],
textTtsEnabled,
sessionMetadataExtras,
onProtocolEvent,
}) => {
@@ -1117,7 +1141,6 @@ export const DebugDrawer: React.FC<{
const [selectedCamera, setSelectedCamera] = useState<string>('');
const [selectedMic, setSelectedMic] = useState<string>('');
const [isSwapped, setIsSwapped] = useState(false);
const [textTtsEnabled, setTextTtsEnabled] = useState(true);
const [aecEnabled, setAecEnabled] = useState<boolean>(() => localStorage.getItem('debug_audio_aec') !== '0');
const [nsEnabled, setNsEnabled] = useState<boolean>(() => localStorage.getItem('debug_audio_ns') !== '0');
const [agcEnabled, setAgcEnabled] = useState<boolean>(() => localStorage.getItem('debug_audio_agc') !== '0');
@@ -1546,6 +1569,7 @@ export const DebugDrawer: React.FC<{
const buildLocalResolvedRuntime = () => {
const warnings: string[] = [];
const services: Record<string, any> = {};
const ttsEnabled = Boolean(textTtsEnabled);
const isExternalLlm = assistant.configMode === 'dify' || assistant.configMode === 'fastgpt';
const knowledgeBaseId = String(assistant.knowledgeBaseId || '').trim();
const knowledge = knowledgeBaseId
@@ -1597,6 +1621,7 @@ export const DebugDrawer: React.FC<{
if (voice) {
const ttsProvider = isSiliconflowVendor(voice.vendor) ? 'siliconflow' : 'edge';
services.tts = {
enabled: ttsEnabled,
provider: ttsProvider,
model: voice.model,
apiKey: ttsProvider === 'siliconflow' ? voice.apiKey : null,
@@ -1605,17 +1630,25 @@ export const DebugDrawer: React.FC<{
};
} else {
services.tts = {
enabled: ttsEnabled,
voice: assistant.voice,
speed: assistant.speed || 1.0,
};
warnings.push(`Voice resource not found in loaded list: ${assistant.voice}`);
}
} else if (!ttsEnabled) {
services.tts = {
enabled: false,
};
}
const localResolved = {
assistantId: assistant.id,
warnings,
sessionStartMetadata: {
output: {
mode: ttsEnabled ? 'audio' : 'text',
},
systemPrompt: assistant.prompt || '',
greeting: assistant.opener || '',
knowledgeBaseId,
@@ -2007,15 +2040,9 @@ export const DebugDrawer: React.FC<{
</div>
<div className="flex items-center justify-between gap-2">
<Badge variant="outline" className="text-xs">WS: {wsStatus}</Badge>
<label className="inline-flex items-center gap-1 text-xs text-muted-foreground px-2 py-1 rounded border border-white/10">
<input
type="checkbox"
checked={textTtsEnabled}
onChange={(e) => setTextTtsEnabled(e.target.checked)}
className="accent-primary"
/>
TTS
</label>
<Badge variant={textTtsEnabled ? 'outline' : 'secondary'} className="text-xs">
TTS: {textTtsEnabled ? 'ON' : 'OFF'}
</Badge>
</div>
<div className="rounded-md border border-white/10 bg-black/20 p-2 space-y-2">
<p className="text-[10px] uppercase tracking-widest text-muted-foreground">Audio 3A</p>

3
web/services/backendApi.ts
View File

@@ -33,6 +33,7 @@ const mapAssistant = (raw: AnyRecord): Assistant => ({
prompt: readField(raw, ['prompt'], ''),
knowledgeBaseId: readField(raw, ['knowledgeBaseId', 'knowledge_base_id'], ''),
language: readField(raw, ['language'], 'zh') as 'zh' | 'en',
voiceOutputEnabled: Boolean(readField(raw, ['voiceOutputEnabled', 'voice_output_enabled'], true)),
voice: readField(raw, ['voice'], ''),
speed: Number(readField(raw, ['speed'], 1)),
hotwords: readField(raw, ['hotwords'], []),
@@ -210,6 +211,7 @@ export const createAssistant = async (data: Partial<Assistant>): Promise<Assista
prompt: data.prompt || '',
knowledgeBaseId: data.knowledgeBaseId || '',
language: data.language || 'zh',
voiceOutputEnabled: data.voiceOutputEnabled ?? true,
voice: data.voice || '',
speed: data.speed ?? 1,
hotwords: data.hotwords || [],
@@ -234,6 +236,7 @@ export const updateAssistant = async (id: string, data: Partial<Assistant>): Pro
prompt: data.prompt,
knowledgeBaseId: data.knowledgeBaseId,
language: data.language,
voiceOutputEnabled: data.voiceOutputEnabled,
voice: data.voice,
speed: data.speed,
hotwords: data.hotwords,

1
web/types.ts
View File

@@ -7,6 +7,7 @@ export interface Assistant {
prompt: string;
knowledgeBaseId: string;
language: 'zh' | 'en';
voiceOutputEnabled?: boolean;
voice: string; // This will now store the ID of the voice from Voice Library
speed: number;
hotwords: string[];