diff --git a/.env.example b/.env.example index d55e123..2ec029e 100644 --- a/.env.example +++ b/.env.example @@ -45,6 +45,9 @@ OWNER_FOLLOWUP_LISTEN_TIMEOUT_MS=3000 OWNER_BARGE_IN_ENABLED=1 OWNER_BARGE_IN_MIN_SPEECH_MS=250 OWNER_BARGE_IN_ECHO_GUARD_MS=500 +OWNER_END_CHIME_ENABLED=1 +OWNER_END_CHIME_FREQUENCY_HZ=880 +OWNER_END_CHIME_DURATION_MS=140 OWNER_WAKE_WORD=小杰小杰 OWNER_SAMPLE_RATE=16000 OWNER_CHANNELS=1 diff --git a/README.md b/README.md index 34c00c5..f477c3a 100644 --- a/README.md +++ b/README.md @@ -72,6 +72,9 @@ OWNER_FOLLOWUP_LISTEN_TIMEOUT_MS=3000 OWNER_BARGE_IN_ENABLED=1 OWNER_BARGE_IN_MIN_SPEECH_MS=250 OWNER_BARGE_IN_ECHO_GUARD_MS=500 +OWNER_END_CHIME_ENABLED=1 +OWNER_END_CHIME_FREQUENCY_HZ=880 +OWNER_END_CHIME_DURATION_MS=140 ``` `OWNER_WAKE_PROVIDER=local_kws` 表示唤醒词“小杰小杰”由本地模型检测。唤醒命中后会先本地播报 `OWNER_WAKE_ACK_TEXT=我在`，再开始听取问题。`OWNER_SPEECH_PROVIDER=local` 表示正式问题 STT、实时字幕和 TTS 都走本地模型或 macOS 本地能力；云端只接收 final 文本和本次会话历史用于 LLM 回复。 @@ -84,6 +87,8 @@ OWNER_BARGE_IN_ECHO_GUARD_MS=500 `OWNER_BARGE_IN_ENABLED=1` 表示播报中允许打断。播放开始后的 `OWNER_BARGE_IN_ECHO_GUARD_MS=500` 毫秒内忽略麦克风输入，之后如果检测到至少 `OWNER_BARGE_IN_MIN_SPEECH_MS=250` 毫秒有效用户语音，并且 realtime STT 给出有效 partial，就停止剩余播报。上下文只记录已经完整播出的 assistant 句子，未播出的内容不会写入临时历史。 +`OWNER_END_CHIME_ENABLED=1` 表示对话自然结束或追问超时恢复待机前会播放一声本地短提示音。提示音不走 TTS，也不会写入上下文；`OWNER_END_CHIME_FREQUENCY_HZ` 和 `OWNER_END_CHIME_DURATION_MS` 可以调整音高和时长，设置 `OWNER_END_CHIME_ENABLED=0` 可以关闭。 + ## 本地模型 首次运行前必须准备本地语音模型；同一脚本会下载 wake、VAD、2025 中文 CTC STT 和 GTCRN denoiser 模型： diff --git a/src/owner_voice_pet/__init__.py b/src/owner_voice_pet/__init__.py index 4c9cf96..d98bbec 100644 --- a/src/owner_voice_pet/__init__.py +++ b/src/owner_voice_pet/__init__.py @@ -27,7 +27,7 @@ from .llm import MockLlmProvider, OpenAICompatibleLlmProvider from .pipeline import PipelineResult, VoicePipeline from .runtime import LiveVoiceRuntime, RuntimeSummary, TerminalRuntimeReporter, TurnResult, build_live_runtime from .simulation import run_simulated_live -from .tts import CloudTtsProvider, MacSayTtsProvider, SentenceBuffer, SineTtsProvider, sanitize_tts_text +from .tts import CloudTtsProvider, MacSayTtsProvider, SentenceBuffer, SineTtsProvider, make_prompt_chime, sanitize_tts_text from .assets import validate_pet_assets from .ui import ConsolePetWindow, PetStateController, PetVisualState @@ -70,6 +70,7 @@ __all__ = [ "MacSayTtsProvider", "SentenceBuffer", "SineTtsProvider", + "make_prompt_chime", "sanitize_tts_text", "validate_pet_assets", "ConsolePetWindow", diff --git a/src/owner_voice_pet/assistant_pipeline.py b/src/owner_voice_pet/assistant_pipeline.py index eaf6851..bb33741 100644 --- a/src/owner_voice_pet/assistant_pipeline.py +++ b/src/owner_voice_pet/assistant_pipeline.py @@ -47,7 +47,7 @@ from .protocols import ( WakeWordProvider, ) from .stt import is_valid_transcript_text -from .tts import SentenceBuffer, sanitize_tts_text +from .tts import SentenceBuffer, make_prompt_chime, sanitize_tts_text from .vad import VadRecorder @@ -344,19 +344,10 @@ class TurnController: else: decision = self._decide_continuation(current_user_text, last_assistant_text, current_turn_id) if not decision.should_continue: - self._event( - CONTINUOUS_SESSION_ENDED, - PipelineState.WAKE_LISTENING, - "", - turn_id=current_turn_id, + self._end_conversation( + current_turn_id, payload={"decision": decision.action, "reason": decision.reason}, ) - self._event( - STANDBY_RESUMED, - PipelineState.WAKE_LISTENING, - "恢复待机：可继续唤醒", - turn_id=current_turn_id, - ) return TurnResult( True, current_user_text, @@ -488,11 +479,29 @@ class TurnController: "追问超时：未检测到用户回答", turn_id=turn_id, ) - self._event(CONTINUOUS_SESSION_ENDED, PipelineState.WAKE_LISTENING, "", turn_id=turn_id) - self._event(STANDBY_RESUMED, PipelineState.WAKE_LISTENING, "恢复待机：可继续唤醒", turn_id=turn_id) + self._end_conversation(turn_id) return None return user_text + def _end_conversation(self, turn_id: int, *, payload: dict[str, object] | None = None) -> None: + self._event(CONTINUOUS_SESSION_ENDED, PipelineState.WAKE_LISTENING, "", turn_id=turn_id, payload=payload) + self._play_end_chime() + self._event(STANDBY_RESUMED, PipelineState.WAKE_LISTENING, "恢复待机：可继续唤醒", turn_id=turn_id) + + def _play_end_chime(self) -> None: + if not self.config.end_chime_enabled: + return + segment = make_prompt_chime( + frequency_hz=self.config.end_chime_frequency_hz, + duration_ms=self.config.end_chime_duration_ms, + sample_rate=self.config.sample_rate, + channels=self.config.channels, + ) + playback = self.transport.play_pcm(segment) + if playback.error: + return + self._drain_input_after_playback() + def _speak(self, sentence: str, turn_id: int) -> SpeakResult: spoken_sentence = sanitize_tts_text(sentence) if not spoken_sentence: diff --git a/src/owner_voice_pet/cli.py b/src/owner_voice_pet/cli.py index 8fddf7f..b3f0cfc 100644 --- a/src/owner_voice_pet/cli.py +++ b/src/owner_voice_pet/cli.py @@ -104,6 +104,9 @@ def main(argv: list[str] | None = None) -> int: "barge_in_enabled": config.barge_in_enabled, "barge_in_min_speech_ms": config.barge_in_min_speech_ms, "barge_in_echo_guard_ms": config.barge_in_echo_guard_ms, + "end_chime_enabled": config.end_chime_enabled, + "end_chime_frequency_hz": config.end_chime_frequency_hz, + "end_chime_duration_ms": config.end_chime_duration_ms, }, ensure_ascii=False, sort_keys=True, diff --git a/src/owner_voice_pet/config.py b/src/owner_voice_pet/config.py index 7e02a6d..ce65442 100644 --- a/src/owner_voice_pet/config.py +++ b/src/owner_voice_pet/config.py @@ -59,6 +59,9 @@ class AppConfig: barge_in_enabled: bool = True barge_in_min_speech_ms: int = 250 barge_in_echo_guard_ms: int = 500 + end_chime_enabled: bool = True + end_chime_frequency_hz: int = 880 + end_chime_duration_ms: int = 140 @classmethod def from_dotenv(cls, path: str | Path = ".env", prefix: str = "OWNER_") -> "AppConfig": @@ -130,6 +133,9 @@ class AppConfig: barge_in_enabled=(get("BARGE_IN_ENABLED", "1") or "1").lower() not in {"0", "false", "no"}, barge_in_min_speech_ms=int(get("BARGE_IN_MIN_SPEECH_MS", "250") or "250"), barge_in_echo_guard_ms=int(get("BARGE_IN_ECHO_GUARD_MS", "500") or "500"), + end_chime_enabled=(get("END_CHIME_ENABLED", "1") or "1").lower() not in {"0", "false", "no"}, + end_chime_frequency_hz=int(get("END_CHIME_FREQUENCY_HZ", "880") or "880"), + end_chime_duration_ms=int(get("END_CHIME_DURATION_MS", "140") or "140"), ) @classmethod @@ -394,6 +400,20 @@ class AppConfig: "startup", ) ) + for name, value in { + "OWNER_END_CHIME_FREQUENCY_HZ": self.end_chime_frequency_hz, + "OWNER_END_CHIME_DURATION_MS": self.end_chime_duration_ms, + }.items(): + if value <= 0: + errors.append( + ProviderError( + ErrorCode.CONFIG_MISSING_VALUE, + f"{name} must be positive", + False, + "config", + "startup", + ) + ) return errors def api_url(self, path: str) -> str: diff --git a/src/owner_voice_pet/runtime.py b/src/owner_voice_pet/runtime.py index 0747341..8be13c0 100644 --- a/src/owner_voice_pet/runtime.py +++ b/src/owner_voice_pet/runtime.py @@ -33,7 +33,7 @@ from .models import AudioSegment, ErrorCode, PipelineState, ProviderError from .protocols import AudioTransport, LlmProvider, RealtimeSttProvider, SttProvider, TtsProvider, WakeWordProvider from .stt import CloudAsrSttProvider, SherpaOnnxSttProvider, is_valid_transcript_text from .transport import SoundDeviceAudioTransport -from .tts import CloudTtsProvider, MacSayTtsProvider, SentenceBuffer, sanitize_tts_text +from .tts import CloudTtsProvider, MacSayTtsProvider, SentenceBuffer, make_prompt_chime, sanitize_tts_text from .vad import EnergyVadProvider, HybridVadProvider, PrimarySpeakerVadRecorder, SherpaOnnxVadProvider, VadRecorder from .wakeword import SherpaOnnxKeywordWakeWordProvider @@ -323,6 +323,7 @@ class LiveVoiceRuntime: turn_id, ) self.context.append_assistant(spoken_text) + self._play_end_chime() self._event(STANDBY_RESUMED, PipelineState.WAKE_LISTENING, "恢复待机：可继续唤醒", turn_id=turn_id) return TurnResult(True, user_text, spoken_text, states=list(self._states)) @@ -339,6 +340,20 @@ class LiveVoiceRuntime: self._drain_input_after_playback() return spoken_sentence + def _play_end_chime(self) -> None: + if not self.config.end_chime_enabled: + return + segment = make_prompt_chime( + frequency_hz=self.config.end_chime_frequency_hz, + duration_ms=self.config.end_chime_duration_ms, + sample_rate=self.config.sample_rate, + channels=self.config.channels, + ) + playback = self.transport.play_pcm(segment) + if playback.error: + return + self._drain_input_after_playback() + def _recover(self, error: ProviderError, turn_id: int) -> TurnResult: self._event(STAGE_ERROR, PipelineState.ERROR_RECOVERING, error.message, turn_id=turn_id, payload={"error": error}) self._event(RECOVERING, PipelineState.ERROR_RECOVERING, "恢复待机：本轮已结束", turn_id=turn_id) diff --git a/src/owner_voice_pet/tts.py b/src/owner_voice_pet/tts.py index a106c32..de2dfd5 100644 --- a/src/owner_voice_pet/tts.py +++ b/src/owner_voice_pet/tts.py @@ -144,6 +144,36 @@ def _normalize_spoken_text(text: str) -> str: return clean.strip(" \t\r\n，,；;：:") +def make_prompt_chime( + *, + frequency_hz: int = 880, + duration_ms: int = 140, + sample_rate: int = 16000, + channels: int = 1, +) -> AudioSegment: + duration_ms = max(20, duration_ms) + sample_rate = max(8000, sample_rate) + channels = max(1, channels) + samples = int(sample_rate * duration_ms / 1000) + fade_samples = max(1, min(samples // 2, int(sample_rate * 0.015))) + pcm = bytearray() + for idx in range(samples): + fade_in = idx / fade_samples if idx < fade_samples else 1.0 + fade_out = (samples - idx - 1) / fade_samples if idx >= samples - fade_samples else 1.0 + envelope = max(0.0, min(1.0, fade_in, fade_out)) + value = int(math.sin(2 * math.pi * frequency_hz * idx / sample_rate) * 9000 * envelope) + packed = struct.pack(" None: self.max_chars = max_chars diff --git a/tests/test_cli_acceptance.py b/tests/test_cli_acceptance.py index 68e6b30..4f6eca0 100644 --- a/tests/test_cli_acceptance.py +++ b/tests/test_cli_acceptance.py @@ -46,6 +46,8 @@ class CliAcceptanceTests(unittest.TestCase): self.assertEqual(data["endpoint_mode"], "primary_speaker") self.assertEqual(data["realtime_transcript_idle_timeout_ms"], 1500) self.assertEqual(data["context_mode"], "session_memory") + self.assertTrue(data["end_chime_enabled"]) + self.assertEqual(data["end_chime_duration_ms"], 140) self.assertNotIn("secret-value", str(data)) def test_security_check_command_has_no_leaks(self) -> None: diff --git a/tests/test_live_runtime.py b/tests/test_live_runtime.py index 5ead38f..0ade4fb 100644 --- a/tests/test_live_runtime.py +++ b/tests/test_live_runtime.py @@ -269,8 +269,10 @@ class LiveRuntimeTests(unittest.TestCase): self.assertEqual(runtime.config.post_playback_drain_ms, 0) self.assertEqual(len(stt.calls), 2) self.assertEqual(len(llm.calls), 2) - self.assertEqual(len(transport.played_segments), 4) - self.assertEqual(transport.flush_count, 4) + self.assertEqual(len(transport.played_segments), 6) + self.assertEqual(transport.flush_count, 6) + self.assertEqual(transport.played_segments[2].metadata["chime"], "end") + self.assertEqual(transport.played_segments[5].metadata["chime"], "end") self.assertEqual(reporter.transcripts, ["第一问", "第二问"]) self.assertIn("应答中：我在", reporter.statuses) self.assertLess(reporter.statuses.index("唤醒命中"), reporter.statuses.index("应答中：我在")) @@ -324,7 +326,7 @@ class LiveRuntimeTests(unittest.TestCase): self.assertEqual(summary.completed_turns, 2) self.assertEqual(ack_tts.synthesized_texts, ["我在"]) self.assertEqual(transport.played_segments[0].metadata["text"], "我在") - self.assertEqual(transport.played_segments[2].metadata["text"], "我在") + self.assertEqual(transport.played_segments[3].metadata["text"], "我在") def test_zero_post_playback_drain_flushes_without_dropping_prefilled_question(self) -> None: runtime, stt, _, transport, reporter = make_runtime(["第一问"]) @@ -335,7 +337,8 @@ class LiveRuntimeTests(unittest.TestCase): self.assertEqual(summary.completed_turns, 1) self.assertEqual(reporter.transcripts, ["第一问"]) self.assertEqual(len(stt.calls), 1) - self.assertEqual(transport.flush_count, 2) + self.assertEqual(transport.flush_count, 3) + self.assertEqual(transport.played_segments[-1].metadata["chime"], "end") def test_no_ack_text_does_not_drain_before_capture(self) -> None: runtime, stt, _, transport, reporter = make_runtime(["第一问"], wake_ack_text="") @@ -345,8 +348,9 @@ class LiveRuntimeTests(unittest.TestCase): self.assertEqual(summary.completed_turns, 1) self.assertEqual(reporter.transcripts, ["第一问"]) self.assertEqual(len(stt.calls), 1) - self.assertEqual(len(transport.played_segments), 1) - self.assertEqual(transport.flush_count, 1) + self.assertEqual(len(transport.played_segments), 2) + self.assertEqual(transport.flush_count, 2) + self.assertEqual(transport.played_segments[-1].metadata["chime"], "end") def test_temporary_context_is_sent_to_second_llm_call(self) -> None: runtime, _, llm, _, _ = make_runtime(["第一问", "第二问"]) @@ -593,6 +597,7 @@ class LiveRuntimeTests(unittest.TestCase): self.assertIn(FOLLOWUP_TIMEOUT, event_types) self.assertEqual(event_types[-1], STANDBY_RESUMED) self.assertEqual(len(llm.calls), 1) + self.assertEqual(transport.played_segments[-1].metadata["chime"], "end") def test_completed_reply_returns_to_standby_without_cloud_classifier_delay(self) -> None: frames = [wake_frame(0, 0)] diff --git a/tests/test_models_config.py b/tests/test_models_config.py index a964714..9eec226 100644 --- a/tests/test_models_config.py +++ b/tests/test_models_config.py @@ -113,6 +113,9 @@ class ModelsConfigTests(unittest.TestCase): self.assertTrue(config.barge_in_enabled) self.assertEqual(config.barge_in_min_speech_ms, 250) self.assertEqual(config.barge_in_echo_guard_ms, 500) + self.assertTrue(config.end_chime_enabled) + self.assertEqual(config.end_chime_frequency_hz, 880) + self.assertEqual(config.end_chime_duration_ms, 140) self.assertTrue(config.llm_stream) self.assertEqual(config.validate_basic(), []) @@ -153,6 +156,8 @@ class ModelsConfigTests(unittest.TestCase): followup_listen_timeout_ms=-1, barge_in_min_speech_ms=-1, barge_in_echo_guard_ms=-1, + end_chime_frequency_hz=0, + end_chime_duration_ms=0, ) errors = config.validate_basic() @@ -161,6 +166,8 @@ class ModelsConfigTests(unittest.TestCase): self.assertTrue(any("OWNER_FOLLOWUP_LISTEN_TIMEOUT_MS" in error.message for error in errors)) self.assertTrue(any("OWNER_BARGE_IN_MIN_SPEECH_MS" in error.message for error in errors)) self.assertTrue(any("OWNER_BARGE_IN_ECHO_GUARD_MS" in error.message for error in errors)) + self.assertTrue(any("OWNER_END_CHIME_FREQUENCY_HZ" in error.message for error in errors)) + self.assertTrue(any("OWNER_END_CHIME_DURATION_MS" in error.message for error in errors)) def test_speaker_similarity_threshold_range_is_validated(self) -> None: config = AppConfig(speaker_similarity_threshold=1.5) diff --git a/tests/test_pipeline_llm_tts.py b/tests/test_pipeline_llm_tts.py index f05a69e..146ece9 100644 --- a/tests/test_pipeline_llm_tts.py +++ b/tests/test_pipeline_llm_tts.py @@ -12,7 +12,7 @@ from owner_voice_pet.models import AudioFrame, ErrorCode, Message, PipelineState from owner_voice_pet.pipeline import VoicePipeline from owner_voice_pet.stt import MetadataSttProvider from owner_voice_pet.transport import MemoryAudioTransport -from owner_voice_pet.tts import CloudTtsProvider, SentenceBuffer, SineTtsProvider, sanitize_tts_text +from owner_voice_pet.tts import CloudTtsProvider, SentenceBuffer, SineTtsProvider, make_prompt_chime, sanitize_tts_text from owner_voice_pet.vad import EnergyVadProvider, VadRecorder from owner_voice_pet.wakeword import KeywordWakeWordProvider @@ -73,6 +73,13 @@ class PipelineLlmTtsTests(unittest.TestCase): self.assertGreater(len(segment.pcm), 0) self.assertGreater(segment.duration_ms, 0) + def test_prompt_chime_generates_short_pcm_audio(self) -> None: + segment = make_prompt_chime(frequency_hz=880, duration_ms=140, sample_rate=16000, channels=1) + + self.assertEqual(segment.duration_ms, 140) + self.assertEqual(segment.metadata["chime"], "end") + self.assertGreater(len(segment.pcm), 0) + def test_cloud_tts_posts_speech_request(self) -> None: class FakeResponse: def __enter__(self):