From e5941926be79f16a515af2a4d356cefb56e8b34d Mon Sep 17 00:00:00 2001 From: Garegin Harutyunyan <14837795+realgarik@users.noreply.github.com> Date: Mon, 27 Apr 2026 16:14:00 +0400 Subject: [PATCH] Krisp tt demo tool (#4335) * VIVA SDK TT v3 support * Format fix. * Renamed the API naming, removed '3' from the name. * Implementation of User turn start strategy using Krisp VIVA Interruption Prediction in scope of TT v3 support. * TT demo tool * Some improvements for demo scripts, audio recordin, etc. * Enhance demo scripts with VAD selection and audio embedding features. Updated HTML report to include annotated audio players and improved response time metrics in summary formatting. Added README for setup and usage instructions. * Refactor interrupt prediction demo to compare multiple interruption strategies (Krisp IP vs VAD). Updated README with usage instructions and output details. Enhanced audio processing with new helper functions for generating beeps and mixing audio. * Refactor demo scripts to improve latency metrics by introducing total_delay property in TurnEvent. Update formatting in reports and visualizations to reflect accurate speech end times, including VAD wait times. Enhance HTML report with detailed latency information and adjust audio processing to account for VAD stop seconds. * Add audio resampling functionality and update demo scripts for improved audio processing - Introduced `resample_audio` function to handle audio resampling with linear interpolation. - Updated `demo_audio_recorder.py` to utilize the new resampling feature, ensuring audio is saved at the requested sample rate. - Modified `demo_interrupt_prediction.py` and `demo_turn_taking.py` to resample audio to 16 kHz for compatibility with Silero VAD. - Adjusted imports in demo scripts to include the new resampling function. - Enhanced error handling for sample rate discrepancies in audio recording. * Enhance demo_interrupt_prediction.py with VAD type selection and improved processing logic - Added support for selecting between "silero" and "krisp" VAD engines in the demo script. - Introduced a new create_vad function to configure VAD analyzers based on the selected type. - Updated audio processing logic to handle VAD type-specific resampling and state management. - Modified the KrispVivaIPUserTurnStartStrategy to utilize a separate vad_flag for per-frame VAD input, improving interruption detection accuracy. * Refactor audio processing scripts for improved readability and consistency - Updated type hinting in `resample_audio` function to use `tuple` instead of `Tuple`. - Simplified print statements in `demo_audio_recorder.py`, `demo_formatting.py`, and `demo_interrupt_prediction.py` for better readability. - Adjusted argument formatting in `demo_audio_recorder.py` and `demo_formatting.py` for consistency. - Cleaned up list comprehensions in `demo_formatting.py`, `demo_html_report.py`, and `demo_interrupt_prediction.py` for clarity. - Enhanced error handling in `__init__.py` for the KrispVivaIPUserTurnStartStrategy import. * Refactor VAD handling in KrispVivaIPUserTurnStartStrategy and update tests for clarity - Simplified the argument formatting in the _handle_vad_started method for improved readability. - Updated test assertions to reflect changes in VAD processing logic, ensuring that the vad_flag is correctly set to False during continuous state processing. - Enhanced test cases to verify that the process method is called appropriately under different conditions. * more format fixes. * removed demo scripts. * reverted wrongly removed file. * Corrected the IP integration logic. * style fix. * Refactor audio processing and state management in KrispVivaIPUserTurnStartStrategy - Removed the unused _vad_flag attribute to streamline state tracking. - Updated the reset method to clear the audio buffer instead of resetting the vad_flag. - Adjusted the process_frame method to use _speech_active for VAD input, enhancing clarity in the logic. - Modified tests to reflect changes in state management and ensure proper functionality of the reset method and audio buffer handling. * FIxed formatting --------- Co-authored-by: Aram Poghosyan --- src/pipecat/turns/user_start/__init__.py | 2 +- .../krisp_viva_ip_user_turn_start_strategy.py | 14 +- .../test_krisp_ip_user_turn_start_strategy.py | 120 +++++++++++++++++- 3 files changed, 124 insertions(+), 12 deletions(-) diff --git a/src/pipecat/turns/user_start/__init__.py b/src/pipecat/turns/user_start/__init__.py index 14de5d28b..c2b1efa73 100644 --- a/src/pipecat/turns/user_start/__init__.py +++ b/src/pipecat/turns/user_start/__init__.py @@ -14,7 +14,7 @@ from .wake_phrase_user_turn_start_strategy import WakePhraseUserTurnStartStrateg try: from .krisp_viva_ip_user_turn_start_strategy import KrispVivaIPUserTurnStartStrategy except ImportError: - KrispVivaIPUserTurnStartStrategy = None + KrispVivaIPUserTurnStartStrategy = None # krisp_audio not installed __all__ = [ "BaseUserTurnStartStrategy", diff --git a/src/pipecat/turns/user_start/krisp_viva_ip_user_turn_start_strategy.py b/src/pipecat/turns/user_start/krisp_viva_ip_user_turn_start_strategy.py index 807bc8154..a366b8847 100644 --- a/src/pipecat/turns/user_start/krisp_viva_ip_user_turn_start_strategy.py +++ b/src/pipecat/turns/user_start/krisp_viva_ip_user_turn_start_strategy.py @@ -43,7 +43,7 @@ except ModuleNotFoundError as e: logger.error( "In order to use KrispVivaIPUserTurnStartStrategy, you need to install krisp_audio." ) - raise Exception(f"Missing module: {e}") + raise ImportError(f"Missing module: {e}") from e class KrispVivaIPUserTurnStartStrategy(BaseUserTurnStartStrategy): @@ -211,22 +211,24 @@ class KrispVivaIPUserTurnStartStrategy(BaseUserTurnStartStrategy): """ logger.trace("Krisp VIVA IP: VAD speech started, collecting audio for classification") self._speech_active = True - self._audio_buffer.clear() self._decision_made = False return ProcessFrameResult.CONTINUE async def _handle_audio(self, frame: InputAudioRawFrame) -> ProcessFrameResult: """Feed audio to the IP model and check for genuine interruption. + Every frame is passed to the IP model regardless of speech state so + that the model maintains continuous internal state (matching the + standalone Krisp SDK behaviour). ``_speech_active`` is forwarded as + the per-frame VAD input to the model and also gates the threshold + evaluation. + Args: frame: Raw audio input frame. Returns: STOP if the model detects a genuine interruption, CONTINUE otherwise. """ - if not self._speech_active or self._decision_made: - return ProcessFrameResult.CONTINUE - self._ensure_session(frame.sample_rate) if self._ip_session is None or self._samples_per_frame is None: @@ -254,7 +256,7 @@ class KrispVivaIPUserTurnStartStrategy(BaseUserTurnStartStrategy): for ip_frame in frames: ip_prob = self._ip_session.process(ip_frame.tolist(), self._speech_active) - if ip_prob >= self._threshold: + if self._speech_active and not self._decision_made and ip_prob >= self._threshold: logger.debug( f"Krisp VIVA IP: genuine interruption detected (prob={ip_prob:.3f}, " f"threshold={self._threshold})" diff --git a/tests/test_krisp_ip_user_turn_start_strategy.py b/tests/test_krisp_ip_user_turn_start_strategy.py index bb34d879a..5e5dd00cc 100644 --- a/tests/test_krisp_ip_user_turn_start_strategy.py +++ b/tests/test_krisp_ip_user_turn_start_strategy.py @@ -103,8 +103,12 @@ class TestKrispVivaIPUserTurnStartStrategy(unittest.IsolatedAsyncioTestCase): api_key="test-key", ) - def _audio_frame(self, sample_rate: int = 16000, frame_duration_ms: int = 20): - samples = int(sample_rate * frame_duration_ms / 1000) + def _audio_frame( + self, sample_rate: int = 16000, frame_duration_ms: int = 20, num_samples: int | None = None + ): + samples = ( + num_samples if num_samples is not None else int(sample_rate * frame_duration_ms / 1000) + ) return InputAudioRawFrame( audio=_int16_silence(samples), sample_rate=sample_rate, @@ -165,7 +169,10 @@ class TestKrispVivaIPUserTurnStartStrategy(unittest.IsolatedAsyncioTestCase): result = await strategy.process_frame(self._audio_frame()) self.assertEqual(result, ProcessFrameResult.CONTINUE) - self.mock_ip_session.process.assert_not_called() + # process() is still called (continuous state), but with speech_active=False + self.mock_ip_session.process.assert_called() + args = self.mock_ip_session.process.call_args[0] + self.assertFalse(args[1]) # speech_active should be False finally: await strategy.cleanup() @@ -182,7 +189,10 @@ class TestKrispVivaIPUserTurnStartStrategy(unittest.IsolatedAsyncioTestCase): result = await strategy.process_frame(self._audio_frame()) self.assertEqual(result, ProcessFrameResult.CONTINUE) - self.mock_ip_session.process.assert_not_called() + # process() is still called (continuous state), but with speech_active=False + self.mock_ip_session.process.assert_called() + args = self.mock_ip_session.process.call_args[0] + self.assertFalse(args[1]) # speech_active should be False finally: await strategy.cleanup() @@ -193,7 +203,11 @@ class TestKrispVivaIPUserTurnStartStrategy(unittest.IsolatedAsyncioTestCase): try: result = await strategy.process_frame(self._audio_frame()) self.assertEqual(result, ProcessFrameResult.CONTINUE) - self.mock_ip_session.process.assert_not_called() + # process() is called (continuous state) even before VAD start, + # but _speech_active=False prevents triggering despite high prob + self.mock_ip_session.process.assert_called() + args = self.mock_ip_session.process.call_args[0] + self.assertFalse(args[1]) # speech_active should be False finally: await strategy.cleanup() @@ -219,6 +233,21 @@ class TestKrispVivaIPUserTurnStartStrategy(unittest.IsolatedAsyncioTestCase): finally: await strategy.cleanup() + async def test_reset_clears_audio_buffer(self): + self.mock_ip_session.process = MagicMock(return_value=0.1) + + strategy = self._make_strategy(threshold=0.5) + try: + await strategy.process_frame(VADUserStartedSpeakingFrame()) + # Feed a partial frame (smaller than samples_per_frame) so it stays in buffer + await strategy.process_frame(self._audio_frame(num_samples=10)) + self.assertGreater(len(strategy._audio_buffer), 0) + + await strategy.process_frame(VADUserStoppedSpeakingFrame()) + self.assertEqual(len(strategy._audio_buffer), 0) + finally: + await strategy.cleanup() + async def test_unrelated_frames_continue(self): strategy = self._make_strategy() try: @@ -231,6 +260,87 @@ class TestKrispVivaIPUserTurnStartStrategy(unittest.IsolatedAsyncioTestCase): finally: await strategy.cleanup() + async def test_reset_method_clears_state(self): + self.mock_ip_session.process = MagicMock(return_value=0.1) + + strategy = self._make_strategy(threshold=0.5) + try: + await strategy.process_frame(VADUserStartedSpeakingFrame()) + await strategy.process_frame(self._audio_frame(num_samples=10)) + self.assertTrue(strategy._speech_active) + self.assertGreater(len(strategy._audio_buffer), 0) + + await strategy.reset() + + self.assertFalse(strategy._speech_active) + self.assertEqual(len(strategy._audio_buffer), 0) + self.assertFalse(strategy._decision_made) + finally: + await strategy.cleanup() + + async def test_cleanup_releases_sdk(self): + strategy = self._make_strategy() + + await strategy.cleanup() + + self.mock_sdk_manager.release.assert_called_once() + self.assertIsNone(strategy._ip_session) + self.assertFalse(strategy._sdk_acquired) + + def test_init_raises_if_no_model_path(self): + with self.assertRaises(ValueError): + KrispVivaIPUserTurnStartStrategy(api_key="test-key") + + def test_init_raises_if_wrong_extension(self): + with self.assertRaises(ValueError): + KrispVivaIPUserTurnStartStrategy(model_path="/some/model.bin", api_key="test-key") + + def test_init_raises_if_file_not_found(self): + with self.assertRaises(FileNotFoundError): + KrispVivaIPUserTurnStartStrategy( + model_path="/nonexistent/model.kef", api_key="test-key" + ) + + def test_init_raises_if_sdk_fails(self): + self.mock_sdk_manager.acquire.side_effect = RuntimeError("SDK error") + with self.assertRaises(RuntimeError): + KrispVivaIPUserTurnStartStrategy(model_path=self.model_path, api_key="test-key") + + def test_init_uses_env_var_for_model_path(self): + with patch.dict(os.environ, {"KRISP_VIVA_IP_MODEL_PATH": self.model_path}): + strategy = KrispVivaIPUserTurnStartStrategy(api_key="test-key") + self.assertEqual(strategy._model_path, self.model_path) + + async def test_vad_stopped_when_speech_inactive_is_no_op(self): + strategy = self._make_strategy() + try: + result = await strategy.process_frame(VADUserStoppedSpeakingFrame()) + self.assertEqual(result, ProcessFrameResult.CONTINUE) + self.assertFalse(strategy._speech_active) + finally: + await strategy.cleanup() + + async def test_interruption_at_exact_threshold_triggers(self): + threshold = 0.5 + self.mock_ip_session.process = MagicMock(return_value=threshold) + + strategy = self._make_strategy(threshold=threshold) + try: + fired = False + + @strategy.event_handler("on_user_turn_started") + async def on_user_turn_started(strategy, params): + nonlocal fired + fired = True + + await strategy.process_frame(VADUserStartedSpeakingFrame()) + result = await strategy.process_frame(self._audio_frame()) + + self.assertTrue(fired) + self.assertEqual(result, ProcessFrameResult.STOP) + finally: + await strategy.cleanup() + if __name__ == "__main__": unittest.main()