wx44wx/pipecat
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

MinWordsUserTurnStartStrategy: avoid using interim transcriptions

Browse Source 
Using a mix of interim and final transcriptions caused incorrect behavior; using
only interim transcriptions alone also produced undesired results.
This commit is contained in:
Aleix Conchillo Flaqué
2026-01-14 18:21:27 -08:00
parent 965466cc09
commit 61a7922bea
2 changed files with 7 additions and 88 deletions
Download Patch File Download Diff File Expand all files Collapse all files

65
tests/test_user_turn_start_strategy.py
View File

@@ -24,7 +24,7 @@ from pipecat.turns.user_start import (
class TestMinWordsInterruptionStrategy(unittest.IsolatedAsyncioTestCase):
async def test_bot_speaking_transcriptions(self):
strategy = MinWordsUserTurnStartStrategy(min_words=2)
strategy = MinWordsUserTurnStartStrategy(min_words=3)
should_start = None
@@ -38,7 +38,7 @@ class TestMinWordsInterruptionStrategy(unittest.IsolatedAsyncioTestCase):
self.assertFalse(should_start)
await strategy.process_frame(
TranscriptionFrame(text=" there!", user_id="cat", timestamp="")
TranscriptionFrame(text=" there friend!", user_id="cat", timestamp="")
)
self.assertTrue(should_start)
@@ -50,54 +50,14 @@ class TestMinWordsInterruptionStrategy(unittest.IsolatedAsyncioTestCase):
await strategy.process_frame(TranscriptionFrame(text="Hello!", user_id="cat", timestamp=""))
self.assertFalse(should_start)
await strategy.process_frame(TranscriptionFrame(text="How", user_id="cat", timestamp=""))
self.assertFalse(should_start)
await strategy.process_frame(
TranscriptionFrame(text="How are you?", user_id="cat", timestamp="")
)
self.assertTrue(should_start)
async def test_bot_speaking_interim_transcriptions(self):
strategy = MinWordsUserTurnStartStrategy(min_words=2)
should_start = None
@strategy.event_handler("on_user_turn_started")
async def on_user_turn_started(strategy, params):
nonlocal should_start
should_start = True
await strategy.process_frame(BotStartedSpeakingFrame())
await strategy.process_frame(
InterimTranscriptionFrame(text="Hello", user_id="cat", timestamp="")
)
self.assertFalse(should_start)
await strategy.process_frame(BotStartedSpeakingFrame())
await strategy.process_frame(
InterimTranscriptionFrame(text="Hello there!", user_id="cat", timestamp="")
)
self.assertTrue(should_start)
async def test_bot_speaking_all_transcriptions(self):
strategy = MinWordsUserTurnStartStrategy(min_words=2)
should_start = None
@strategy.event_handler("on_user_turn_started")
async def on_user_turn_started(strategy, params):
nonlocal should_start
should_start = True
await strategy.process_frame(BotStartedSpeakingFrame())
await strategy.process_frame(
InterimTranscriptionFrame(text="Hello", user_id="cat", timestamp="")
)
self.assertFalse(should_start)
await strategy.process_frame(
TranscriptionFrame(text="Hello there!", user_id="cat", timestamp="")
)
self.assertTrue(should_start)
async def test_bot_not_speaking_transcriptions(self):
strategy = MinWordsUserTurnStartStrategy(min_words=2)
@@ -111,21 +71,6 @@ class TestMinWordsInterruptionStrategy(unittest.IsolatedAsyncioTestCase):
await strategy.process_frame(TranscriptionFrame(text="Hello", user_id="cat", timestamp=""))
self.assertTrue(should_start)
async def test_bot_not_speaking_interim_transcriptions(self):
strategy = MinWordsUserTurnStartStrategy(min_words=2)
should_start = None
@strategy.event_handler("on_user_turn_started")
async def on_user_turn_started(strategy, params):
nonlocal should_start
should_start = True
await strategy.process_frame(
InterimTranscriptionFrame(text="Hello", user_id="cat", timestamp="")
)
self.assertTrue(should_start)
class TestVADUserTurnStartStrategy(unittest.IsolatedAsyncioTestCase):
async def test_vad_strategy(self):