wx44wx/pipecat
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity
6,320 Commits 276 Branches 111 Tags
2f2bde9856cc13b060cecd34a60cc75d106906ae
Commit Graph

187 Commits

Author SHA1 Message Date
Paul Kompfner
913194844e Make the mechanism of adding spaces when concatenating TTS (or speech-to-speech LLM) output text explicit and deterministic, rather than heuristic-based. 
This fixes a bug where spaces were sometimes missing from assistant messages in context.
 2025-11-10 14:22:32 -05:00
Mark Backman
2300941bb8 Revert "Merge pull request #3004 from pipecat-ai/mb/improve-concat-aggregated-text" 
This reverts commit 5e7f59a0b0, reversing
changes made to 2ad4122b77.
 2025-11-10 09:58:19 -05:00
Mark Backman
03001f8047 Update TranscriptProcessor unit tests 2025-11-07 11:44:04 -05:00
Mark Backman
f6b6aa8766 fix: STTMuteFilter no longer sends STTMuteFrame 2025-11-06 11:53:32 -05:00
Aleix Conchillo Flaqué
d844829538 Revert "fix interruption task frame context ordering" 2025-11-05 12:14:03 -08:00
Aleix Conchillo Flaqué
e4212fb3c0 tests: add interruption strategies context ordering tests 2025-11-05 09:38:18 -08:00
Aleix Conchillo Flaqué
0f5030bafa tests: add unit test to check for idle timeout on swallowed frames 2025-10-31 08:45:56 -07:00
Paul Kompfner
ac5734d0ed Deprecate expect_stripped_words option from LLMAssistantAggregatorParams, when used with the newer LLMAssistantAggregator, which now handles word spacing automatically. 
This commit does not change how it works in the older `LLMAssistantContextAggregator`.
 2025-10-30 17:22:47 -04:00
Paul Kompfner
ea6e146f2d Update TestServiceSwitcher to exercise targeting system frames only to the active service 2025-10-22 11:14:27 -04:00
Aleix Conchillo Flaqué
1b9e96c016 PipelineTask: fix task cancellation issues 2025-10-16 13:16:19 -07:00
Aleix Conchillo Flaqué
62b7c3d3b2 PipelineTask: add on_pipeline_error event 2025-10-07 18:36:38 -07:00
Aleix Conchillo Flaqué
7c15a8c800 Revert "fix context order when using interruption strategies" 
This reverts commit de8ee96927.
 2025-10-07 17:42:35 -07:00
Aleix Conchillo Flaqué
de8ee96927 fix context order when using interruption strategies 2025-10-06 22:43:01 -07:00
Aleix Conchillo Flaqué
4dc1e15a99 frames: use OutputTransportMessage(Urgent)Frame instead of TransportMessage(Urgent)Frame 2025-10-01 16:00:42 -07:00
Paul Kompfner
6faa50ae5b Remove remaining usage of OpenAILLMContext throughout the codebase in favor of LLMContext, except for: 
- Usage in classes that are already deprecated
- Usage related to realtime LLMs, which don't yet support `LLMContext`
- Usage in (soon-to-be-deprecated) code paths related to `OpenAILLMContext` itself and associated machinery
 2025-09-24 16:35:03 -04:00
Aleix Conchillo Flaqué
2571cb2e69 tests: fix formatting 2025-09-23 15:28:07 -07:00
Aleix Conchillo Flaqué
15782be27c Merge pull request #2676 from golbin/main 
Fix audio buffer flush and silence handling
 2025-09-23 15:27:31 -07:00
Aleix Conchillo Flaqué
3d88b42e0b FrameProcessor: add before/after events for processed/pushed frames 2025-09-19 20:47:21 -07:00
Jin Kim
58f70e7e0d Add tests for audio buffer processor flush alignment 2025-09-18 22:19:32 +09:00
Aleix Conchillo Flaqué
42344125b1 tests: add unit tests for push_interruption_task_frame_and_wait() 2025-09-17 13:38:22 -07:00
Paul Kompfner
8bc3c89140 Fix a bug preventing usage of multiple ServiceSwitchers in a pipeline 2025-09-17 16:09:18 -04:00
Paul Kompfner
2cd2567a37 Add a unit tests validating that multiple ServiceSwitchers can be used in the same pipeline (currently failing) 2025-09-17 16:04:30 -04:00
Paul Kompfner
a12392182c Simplify, undoing the change allowing controlling ServiceSwitcher with immediate frames (SystemFrames). Service switcher frames are ControlFrames, which are easier to reason about. We can always build the immediate option later if needed (i.e. if there's sufficient user pull for it) 2025-09-17 15:35:02 -04:00
Paul Kompfner
b814b70e1e Allow controlling ServiceSwitcher with either immediate frames (SystemFrames) or in-order frames (ControlFrames). 
Immediate is the "default", i.e. has the more obvious name (e.g. `ManuallySwitchServiceFrame` v `ManuallySwitchServiceControlFrame`), since that's *probably* what users will want to reach for. Also, the immediate frames are more likely to behave like what we had before the last few commits, where the service switch would always "jump the queue" by having an immediate effect once it hit the `ServiceSwitcher` in the pipeline, jumping ahead of frames in front of it destined for the service.
 2025-09-17 15:35:02 -04:00
Paul Kompfner
a1f84e1b50 Remove extraneous unit tests 2025-09-17 15:35:02 -04:00
Paul Kompfner
e1b1dc16ec Add unit tests for ServiceSwitcher 2025-09-17 15:35:02 -04:00
Aleix Conchillo Flaqué
db71089f5e PipelineTask: add on_pipeline_finished event 
This deprecates `on_pipeline_stopped`, `on_pipeline_ended` and
`on_pipeline_cancelled`.
 2025-09-15 15:28:33 -07:00
kompfner
81cede2c60 Merge pull request #2653 from pipecat-ai/pk/llm-context-adapting-tests 
`LLMContext`-adapting unit tests
 2025-09-15 16:38:46 -04:00
Paul Kompfner
fe42187dc1 Implement LLMService.create_llm_specific_message() so that users don't need to just know what value of llm to provide to the LLMSpecificMessage constructor 2025-09-15 14:15:22 -04:00
Paul Kompfner
999e88c942 Add unit tests for AWSBedrockLLMAdapter.get_llm_invocation_params(), focusing on messages specifically 2025-09-15 12:08:21 -04:00
Paul Kompfner
c04df2f28b Add unit tests for AnthropicLLMAdapter.get_llm_invocation_params(), focusing on messages specifically 2025-09-15 11:55:48 -04:00
Paul Kompfner
100ef0ab5c Add unit tests for GeminiLLMAdapter.get_llm_invocation_params(), focusing on messages specifically 2025-09-15 11:38:23 -04:00
Paul Kompfner
42886d7105 Add unit tests for OpenAILLMAdapter.get_llm_invocation_params(), focusing on messages specifically. Also, fix a bug in OpenAILLMAdapter (found thanks to the unit tests) where it wasn't "unwrapping" LLMSpecificMessages. 2025-09-15 11:17:11 -04:00
Paul Kompfner
9f82c6b4a4 Add unit tests for run_inference 2025-09-12 11:07:11 -04:00
Aleix Conchillo Flaqué
57add41971 tests: add unit test for MinWordsInterruptionStrategy 2025-09-11 13:07:30 -07:00
Aleix Conchillo Flaqué
74b38b59d6 tests(utils): allow passing PipelineParams to run_test() 2025-09-11 13:02:21 -07:00
Aleix Conchillo Flaqué
0b21f8a6bd FrameProcessor: add push_interruption_task_frame_and_wait() 2025-09-11 09:19:44 -07:00
Aleix Conchillo Flaqué
9d9f10ae0e frames: StartInterruptionFrame is deprecated, use InterruptionFrame 2025-09-11 09:01:54 -07:00
Mark Backman
972546b24f Add IVR navigation (#2529) 2025-08-29 20:08:17 -04:00
Aleix Conchillo Flaqué
a5c5e069ba move pipecat.frames.frames.KeypadEntry to pipecat.audio.dtmf.types.KeypadEntry 2025-08-28 17:43:29 -07:00
Aleix Conchillo Flaqué
8ab81d289a PipelineTask: warn if CancelFrame doesn't reach the end 2025-08-26 10:36:33 -07:00
Aleix Conchillo Flaqué
8044c4170d PipelineRunner/PipelineTask: fix asyncio task cancellation 2025-08-21 23:50:00 -07:00
Aleix Conchillo Flaqué
24a628c85e remove watchdog timers and specific asyncio implementations 
Watchdog timers have been removed. They were introduced in 0.0.72 to help
diagnose pipeline freezes. Unfortunately, they proved ineffective since they
required developers to use Pipecat-specific queues, iterators, and events to
correctly reset the timer, which limited their usefulness and added friction.
 2025-08-21 18:09:56 -07:00
Aleix Conchillo Flaqué
28c729cc36 replace asyncio.wait_for for wait_for2.wait_for 2025-08-20 15:26:57 -07:00
Aleix Conchillo Flaqué
f387776985 add custom asyncio.wait_for() 
This patch uses `wait_for2` package to implement `asyncio.wait_for()` for
Python < 3.12.

In Python 3.12, `asyncio.wait_for()` is implemented in terms of
`asyncio.timeout()` which fixed a bunch of issues. However, this was never
backported (because of the lack of `async.timeout()`) and there are still many
remainig issues, specially in Python 3.10, in `async.wait_for()`.

See https://github.com/python/cpython/pull/98518
 2025-08-20 14:09:05 -07:00
Aleix Conchillo Flaqué
4e3af00b6d tests: try to use default SleepFrame time 2025-08-20 10:08:54 -07:00
Aleix Conchillo Flaqué
4015aedb86 tests: fix unit tests 2025-08-20 10:08:54 -07:00
Aleix Conchillo Flaqué
0508ddddfb WatchdogPriorityQueue: fix watchdog sentinel insertion 
We now force each inserted item in the priority queue to be a tuple and the
actual value to be last in the tuple. All the previous values in the tuple also
need to be numeric.
 2025-08-12 17:40:58 -07:00
Aleix Conchillo Flaqué
3f3d757581 tests: added WatchdogQueue and WatchdogPriorityQueue unit tests 2025-08-12 11:48:22 -07:00
Captain Caius
3fae00e067 Add VADUserStartSpeakingFrame VADUserStopSpeakingFrame to STTMuteFilter 2025-08-10 19:35:04 -07:00