Commit Graph

2836 Commits

Author SHA1 Message Date
Mark Backman
998ac32627 Merge pull request #2413 from captaincaius/fix-stt-mute-filter-vad-frames-20250810
Add VADUserStartSpeakingFrame VADUserStopSpeakingFrame to STTMuteFilter (fix #2412)
2025-08-11 13:54:34 -07:00
Filipi Fuchter
b7f12a96f1 Improving the latency of the HeyGenVideoService. 2025-08-11 09:11:17 -03:00
Filipi Fuchter
55d200e2d1 Added TTFB metrics for HeyGenVideoService and TavusVideoService. 2025-08-11 09:07:21 -03:00
Captain Caius
3fae00e067 Add VADUserStartSpeakingFrame VADUserStopSpeakingFrame to STTMuteFilter 2025-08-10 19:35:04 -07:00
Abdulrahman Alrahma
fc67cc3302 Support endpoint_id for AzureSTTService 2025-08-10 22:24:47 +01:00
Aleix Conchillo Flaqué
11119430cd TTSService: deprecate say() method 2025-08-07 13:48:32 -07:00
Paul Kompfner
9ea06c33f7 Bump deprecation version of LLMMessagesFrame, LLMUserResponseAggregator, and LLMAssistantResponseAggregator (the deprecation slipped past the 0.0.78 release) 2025-08-07 14:56:50 -04:00
Paul Kompfner
809ab0b7b6 Improve printed deprecation warning 2025-08-07 14:45:35 -04:00
Paul Kompfner
2b5db9c562 Remove redundant deprecation warning in docstring 2025-08-07 14:45:35 -04:00
Paul Kompfner
b4a886b59f Remove redundant deprecation warning in docstring 2025-08-07 14:45:35 -04:00
Paul Kompfner
711f740d9e Update UserResponseAggregator to avoid using the now-deprecated LLMUserResponseAggregator 2025-08-07 14:43:37 -04:00
Paul Kompfner
a0bda98c20 Update langchain to avoid using the now-deprecated LLMMessagesFrame, LLMUserResponseAggregator, and LLMAssistantResponseAggregator 2025-08-07 14:43:37 -04:00
Paul Kompfner
1c1bae35ab Mention deprecation in docstring for LLMMessagesFrame 2025-08-07 14:43:37 -04:00
Paul Kompfner
56c52c2cf2 Deprecate LLMUserResponseAggregator and LLMAssistantResponseAggregator, which depend on the now-deprecated LLMMessagesFrame. 2025-08-07 14:43:37 -04:00
Paul Kompfner
740aee1a1a Fix an issue in AnthropicLLMContext where we would never initialize turns_above_cache_threshold if we were upgrading from an OpenAILLMContext.
I noticed this when working on 22c-natural-conversation-mixed-llms.py
2025-08-07 14:43:37 -04:00
Paul Kompfner
64e48e4660 Deprecate LLMMessagesFrame.
The same functionality can be achieved using either:
- `LLMMessagesUpdateFrame` with the desired messages, with `run_llm` set to `True`
- `OpenAILLMContextFrame` with a new context initialized with the desired messages
2025-08-07 14:43:37 -04:00
Mark Backman
a5f5e418a8 feat: Add option to ignore emulated user speech while the bot is speaking 2025-08-07 14:08:11 -04:00
Filipi da Silva Fuchter
cec8a74293 Merge pull request #2386 from pipecat-ai/filipi/parallel_pipeline
Only push the StartFrame when all parallel pipelines have processed it
2025-08-07 14:20:30 -03:00
Filipi Fuchter
228a55ac1e Only push the StartFrame when all parallel pipelines have processed it. 2025-08-07 14:18:21 -03:00
Vanessa Pyne
ab9831daf0 Merge pull request #2382 from pipecat-ai/vp-trace-ignore-message
log: warning -> trace for elevenlabs tts unavailable context
2025-08-07 09:35:57 -05:00
Vanessa Pyne
e8c3f5dea6 Update src/pipecat/services/elevenlabs/tts.py
Co-authored-by: Mark Backman <mark@daily.co>
2025-08-07 09:23:33 -05:00
Mark Backman
4288b5e780 Merge pull request #2381 from pipecat-ai/aleix/runner-args-pipeline-idle-timeout
allow specifying PipelineTask idle timeout to runner arguments
2025-08-07 04:47:08 -07:00
Mark Backman
88de5dd415 Merge pull request #2383 from pipecat-ai/aleix/riva-stt-iterator-exception
properly handle concurrent.futures.CancelledError
2025-08-07 04:39:56 -07:00
Aleix Conchillo Flaqué
7ed14ad91f scripts(evals): added soniox, inworld and asyncai 2025-08-06 23:14:50 -07:00
Aleix Conchillo Flaqué
86c6141580 DailyTransport: handle future cancellation 2025-08-06 23:03:20 -07:00
Aleix Conchillo Flaqué
c97643c797 RivaSTTService: always use WatchdogQueue 2025-08-06 23:00:03 -07:00
Aleix Conchillo Flaqué
434d346079 RivaSTTService: handle future cancellation 2025-08-06 22:59:52 -07:00
vipyne
64ae8d2394 log: warning -> trace for elevenlabs tts unavailable context 2025-08-06 22:40:47 -05:00
Aleix Conchillo Flaqué
0b2b9f5f1b RunnerArguments: add pipeline_idle_timeout_secs 2025-08-06 19:35:40 -07:00
Filipi Fuchter
84fecabac5 Removing audio sleep from FastAPI and WebSocket server when they are not connected. 2025-08-06 14:02:51 -03:00
Filipi Fuchter
bbe01d10ef Fixed an issue in BaseOutputTransport where the loop could consume all CPU. 2025-08-06 12:42:58 -03:00
Mark Backman
4364990fd0 Merge pull request #2375 from fabrice404/gladia-region-selection
Gladia region selection
2025-08-06 07:01:24 -07:00
Fabrice Lamant
0594a203fc Add new region parameter to Gladia 2025-08-06 14:28:06 +02:00
Mark Backman
42094fb206 Update docs auto-generation to use uv 2025-08-05 20:37:27 -04:00
Aleix Conchillo Flaqué
58c41f112a DailyRunnerArguments: make body optional (fix) 2025-08-05 16:59:36 -07:00
Aleix Conchillo Flaqué
313fdc92a1 DailyRunnerArguments: make body optional 2025-08-05 16:39:18 -07:00
Aleix Conchillo Flaqué
d22d2da03d PipelineTask: always try to cancel things
In a previous commit we only cleanup things if the user run
`task.cancel()`. However, if the task finishes cleanly we were not cancelling
anything.
2025-08-05 16:24:59 -07:00
Aleix Conchillo Flaqué
f14cbae9b5 DailyRunnerArguments: make token optional
DailyTransport can get a None token value.
2025-08-05 15:46:12 -07:00
Aleix Conchillo Flaqué
8fe906438a Merge pull request #2358 from pipecat-ai/aleix/system-frames-queued
system frames are now queued
2025-08-05 15:09:52 -07:00
Aleix Conchillo Flaqué
a5ea6e1642 FrameProcessor: system frames are now queued
System frames are now queued. Before, system frames could be generated from any
task and would not guarantee any order which was causing undesired
behavior. Also, it was possible to get into some rare recursion issues because
of the way system frames were executed (they were executed in-place, meaning
calling `push_frame()` would finish after the system frame traversed all the
pipeline). This makes system frames more deterministic.
2025-08-05 15:05:50 -07:00
richtermb
e777e78510 Rename 'source' parameter to 'processor' in ErrorFrame class documentation for clarity. 2025-08-05 15:02:00 -07:00
Aleix Conchillo Flaqué
49a5a1e375 PipelineTask: improve task cancellation 2025-08-05 14:49:23 -07:00
Aleix Conchillo Flaqué
61cb45d61b PipelineTask: also wait on CancelFrame
Before CancelFrames didn't need to be waited for because system frames were
processed in-place and therefore calling push_frame() would finalize after it
traversed all the pipeline. Now, system frames are queued so we need to wait
until CancelFrame reaches the end of the pipeline.
2025-08-05 14:49:23 -07:00
Aleix Conchillo Flaqué
21e4f0d56d PipelineRunner: argument ordering 2025-08-05 14:38:55 -07:00
Aleix Conchillo Flaqué
627b44bac2 runner: use new RunnerArguments handle_sigint/handle_sigterm
This allow us to control applications behavior from the runner arguments, which
depen on the environment they run.
2025-08-05 14:38:55 -07:00
Aleix Conchillo Flaqué
e2a576beca RunnerArguments: add handle_sigint/handle_sigterm 2025-08-05 14:32:28 -07:00
Mark Backman
d422c57b52 Merge pull request #2304 from pipecat-ai/mb/cartesia-cjk-lang-support
CartesiaTTSService: Add CJK lang support for word timestamps
2025-08-05 14:08:53 -07:00
Mark Backman
a0e2a2754a Merge pull request #2327 from richtermb/richtermb/push-more-error-frames
Add source parameter to ErrorFrame and set it in FrameProcessor. Upda…
2025-08-05 14:04:52 -07:00
Mark Backman
b8d620c8bb Merge pull request #2362 from pipecat-ai/mb/aws-stt-languages
AWSTranscribeSTTService add support for new languages
2025-08-05 14:00:50 -07:00
Mark Backman
52cb23f8d5 Merge pull request #2364 from pipecat-ai/mb/11labs-default-model
ElevenLabs TTS services: revert to Turbo v2.5 as default model
2025-08-05 13:59:59 -07:00