wx44wx/pipecat
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity
8,213 Commits 276 Branches 111 Tags
3e5be23bd80c99adfb3ab55f511f927f07599b9b
Commit Graph

506 Commits

Author SHA1 Message Date
Ian Lee
3e5be23bd8 fix(inworld): close context at end of turn instead of relying on idle timeout 
The Inworld WS TTS plugin previously relied on the base TTS service's 3-second AUDIO_CONTEXT_TIMEOUT to detect when audio was done, then sent close_context in on_audio_context_completed. This added unnecessary latency before TTSStoppedFrame was emitted.

The original implementation likely borrowed this idea from the 11labs' impelementation. But it's likely better to mirror the Cartesia plugin pattern where on_audio_context_completed is a no-op because the server signals completion directly.

Now close_context is sent in on_turn_context_completed (right after flush_context), so the server responds with contextClosed immediately after the last audio byte. The existing receive handler already calls remove_audio_context on contextClosed, which exits the audio context handler cleanly.
 2026-03-13 12:52:07 -07:00
kompfner
f2debd9b1d Merge pull request #3963 from pipecat-ai/pk/improve-claude-changelog-skill 
Improve changelog skill: prioritize user-facing language and update e…
 2026-03-09 13:00:32 -04:00
Filipi da Silva Fuchter
0c5e936aa5 Merge pull request #3936 from pipecat-ai/filipi/fix_push_aggregation 
Fixed TTS context not being appended to the assistant message history
 2026-03-09 11:14:38 -04:00
Paul Kompfner
7f9169269c Improve changelog skill: prioritize user-facing language and update example changelog 2026-03-09 10:45:33 -04:00
filipi87
8ec160f71e Making the changelog more user friendly. 2026-03-09 11:37:11 -03:00
Filipi da Silva Fuchter
1e615cd095 Merge pull request #3962 from pipecat-ai/filipi/smallwebrtc_queue 
Queuing the messages received before the data channel is ready
 2026-03-09 10:29:05 -04:00
filipi87
097f9c0896 Fixed to push LLMAssistantPushAggregationFrame when the base TTSService class is responsible for pushing the TTSStoppedFrame. 2026-03-09 11:04:09 -03:00
Filipi da Silva Fuchter
16336a3ea4 Merge pull request #3937 from pipecat-ai/filipi/fix_orphan_function_call 
Fix context summarization leaving orphaned tool responses in kept context.
 2026-03-09 09:19:17 -04:00
Mark Backman
9eaa99c8e2 Merge pull request #3957 from pipecat-ai/mb/user-turn-completion-system-instruction 
Move turn completion instructions to system_instruction
 2026-03-09 09:17:06 -04:00
filipi87
aa693bb5ee Adding changelog entry for the SmallWebRTCConnection fix. 2026-03-09 10:11:40 -03:00
Mark Backman
8a02e6fbc5 Merge pull request #3959 from ajmeraharsh/fix/livekit-call-state-updated-args 
fix(livekit): remove redundant self arg in on_call_state_updated
 2026-03-09 08:38:12 -04:00
Mark Backman
d85ba75dda Merge pull request #3953 from pipecat-ai/mb/deepgram-flux-on-the-fly 
Add on-the-fly Configure support for Deepgram Flux STT
 2026-03-09 08:36:00 -04:00
ajmeraharsh
ae6f159b18 chore: add changelog entry for #3959 2026-03-09 09:15:03 +04:00
Aleix Conchillo Flaqué
3b947b7844 Add changelog for #3947 2026-03-08 19:02:51 -07:00
Mark Backman
64155e8f06 Add changelog for #3957 2026-03-08 10:44:45 -04:00
radhikagpt1208
b14c8e0e94 Fix turn completion mixin not resetting state after each LLM response 2026-03-08 08:46:45 -04:00
Mark Backman
4ebdacdea2 Add changelog for #3953 2026-03-07 08:48:11 -05:00
Mark Backman
1cfaea2007 Address code review feedback 2026-03-07 07:42:42 -05:00
filipi87
3000037dec Changelog entries for the TTS improvements and fixes. 2026-03-06 16:16:25 -03:00
Aleix Conchillo Flaqué
593b75bc8b Update foundational examples to use "user" role 
Use system_instruction on LLM service constructors instead of adding
system messages to LLMContext. Messages added to context now use
"user" role.
 2026-03-06 09:53:33 -08:00
filipi87
524b87f087 Adding changelog entry for the summarization fix. 2026-03-06 11:45:20 -03:00
Mark Backman
79131dd6c6 Merge pull request #3930 from dakshdua/main 
Add `push_empty_transcripts` param to `BaseWhisperSTTService` to push received empty transcripts downstream
 2026-03-05 19:25:15 -05:00
Aleix Conchillo Flaqué
fda4cb6732 Add changelog for #3932 2026-03-05 16:16:41 -08:00
Daksh Dua
789ce2fd5e Add param to push empty transcripts 2026-03-05 16:16:24 -08:00
zack
380726cfd3 Update AssemblyAI turn detection example to use keyterms_prompt 
Change the commented example from prompt string format to keyterms_prompt
list format for better clarity and consistency with API best practices.
 2026-03-05 15:47:54 -05:00
zack
11024ccc2c Add changelog entries for vad_threshold and parameter cleanup 2026-03-05 15:32:09 -05:00
Aleix Conchillo Flaqué
1aadb8bd73 Merge pull request #3918 from pipecat-ai/aleix/system-instruction-openai-anthropic 
Wire up system_instruction in OpenAI, Anthropic, and AWS Bedrock
 2026-03-04 17:40:00 -08:00
Aleix Conchillo Flaqué
3c60b0c8af Add changelog for #3918 2026-03-04 17:37:32 -08:00
vipyne
aa31ced32f add changelog for 3916 2026-03-04 16:58:28 -06:00
Mark Backman
4b9fe5afe3 Merge pull request #3915 from pipecat-ai/mb/function_call_timeout_secs-error-msg 
Add per-tool function call timeout_secs
 2026-03-04 15:01:34 -05:00
Mark Backman
97e4e7c647 Add changelog for #3915 2026-03-04 09:42:01 -05:00
Mark Backman
6789aee9e8 Add changelog for #3914 2026-03-03 20:09:26 -05:00
Mark Backman
ca0ec16373 Merge pull request #3889 from ai-coustics/goedev/aic-voice-focus-and-memoryview-fix 
AIC Voice Focus version update & concurrency safety issue on audio buffer.
 2026-03-03 09:28:13 -05:00
filipi87
fc905a7ef5 Merge branch 'main' into filipi/deepgram 
# Conflicts:
#	src/pipecat/services/deepgram/stt_sagemaker.py
 2026-03-03 10:54:30 -03:00
aconchillo
62260454a2 Update changelog for version 0.0.104 2026-03-02 21:23:39 -08:00
Mark Backman
c79a739c85 Merge pull request #3856 from zkleb-aai/assemblyai-u3-rt-pro 
Assemblyai u3 rt pro
 2026-03-02 20:28:28 -05:00
Mark Backman
aad1211a57 Merge pull request #3885 from pipecat-ai/mb/latency-breakdown 
Add latency breakdown to UserBotLatencyObserver
 2026-03-02 19:27:35 -05:00
Aleix Conchillo Flaqué
141b0ee014 Merge pull request #3902 from pipecat-ai/aleix/deepgram-sagemaker-move 
Move Deepgram SageMaker modules to sagemaker/ subpackage
 2026-03-02 15:25:17 -08:00
Aleix Conchillo Flaqué
303616599f Add changelog for #3902 2026-03-02 15:20:52 -08:00
zack
32773b42d6 Improve terminology: rename file and replace 'STT mode' with 'AssemblyAI turn detection' 
- Rename 07o-interruptible-assemblyai-stt.py -> 07o-interruptible-assemblyai-turn-detection.py
- Replace 'STT mode' with 'AssemblyAI turn detection mode' throughout codebase
- Replace 'Mode 1'/'Mode 2' with descriptive 'Pipecat turn detection'/'AssemblyAI turn detection'
- Update changelog to use 'built-in turn detection' terminology
- Addresses PR feedback about confusing terminology
 2026-03-02 18:08:46 -05:00
filipi87
49c73bb0a3 Merge branch 'main' into filipi/lemonslice 
# Conflicts:
#	README.md
#	uv.lock
 2026-03-02 19:24:52 -03:00
zkleb-aai
5c2ca0ce64 Update changelog/3856.changed.md 
Co-authored-by: Mark Backman <m.backman@gmail.com>
 2026-03-02 17:04:54 -05:00
Mark Backman
a4796a2373 Merge pull request #3898 from pipecat-ai/mb/revert-processing-metrics-deprecation 
Revert processing metrics deprecation
 2026-03-02 16:39:02 -05:00
Aleix Conchillo Flaqué
44466cfa07 Merge pull request #3896 from pipecat-ai/aleix/broadcast-interruption 
Add broadcast_interruption() to FrameProcessor
 2026-03-02 13:36:39 -08:00
Aleix Conchillo Flaqué
741ff14d3a Rename changelog files to use PR #3896 and mark breaking change 2026-03-02 13:26:45 -08:00
Aleix Conchillo Flaqué
4a61d5bfad Add broadcast_interruption() to FrameProcessor 
Replace the round-trip push_interruption_task_frame_and_wait() mechanism
with broadcast_interruption(), which pushes an InterruptionFrame both
upstream and downstream directly from the calling processor.

This eliminates race conditions (transcription arriving before the
InterruptionFrame comes back), swallowed-event timeouts (frame blocked
before reaching the sink), and the complexity of _wait_for_interruption
flag / queue bypass / frame.complete() obligations.

- Add broadcast_interruption() to FrameProcessor
- Deprecate push_interruption_task_frame_and_wait() (delegates to new method)
- Remove event field and complete() from InterruptionFrame/InterruptionTaskFrame
- Remove _wait_for_interruption flag and all special-case logic
- Remove frame.complete() calls in stt_mute_filter and llm_response_universal
- Update all 17 call sites to use broadcast_interruption()
- Update tests
 2026-03-02 13:26:45 -08:00
Mark Backman
d0ecb3c7a8 Revert "Deprecate processing metrics (ProcessingMetricsData)" (#3852) 
This reverts commit 127b52bad5.
 2026-03-02 16:26:29 -05:00
Mark Backman
8f66272de7 Update changelog 2026-03-02 16:16:38 -05:00
Mark Backman
a738a4d82b Add function call latency tracking to LatencyBreakdown 2026-03-02 16:11:43 -05:00
Mark Backman
ddba1b84a9 Add first-bot-speech latency to UserBotLatencyObserver 
Measure time from ClientConnectedFrame to first BotStartedSpeakingFrame,
emitting a one-time on_first_bot_speech_latency event with breakdown.
 2026-03-02 16:11:43 -05:00