wx44wx/pipecat
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity
7,953 Commits 276 Branches 111 Tags
dee94b3cb86101d395b3d1ea8865ea2f1ca458f3
Commit Graph

7953 Commits

This Branch
This Branch
All Branches
Author SHA1 Message Date
Mark Backman
dee94b3cb8 Merge pull request #3795 from omChauhanDev/fix/realtime-cancel-not-active 
fix(realtime): handle response_cancel_not_active as non-fatal
 2026-03-01 07:29:59 -05:00
Om Chauhan
ece4343839 changed log level to debug 2026-03-01 12:25:42 +05:30
Mark Backman
17205c1647 Merge pull request #3871 from rupesh-svg/fix/rtvi-processor-double-insert 
Fix PipelineTask double-inserting RTVIProcessor with custom RTVIObserver
 2026-02-27 19:34:46 -05:00
Mark Backman
2a776d0c1e Merge pull request #3873 from rimelabs/matt/rime/add_speedAlpha_param_to_arcana 
[RimeTTS] Add `speedAlpha` parameter support to the `arcana` model
 2026-02-27 19:27:56 -05:00
Rupesh
56f2564ed1 Use local variable instead of instance variable for RTVI prepend decision 
Replace _rtvi_external instance variable with a local prepend_rtvi flag
since it is only used during __init__ to decide whether to prepend the
RTVIProcessor to the pipeline.
 2026-02-27 14:45:37 -08:00
macaki
000d38e253 [Rime] Both mist and arcana now support the speedAlpha parameter. 2026-02-27 15:17:23 -07:00
Filipi da Silva Fuchter
36edef489e Merge pull request #3863 from pipecat-ai/filipi/manual_summarization 
Manual context summarization
 2026-02-27 16:46:37 -05:00
filipi87
d077a810ae Fixing context summarization tests 2026-02-27 18:42:50 -03:00
filipi87
0839e3813f Refactoring the examples to use the new context summarization classes. 2026-02-27 18:42:39 -03:00
filipi87
69414e8a5a Added example 54b-context-summarization-manual-openai.py demonstrating on-demand summarization triggered via a function call tool. 2026-02-27 18:42:23 -03:00
filipi87
dfd0a515f3 Changelog entries for the context summarization improvements. 2026-02-27 18:42:13 -03:00
filipi87
ed7f0a2c08 Adding support for on-demand summarization 2026-02-27 18:41:55 -03:00
filipi87
08d93ce9b6 Renamed LLMAssistantAggregatorParams fields for clarity. 2026-02-27 18:41:17 -03:00
filipi87
f11d4b6944 Refactored LLMContextSummarizationConfig into two focused classes, LLMContextSummaryConfig and LLMAutoContextSummarizationConfig. 2026-02-27 18:40:41 -03:00
filipi87
51a3310e78 Added LLMSummarizeContextFrame: push this frame anywhere in the pipeline to trigger on-demand context summarization (e.g. from a function call tool). 2026-02-27 18:39:57 -03:00
Rupesh
6f33aff0c6 Fix PipelineTask double-inserting RTVIProcessor when custom RTVIObserver is provided 
When the user places an RTVIProcessor inside their pipeline and provides
a custom RTVIObserver subclass in observers, PipelineTask correctly
detects both and logs "skipping default ones." However it then
unconditionally prepends self._rtvi to the pipeline, causing the
processor to appear twice in the frame chain.

Track whether the RTVIProcessor was found externally (inside the user
pipeline) vs created internally. Only prepend it when created internally.

Fixes #3867
 2026-02-27 13:29:01 -08:00
Mark Backman
4eb993c980 Merge pull request #3868 from wollerman/wollerman/numba-version-pin-update 
fix: Update numba version pin from == to >=0.61.2
 2026-02-27 16:04:20 -05:00
Mark Backman
83e29eb478 Merge pull request #3855 from pipecat-ai/mb/context-summarization-improvements 
Improve context summarization with dedicated LLM, timeout, and observability
 2026-02-27 15:24:38 -05:00
Matt
acff172bf2 create changelog entry 2026-02-27 14:52:37 -05:00
Mark Backman
9747e8da4a Merge pull request #3866 from pipecat-ai/mb/fix-docs-workflow-version 
Fix docs workflow to add auto-docs label
 2026-02-27 13:09:36 -05:00
Mark Backman
8fc63352d9 Merge pull request #3865 from pipecat-ai/mb/elevenlabs-realtime-stt-finalized 
Set finalized flag on ElevenLabs Realtime STT for manual commit strategy
 2026-02-27 13:09:17 -05:00
Matt
6ebfea4746 update numba version pin to >= 2026-02-27 12:44:31 -05:00
Mark Backman
f74af9b9c7 Always apply a timeout to summarization LLM calls 
Even when summarization_timeout is explicitly set to None, use a
DEFAULT_SUMMARIZATION_TIMEOUT (120s) fallback so the LLM call can
never hang indefinitely. Applied in both LLMService and the dedicated
LLM path in LLMContextSummarizer.
 2026-02-27 12:09:00 -05:00
Mark Backman
82c249608f Move dedicated LLM summarization into LLMContextSummarizer 
The dedicated LLM logic lived in LLMAssistantAggregator, creating two
code paths and requiring the aggregator to call a private LLMService
method. Move it into the summarizer which already owns the config and
summarization lifecycle, keeping the aggregator handler as a single-line
upstream push.
 2026-02-27 12:09:00 -05:00
Mark Backman
98e737b4e9 Add tests for context summarization improvements 
Cover summary message role, template, on_summary_applied event,
summarization timeout, and dedicated LLM routing/error handling.
 2026-02-27 12:08:43 -05:00
Mark Backman
ec9ddb3199 Add changelog entries for context summarization improvements (#3855) 2026-02-27 12:07:34 -05:00
Mark Backman
712305c5b1 Add example 54c showing custom context summarization 2026-02-27 12:07:34 -05:00
Mark Backman
be8ea818c8 Add on_summary_applied event for observability 
Emits a SummaryAppliedEvent after context summarization completes,
  providing message counts so applications can track compression
  metrics.
 2026-02-27 12:07:34 -05:00
Mark Backman
50710e9c3f Add summarization timeout to prevent hung LLM calls 
Adds a configurable summarization_timeout (default 120s) that cancels
  summary generation if the LLM hangs. On timeout, an error result is
  returned so _summarization_in_progress resets and future
  summarizations are unblocked.
 2026-02-27 12:07:34 -05:00
Mark Backman
a489bfaf00 Add optional dedicated LLM for context summarization 
Adds an  field to LLMContextSummarizationConfig that allows
  routing summarization to a separate LLM service (e.g., Gemini Flash)
  instead of the pipeline's primary model. This avoids paying for
  expensive inference when compressing context in long-running sessions.
 2026-02-27 12:07:34 -05:00
Mark Backman
945a523eed Add configurable summary_message_template to LLMContextSummarizationConfig 
Allows applications to customize how the summary is wrapped when
  injected into context (e.g., XML tags, custom delimiters) so system
  prompts can distinguish summaries from live conversation.
 2026-02-27 12:07:34 -05:00
Mark Backman
790c434a08 Update summary message role: use user instead of assistant 
The context summary is information provided to the assistant, not
  something the assistant said.
 2026-02-27 12:07:34 -05:00
Filipi da Silva Fuchter
db40a354be Merge pull request #3794 from omChauhanDev/fix/context-summarization-llm-specific-message 
skipping provider-specific messages during summarization
 2026-02-27 10:57:34 -05:00
filipi87
aa6d3b38b3 Add explanatory comments for LLMSpecificMessage guards in context summarization, amd fixed the missing guard in LLMContextSummarizer._apply_summary when searching for the first system message. 2026-02-27 12:53:25 -03:00
Mark Backman
41d6470e4a Fix docs workflow: add auto-docs label, remove version info 2026-02-27 10:39:37 -05:00
Mark Backman
601822e3e5 Add changelog for PR #3865 2026-02-27 10:25:48 -05:00
Mark Backman
3a32d91c66 Set finalized flag on ElevenLabs Realtime STT transcriptions for manual commit strategy 2026-02-27 10:21:10 -05:00
Filipi da Silva Fuchter
35b3803ebc Merge pull request #3845 from pipecat-ai/filipi/fix_tts_speak_frame 
Add TTSSpeakFrame.push_assistant_aggregation to force context flush after TTS.
 2026-02-27 09:59:33 -05:00
filipi87
3b427a47b6 Fixing Piper test. 2026-02-27 11:57:11 -03:00
filipi87
d701c3427c Changelog entry for the TTSSpeakFrame fix. 2026-02-27 11:57:03 -03:00
filipi87
1f45e80f9d Updated the 52-live-translation.py example to demonstrate the fix 2026-02-27 11:56:52 -03:00
filipi87
bc6f8e51de Fixed TTSSpeakFrame not automatically committing spoken text to the conversation context when used outside of an LLM response (e.g., for bot greeting messages or injected speech) 2026-02-27 11:56:44 -03:00
filipi87
deba2515f9 Added a new LLMAssistantPushAggregationFrame control frame that signals LLMAssistantAggregator to immediately flush its text buffer to the conversation context 2026-02-27 11:56:36 -03:00
Mark Backman
127b52bad5 Merge pull request #3852 from pipecat-ai/mb/deprecate-processing-metrics 
Deprecate processing metrics (ProcessingMetricsData)
 2026-02-27 09:50:29 -05:00
Mark Backman
0697f72dae Merge pull request #3864 from pipecat-ai/mb/auto-docs-update 
Add automated docs update workflow
 2026-02-27 09:36:27 -05:00
Mark Backman
c259a6a73b Deprecate processing metrics (ProcessingMetricsData) 
Add deprecation warnings to start_processing_metrics() and
stop_processing_metrics() on FrameProcessorMetrics and FrameProcessor.
Mark ProcessingMetricsData as deprecated in docstring. All existing
behavior is preserved — the warnings inform users that these will be
removed in a future version.
 2026-02-27 09:22:29 -05:00
Mark Backman
3e04f5d05f Add GitHub Actions workflow to auto-update docs on PR merge 
Runs Claude Code Action after PRs merge to main when source files
in services/transports/serializers/processors/audio/turns/observers/pipeline
are changed. Creates a docs PR on pipecat-ai/docs with targeted edits
following the existing update-docs skill instructions.
 2026-02-27 09:18:15 -05:00
Mark Backman
2a6a993869 Merge pull request #3850 from rupesh-svg/fix/genesys-remove-audio-chunk-logging 
Remove verbose audio chunk logging from GenesysAudioHookSerializer
 2026-02-26 21:52:54 -05:00
Rupesh
bbaa79fef0 Add changelog for PR #3850 2026-02-26 14:00:34 -08:00
Rupesh
fff9db0d8f Remove verbose audio chunk logging from GenesysAudioHookSerializer 
Fixes #3777
 2026-02-26 13:51:05 -08:00