wx44wx/pipecat
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity
7,583 Commits 276 Branches 111 Tags
9bb712a47b4ece26dde95fb2b41e8e866c4ed1df
Commit Graph

4386 Commits

Author SHA1 Message Date
Neil Ruaro
003c24ca6e Make model parameter explicit in docstring example 2026-01-16 01:18:37 +08:00
Neil Ruaro
ed120d014d Add model-specific sample rates, transport example, and fix audio buffer alignment 2026-01-16 01:18:37 +08:00
Neil Ruaro
e76a3d04f0 Update Camb TTS to 48kHz sample rate 2026-01-16 01:18:37 +08:00
Neil Ruaro
641d17007f Clean up Camb TTS service and tests 2026-01-16 01:18:37 +08:00
Neil Ruaro
9293b5f24a Migrate Camb TTS service from raw HTTP to official SDK 
- Replace aiohttp with camb SDK (AsyncCambAI client)
- Add support for passing existing SDK client instance
- Simplify API: no longer requires aiohttp_session parameter
- Update example to use simplified initialization
- Rewrite tests to mock SDK client instead of HTTP servers
 2026-01-16 01:18:37 +08:00
Neil Ruaro
c1f3cbd1d4 Yield TTSAudioRawFrame directly instead of calling private method 2026-01-16 01:18:37 +08:00
Neil Ruaro
78fa2ab65e Update default voice ID, fix MARS naming, and clean up example 2026-01-16 01:18:37 +08:00
Neil Ruaro
56da2caeed Update Camb.ai TTS inference options 2026-01-16 01:18:37 +08:00
Neil Ruaro
a541d65255 Update MARS model names to mars-flash, mars-pro, mars-instruct 
Rename model identifiers from mars-8-* to the new naming convention:
- mars-8-flash -> mars-flash (default)
- mars-8 -> removed
- mars-8-instruct -> mars-instruct
- Added mars-pro
 2026-01-16 01:18:37 +08:00
Neil Ruaro
7ae0d651d6 added cambai tts integration 2026-01-16 01:18:36 +08:00
kompfner
24082b84f2 Merge pull request #3453 from pipecat-ai/pk/consistency-pass-on-user-started-stopped-speaking-frames 
Do a consistency pass on how we're sending `UserStartedSpeakingFrame`…
 2026-01-15 09:24:14 -05:00
Aleix Conchillo Flaqué
9e705ce768 UserTurnController: reset user turn start strategies when turn triggered 2026-01-14 18:20:29 -08:00
Mark Backman
f3993f1775 fix to make on_user_turn_stop_timeout work with ExternalUserTurnStrategies 2026-01-14 20:10:56 -05:00
Paul Kompfner
e107902b14 Do a consistency pass on how we're sending UserStartedSpeakingFrames and UserStoppedSpeakingFrames. The codebase is now consistent in broadcasting both types of frames up and downstream. 2026-01-14 18:47:15 -05:00
Ashot
c4ae4025f3 Adjustments of Async TTS for multicontext websocket support 2026-01-14 16:33:30 +04:00
Ashot
15067c678d adapt Async TTS to updated AudioContextTTSService 2026-01-14 15:45:27 +04:00
Ashot
5ae592f38e Improve Async TTS interruption handling by using AudioContextTTSService class and add changelog fragments 2026-01-14 15:45:27 +04:00
Ashot
9cdbc56be3 Fix TTFB metric and add multi-context WebSocket support for Async TTS 2026-01-14 15:45:27 +04:00
Aleix Conchillo Flaqué
0d6bdbee10 LLMContextAggregatorPair: make strategy logs less verbose 2026-01-13 15:11:22 -08:00
Aleix Conchillo Flaqué
248dac3a9d Merge pull request #3420 from pipecat-ai/pk/fix-gemini-3-parallel-function-calls 
Fix parallel function calling with Gemini 3.
 2026-01-13 14:40:33 -08:00
Paul Kompfner
be49a54856 Fast-exit in the fix for parallel function calling with Gemini 3, if we can determine up-front that there's no work to do 2026-01-13 17:32:20 -05:00
Aleix Conchillo Flaqué
bd9ee0d646 Merge pull request #3434 from pipecat-ai/aleix/context-appregator-pair-tuple 
context aggregator pair tuple
 2026-01-13 14:12:51 -08:00
Mark Backman
442e0e582d Merge pull request #3431 from pipecat-ai/mb/update-realtime-examples-transcript-handler 
Update GeminiLiveLLMService to push thought frames, update 26a for new transcript events
 2026-01-13 17:10:40 -05:00
Aleix Conchillo Flaqué
1ab3bf2ef6 LLMContextAggregatorPair: instances can now return a tuple 2026-01-13 14:01:53 -08:00
Mark Backman
bb00d223c9 Update 26a to use context aggregator transcription events 2026-01-13 17:01:10 -05:00
Aleix Conchillo Flaqué
5612bf513b LLMContext: fix create_audio_message 2026-01-13 13:53:34 -08:00
Paul Kompfner
30fbcfbf71 Rework fix for parallel function calling with Gemini 3 2026-01-13 16:33:59 -05:00
Mark Backman
5d90f4ea06 Merge pull request #3428 from pipecat-ai/mb/fix-tracing-none-values 
Fix TTS, realtime LLM services could return unknown for model_name
 2026-01-13 15:40:10 -05:00
Aleix Conchillo Flaqué
e268c73c41 LLMAssistantAggregator: cache function call requested images 2026-01-13 12:10:08 -08:00
Aleix Conchillo Flaqué
d3c57e2da0 UserImageRawFrame: don't deprecate request field 2026-01-13 11:56:13 -08:00
Aleix Conchillo Flaqué
02eace5a16 UserImageRequestFrame: don't deprecate function call related fields 2026-01-13 11:55:55 -08:00
Mark Backman
15bc1dd999 Update GeminiLiveLLMService to push Thought frames when thought content is returned 2026-01-13 14:13:00 -05:00
Paul Kompfner
b937956dc8 Fix request_image_frame and usage 2026-01-13 13:23:01 -05:00
Mark Backman
efbc0c8510 Fix TTS, realtime LLM services could return unknown for model_name 2026-01-13 12:12:15 -05:00
Himanshu Gunwant
d0f227189c fix: openai llm model name is unknown (#3422) 2026-01-13 11:55:52 -05:00
Mark Backman
ae59b3ba36 Merge pull request #3404 from poseneror/feature/gladia-vad-events 
feat(gladia): add VAD events support
 2026-01-13 11:26:56 -05:00
Paul Kompfner
8812686b17 Fix parallel function calling with Gemini 3. 
Gemini expects parallel function calls to be passed in as a single multi-part `Content` block. This is important because only one of the function calls in a batch of parallel function calls gets a thought signature—if they're passed in as separate `Content` blocks, there'd be one or more missing thought signatures, which would result in a Gemini error.
 2026-01-13 11:03:38 -05:00
kompfner
8b0f0b5bb4 Merge pull request #3425 from pipecat-ai/pk/gemini-3-flash-new-thinking-levels 
Add Gemini 3 Flash-specific thinking levels
 2026-01-13 11:02:53 -05:00
Mark Backman
a298ce3b41 Merge pull request #3424 from pipecat-ai/mb/tts-append-trailing-space 
Add append_trailing_space to TTSService to prevent vocalizing trailin…
 2026-01-13 10:42:40 -05:00
Mark Backman
31daa889e8 Add append_trailing_space to TTSService to prevent vocalizing trailing punctuation; update DeepgramTTSService and RimeTTSService to use the arg 2026-01-13 10:38:54 -05:00
Paul Kompfner
76a058178e Add Gemini 3 Flash-specific thinking levels 2026-01-13 09:50:59 -05:00
poseneror
3304b18ac2 Add should_interrupt + broadcast user events 2026-01-13 14:27:35 +02:00
poseneror
b95a6afe77 feat(gladia): add VAD events support 
Add support for Gladia's speech_start/speech_end events to emit
UserStartedSpeakingFrame and UserStoppedSpeakingFrame frames.

When enable_vad=True in GladiaInputParams:
- speech_start triggers interruption and pushes UserStartedSpeakingFrame
- speech_end pushes UserStoppedSpeakingFrame
- Tracks speaking state to prevent duplicate events

This allows using Gladia's built-in VAD instead of a separate VAD
in the pipeline.
 2026-01-13 14:27:35 +02:00
Mark Backman
cd3290df1c Small cleanup for task creation in SpeechmaticsSTTService 2026-01-12 16:00:32 -05:00
Mark Backman
90ded6658d Merge pull request #3403 from pipecat-ai/mb/inworld-tts-add-keepalive 
InworldTTSService: Add keepalive task
 2026-01-12 13:31:24 -05:00
Mark Backman
7e97fb80a5 Merge pull request #3392 from pipecat-ai/mb/websocket-service-connection-closed-error 
Add reconnect logic to WebsocketService in the event of ConnectionClo…
 2026-01-12 13:11:43 -05:00
Aleix Conchillo Flaqué
46b4f9f29b Merge pull request #3413 from pipecat-ai/aleix/fix-assistant-thought-aggregation 
LLMAssistantAggregator: reset aggregation after adding the thought, not before
 2026-01-12 09:21:42 -08:00
Aleix Conchillo Flaqué
ec20d72aba LLMAssistantAggregator: reset aggregation after adding the thought, not before 2026-01-12 09:18:13 -08:00
Varun Pratap Singh
3e982f7a4a refactor: rename audio_packet_bytes to fixed_audio_packet_size 2026-01-12 22:11:39 +05:30
Varun Pratap Singh
e96595fe59 feat: update FastAPI WebSocket transport and add Vonage serializer 2026-01-12 17:50:38 +05:30