Skip to content

[VIDPA-832] Sync public fork#2

Merged
asilvestre merged 136 commits into
mainfrom
feature/VIDPA-832/sync_public_fork
Feb 4, 2026
Merged

[VIDPA-832] Sync public fork#2
asilvestre merged 136 commits into
mainfrom
feature/VIDPA-832/sync_public_fork

Conversation

@asilvestre

Copy link
Copy Markdown

Please describe the changes in your PR. If it is addressing an issue, please reference that as well.

VIDPA-832

kompfner and others added 30 commits January 14, 2026 15:33
…changelog

Add PR 3420 to CHANGELOG (it was missing)
…s and `UserStoppedSpeakingFrame`s. The codebase is now consistent in broadcasting both types of frames up and downstream.
…trategies-timeout

fix to make on_user_turn_stop_timeout work with ExternalUserTurnStrat…
…urn-start-strategies

UserTurnController: reset user turn start strategies when turn triggered
…s-on-user-started-stopped-speaking-frames

Do a consistency pass on how we're sending `UserStartedSpeakingFrame`…
The Daily transport example had authentication issues. Keeping the
local audio example (07zb-interruptible-camb-local.py) which works.
- Add --voice-id CLI argument to example (default: 2681)
- Remove test_camb_quick.py from examples/ (tests belong in tests/)
- Update docstring with new usage
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
- 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
lukepayyapilli and others added 26 commits January 20, 2026 09:19
… triggers the bot response) only after sending the audio input start event, per the AWS team's recommendation
…reset-conversation

Fix `AWSNovaSonicLLMService.reset_conversation()`
If AIService subclasses implement start()/stop()/cancel() and exception are not
handled, execution will not continue and therefore the originator frames will
not be pushed. This would cause the pipeline to not be started (i.e. StartFrame
would not be pushed downstream) or stopped properly.
…tart-end-cancel

AIService: handle StartFrame/EndFrame/CancelFrame exceptions
Initialize client on StartFrame so errrors are reported within the pipeline.
Initialize client on StartFrame so errrors are reported within the pipeline.
…ts-error-handling

NVIDIA STT/TTS error handling
Log Daily participant and meeting session IDs upon successful join in…
…ble-frame-handling

fix: preserve UninterruptibleFrames in __reset_process_queue
…deo-queue-memory-leak

fix(livekit): prevent memory leak when video_in_enabled is False
Process audio as soon as we receive it from the generator. Previously, we were
reading from the generator and adding elements into a queue until there was no
more data, then we would process the queue.
…ts-improvements

NVIDIA STT/TTS performance improvements
…ent-on-start

CambTTSService: initialize client during StartFrame
…handle_bot_stopped_speaking

The _bot_speaking flag does not need to be set in this method,
so the redundant assignment has been removed.
refactor(user_mute): remove unnecessary _bot_speaking assignment in _handle_bot_stopped_speaking

@jgoconnell jgoconnell left a comment

Copy link
Copy Markdown

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

LGTM

@asilvestre asilvestre merged commit 0b123b4 into main Feb 4, 2026
10 checks passed
@asilvestre asilvestre deleted the feature/VIDPA-832/sync_public_fork branch February 4, 2026 16:58
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.