Skip to content

fix: flush supported stt streams on vad end of speech - #7285

Open
chenghao-mou wants to merge 5 commits into
mainfrom
chenghao/fix/AGT-3514-vad-eos-stt-flush
Open

chenghao-mou wants to merge 5 commits into
mainfrom
chenghao/fix/AGT-3514-vad-eos-stt-flush

Conversation

@chenghao-mou

Copy link
Copy Markdown
Member

Flush opted-in STT streams on local VAD EOS so ElevenLabs manual commits produce final transcripts. Preserve the audio-only stt_node contract and stream reuse across handoffs.

Buffering custom nodes must disable manual_flush. Mixed fallback chains require every provider to opt in. Live-provider validation is pending.

Related: @IngLP's report #4087 and @longcw's streaming-adapter proposal #4043. This implements session-level flushing.

Addresses AGT-3514 (Linear)

Initial prompt and agent context

Model: GPT-6

do we actually send _FlushSentinel to 11labs stt from VAD EOS?

okay, create a fix to flush STT with VAD EOS when STT supports it.

we can't change stt node contract, right? that means breaking change for any custom stt node implementation?

i guess it is okay, if they are buffering, they can disable by turn off manual flush.

@chenghao-mou chenghao-mou added the review_effort:high Needs broad context and substantial validation label Sep 15, 2026
@chenghao-mou
chenghao-mou marked this pull request as ready for review September 15, 2026 13:22
@chenghao-mou
chenghao-mou requested a review from a team as a code owner September 15, 2026 13:23
devin-ai-integration[bot]

This comment was marked as resolved.

devin-ai-integration[bot]

This comment was marked as resolved.

devin-ai-integration[bot]

This comment was marked as resolved.

devin-ai-integration[bot]

This comment was marked as resolved.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

review_effort:high Needs broad context and substantial validation

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant