Replaced deprecated amazon-transcribe SDK with new aws-sdk-transcribe-streaming #4111
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
…-streaming - Replace `amazon-transcribe` dependency with `aws_sdk_transcribe_streaming` in pyproject.toml - Bump minimum Python version to 3.12, as requried by the official AWS SDK. - Update `STT` implementation to use the new SDK client and models. - Refactor `SpeechStream._run` to use `TranscribeStreamingClient` from the new SDK. - Update audio input streaming to use `AudioStreamAudioEvent`. - Update transcript event processing to match the new SDK's event structure. - Remove custom credential resolver logic in favor of `EnvironmentCredentialsResolver`
Prevent race conditions and unhandled exceptions during participant disconnects: Changes: - Implement a graceful shutdown sequence: close input stream first, then wait for output stream to finish. - Use `asyncio.shield` to protect inner tasks from immediate cancellation, allowing for proper cleanup. - Suppress `concurrent.futures.InvalidStateError` in `handle_transcript_events` to avoid noise from AWS CRT bindings during shutdown. - Ensure `gather_future` is awaited to prevent "exception never retrieved" warnings. - Use `contextlib.suppress` for cleaner exception handling during resource cleanup.
Contributor
Author
|
Unit test I don't think this has anything to do with the changes in my PR. |
longcw
reviewed
Nov 28, 2025
Contributor
longcw
left a comment
longcw
left a comment
There was a problem hiding this comment.
Thanks for creating the pr! Looks good to me, something nit:
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
3 participants
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
The AWS Python SDK used by livekit-plugins-aws in the STT implementation (
amazon-transcribe) has known critical CPU bottleneck performance issues and has been deprecated by AWS in favor ofaws-sdk-transcribe-streaming.amazon-transcribeissue: [Bug] BufferableByteStream causes 100% CPU usage in non-blocking scenarios awslabs/amazon-transcribe-streaming-sdk#121livekit/agentsissue: AWS STT plugin extremely slow when using multi-user transcription #3739This PR replaces the old, deprecated, low-performance AWS SDK with the new official AWS SDK. I have tested the problematic scenario explained in this issue (#3739) and it is no longer a problem when using this PR. Transcribing multiple users from the same LiveKit Agent using livekit-plugin-aws STT is now possible, with no delays. The scope of the commit is quite small, so I do not expect any significant side effects.