Skip to content

Stream audio transcriptions and translations to the upstream. - #440

Draft
Mohammad-nassar10 wants to merge 2 commits into
vllm-project:mainfrom
Mohammad-nassar10:feat/audio-passthrough
Draft

Mohammad-nassar10 wants to merge 2 commits into
vllm-project:mainfrom
Mohammad-nassar10:feat/audio-passthrough

Conversation

@Mohammad-nassar10

@Mohammad-nassar10 Mohammad-nassar10 commented Oct 12, 2026 •

Copy link
Copy Markdown
Contributor

Summary

Forward POST /v1/audio/transcriptions and /v1/audio/translations to the upstream without buffering the multipart upload. A declared length over max_request_body_size_bytes is rejected with 413 before the upstream is called. An undeclared one is cut off with 413 once it passes the limit.
SSE responses for stream=true are relayed as they arrive.

Closes #404

Test Plan

Signed-off-by: Mohammad <mohammad.nassar@ibm.com>

This branch has not been deployed

No deployments
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.

Pass through remaining vLLM-served OpenAI endpoints

1 participant