fix(httpserver): use chunked transfer encoding for SSE streaming responses#101
Merged
Merged
Conversation
…onses SSE responses previously skipped Transfer-Encoding: chunked and relied on Connection: close to signal body end. While valid per HTTP/1.1 spec, this breaks intermediate proxies (notably Go's httputil.ReverseProxy used by NPS) that expect either Content-Length or chunked framing — they misparse the raw SSE data as chunked encoding, producing errors like "invalid byte in chunk length". Unify all StreamingResponse output to use chunked encoding, matching the behavior of production SSE servers (OpenAI, Anthropic APIs). Closes #100
This was referenced Jun 12, 2026
Oaklight
added a commit
to Oaklight/llm-rosetta
that referenced
this pull request
Jun 12, 2026
Re-vendor httpserver from zerodep to pick up the fix for SSE streaming responses not using Transfer-Encoding: chunked. This caused Go-based reverse proxies (NPS httputil.ReverseProxy) to misparse SSE data, producing "invalid byte in chunk length" errors and intermittent ConnectionRefused under concurrent load. Upstream: Oaklight/zerodep#101 Closes #274
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
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Summary
text/event-stream) streaming responses now useTransfer-Encoding: chunkedlike all other streaming responses, instead of flushing raw bytes withConnection: closehttputil.ReverseProxy(used by NPS, and other Go-based reverse proxies) that misparse raw SSE data as chunked encodingContext
Discovered when running
llm-rosetta-gatewaybehind an NPS tunnel. NPS logs showed repeated errors:The old approach (no
Content-Length, noTransfer-Encoding, rely on connection close) is valid per HTTP/1.1 spec but incompatible with proxies that re-frame the response body.Test plan
ruff check+ruff formatcleanCloses #100