Skip to content
This repository was archived by the owner on Nov 8, 2026. It is now read-only.

Stream Sink results without buffering complete responses - #113

Merged
liran merged 6 commits into
mainfrom
randy/streaming-responses
Sep 20, 2026
Merged

liran merged 6 commits into
mainfrom
randy/streaming-responses

Conversation

@liran

@liran liran commented Sep 20, 2026 •

Copy link
Copy Markdown
Collaborator

Stream Read, Write, Query and Scan through storage, Engine and Gateway without retaining complete responses. Private Forward uses typed results, native gRPC status and EOF; remove Budget accounting and the legacy unary/chunk protocols. Preserve safe endpoint failover before decoding and count streamed business failures without retaining payloads. Existing local message limits remain, with no new memory setting or Worker change. Pin the merged SDK #25 and matching qualification #29.

All six previously failing conformance groups passed against local Elasticsearch 8.19 and OpenSearch 3.8 in 109 seconds. Updated assertions enforce per-result limits, rejection of oversized returned documents before commit, and terminal errors without continuation for incomplete pages. Affected unit/component race tests, vet/staticcheck, workflow lint and performance-tool compilation passed; search coverage remains 89.1% against the unchanged 88% floor. GitHub CI runs on the updated revisions.

@liran
liran merged commit 4479caf into main Sep 20, 2026
15 checks passed
@liran
liran deleted the randy/streaming-responses branch September 20, 2026 10:44
Sign up for free to subscribe to this conversation on GitHub. Already have an account? Sign in.

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant