Repository navigation
docs(rfc): mark rfc-host-memory-guard accepted - #39290
iamwhatever wants to merge 1 commit into
Conversation
#39188 was approved and merged 2026-11-18 20:57Z but kept status: draft. The repository owner approved the design as proposed in the conductor chat on 2026-11-18. Record it as accepted so the implementation PRs #39270 and #39289 can cite an accepted RFC. Co-authored-by: Joe Guo <zejiangg@amazon.com>
First Principles Review (Opus 5.5) — ⏭️ skippedRevision |
UX Review (Opus 5.5) — ⏭️ skippedRevision |
Design Review (Opus 5.5) — ✅ PASSDesign-level review of Design-Verdict: PASS Docs-only status flip. Acceptance takes effect only when a maintainer approves and merges this PR, which is what the RFC asks for. Notes
[DESIGN-REVIEWED] a22d72e Description read[DESCRIPTION-READ] 8c1b6f15e4df38874aca5ba3a8ccfd2f5adf9ae7033dba4f674f22e535ce69f8 That sha256 names everything this verdict read: the PR description, plus the 1 evidence file(s) the workflow collected from it and handed this review. It covers that evidence because the media strip replaces every attachment URL with the same placeholder, so an edit swapping one attachment for another would leave the description's captured text identical while what this review judged changed. Recompute it with |
Opus 5.5 Review — ✅ no blocking findingsReviewed Verdict parsed from the review's SHA-scoped output markers for commit False positive or not applicable? A repository writer can comment: |
GPT 6.1 Review — ✅ no blocking findingsGPT 6.1 completed its review of This comment is updated in place on each push. Review detailsNo findings. False positive or not applicable? A repository writer can comment: |
|
This PR is the step that accepts the host memory guard RFC. The RFC names the repository owner as the decider, so that decision stays with the owner or a human maintainer. An automated approval should not record it. The owner or a maintainer will post the decision here. What the review checked on this head:
The status flip is small and consistent: the front matter, the Status line and the index row all move together. The body says plainly that the merge review is what records the decision. Optional notes for when a human approves:
|
Code change approver recommendation: A person should decideWhy: the decider estimates a 10% chance that this change introduces a defect, within the 12% limit for approving on its own, but it is held at tier 2: design-doc-gate: D. Docs-only change that moves rfc-host-memory-guard from draft to accepted in its front matter, its Status line and the index row; the design text is unchanged. The repo's rules (request-for-change/README.md:313-318, ci-and-reviews.md:3486-3489) make a standalone status-flip PR a proposal that takes effect when a maintainer merges it, and the First Principles lane reads status from the base commit, so this PR cannot unblock its own implementation PRs. The design it accepts lets Crew stop running process trees without the user opting in, and the limits are sound: only processes started from Crew's sessions, with the gateway, session hosts, MCP servers and agent runtimes excluded; processes re-identified before signalling; off when agent.resource_critical_gb is 0; Linux only; and required tests cover these. No code, config, limits or stored data change. CI and all review lanes pass; the one failed check (Publish readiness signal) hit GitHub API rate limits. Remaining points are about wording and where the approval is recorded, and do not block. Tier: 2, held for a person: design-doc-gate: D. An automated recommendation, not a review: a maintainer still decides. It approves on its own only when the estimated chance of a defect is at most 12% and the change has at most 400 changed lines. Decider v3c (217fe607) · commit |
Problem / Motivation
Goal: Record the host memory guard RFC as
accepted, so its implementation PRs can cite an accepted RFC.docs/request-for-change/rfc-host-memory-guard.mdwas approved and merged in #39188 (2026-11-18T20:57:55Z) but its front matter and Status line still readdraft.Why it matters
The First Principles lane reads the RFC status, so while it says
draftit blocks the two implementation PRs #39270 and #39289.Not a goal
/ai-reviewoverride.What changed (motivation → approach → change)
The repository owner approved this design as proposed in the conductor chat on 2026-11-18: stop internal non-gateway trees, heaviest test trees first, never the gateway; write a log line and a chat message; the conductor is told and actively manages memory and disk with high-level guidance.
The RFC says a maintainer records acceptance on its Status line. This PR proposes that line; the maintainer records it by approving and merging this PR.
Changes, in the same format as #33417:
docs/request-for-change/rfc-host-memory-guard.md: front matterstatus: draft->status: accepted; theStatus:paragraph now says accepted and cites docs(rfc): host memory guard stops the heaviest non-gateway tree, and the conductor keeps the host healthy #39188 and the owner's approval.docs/request-for-change/README.md: the index row for this RFC moves fromdrafttoacceptedand drops "acceptance requested from a maintainer".Backwards compatibility
Compatible: docs-only status change; nothing that works today stops working.
Removes nothing: only the RFC status text changes.
Tests
python3 scripts/docs_lint.py: All documentation checks passed.Proof: not provable -- docs or config only
Manual verification
N/A — docs-only change;
docs_lint.pycovers it.Real-machine verification
N/A -- the diff changes two Markdown files and runs nothing.
Related Issues
Refs #22681. Follows #39188. Unblocks #39270 and #39289.
Checklist
feat|fix|docs|style|refactor|perf|test|chore|ci|build|revert: ...); PR Hygiene does not count commitsContribution License Agreement
Contributed under the project's Apache-2.0 license.