Skip to content

SocratiCode 1.16.0: qualification receipt and dated decision update (the Linux pin stays 1.15.0) - #523

Merged
seathatflowsinourveins merged 2 commits into
mainfrom
res-scq-1160-receipt-20260930
Sep 30, 2026
Merged

seathatflowsinourveins merged 2 commits into
mainfrom
res-scq-1160-receipt-20260930

Conversation

@seathatflowsinourveins

@seathatflowsinourveins seathatflowsinourveins commented Sep 30, 2026 •

Copy link
Copy Markdown
Collaborator

Scope

  • What this PR changes: adds the qualification receipt for SocratiCode 1.16.0 (installed in an unreferenced host prefix, qualified, not adopted) and a dated update in the workstation-refresh decision. The Linux pin stays 1.15.0. An independent review of the branch that moves the pin found that a pin change itself starts cutover paths (adoption/update.md re-runs the bootstrap after one and relinks bin/socraticode; tools/adoption/render_config.py and the landscape sweep's scratch Codex homes render the pinned prefix into any configuration rendered from main), so that 26-file branch is kept unmerged for the cutover, which waits for Gate A's window W and for the 7-day cooldown (earliest 2026-10-05T12:23Z; the user's waiver of 2026-09-27 covered 1.15.0 only).
  • Base commit: f77612b6 (main, after OpenHands recipe lock: PyJWT 2.14.0 relock closes ten advisories (unblocks the required osv-scanner check) #525 cleared the repo-wide osv-scanner failure; rebased 2026-09-30 per the hot-file protocol, result byte-identical to the automatic merge).
  • Lane: lane:foundation (registering the receipt in manifests/evidence.json through the hot-file protocol does not make it shared).
  • Owned paths touched: evidence/receipts/, docs/decisions/2026-09-25-workstation-sota-refresh.md, manifests/evidence.json (last commit).

SOTA sources

  • SocratiCode v1.16.0 (tag v1.16.0, annotated tag object a50f72c5b79dcb8f69ac9bc68b21fe744d82d45b, commit 3d3a4a4d427cb0f29de4f5dd1360da52896f62df, released 2026-09-28T12:19:47Z): release notes and CHANGELOG, .github/workflows/ci.yml (upstream's job list and the Qdrant-backed regression step at the tag), src/constants.ts (SOCRATICODE_WATCHER default auto; new value git).
  • npm package socraticode@1.16.0: dist.integrity sha512-Lsei32zdHES3fCIRROn7kb1FY7sia0Oe2ZZMRUShZy8AytaqUB2F7/fF2+2dDLNNZxNAURP2Sb0doSpHv8JbMA==, dist.shasum 44a18caf99a4f4f466e3d618860143dec51bfd71, tarball sha256 3cb602cf2f50f863ef79dde56923d65b7ce840e2cc26070efa41743d2b4aaafe, published 2026-09-28T12:23:47Z; the registry gitHead is the tag commit's parent (the receipt records the content-identity check).
  • Upstream CI at the tag: 13 check runs, all successful (upstream's record, not our execution; no upstream workflow runs npm run test:integration or test:e2e).
  • Precedent: the 1.15.0 pin and cooldown handling, docs/decisions/2026-09-25-workstation-sota-refresh.md and PR Pin socraticode 1.15.0 and ccusage 20.0.26 with upstream-suite qualification receipts #446.

Evidence-class table

Claim Evidence class Command / receipt
The upstream suite passes at the tag: npm ci, npm run test:unit (107 files, 2,517 tests), check:release-versions, biome, tsc --noEmit, and the four Qdrant-backed regressions (12 tests) unchanged upstream tests, run by a scout and re-run by an independent verifier receipt data.upstream_suite
The same unit suite and tsc pass on the dependency tree that the recipe's --before=2026-09-28T12:30:00Z resolves (lockfile removed; 20 name@version pairs differ from the lockfile) unchanged upstream test code on another tree (our run) receipt data.dependency_tree
The npm tarball's sha256, sha512 and sha1 match the registry; its gitHead is the tag's parent with identical src/, and its dist/ equals a build of the tag local integration (registry read and content comparison) receipt data.registry_check, data.content_identity
A regression that needs Qdrant fails, rather than skips, when Qdrant is unreachable (REQUIRE_QDRANT=1): exit 1 local integration (failing control, run 2026-09-30) receipt limitation 3
The prefix is installed and no client configuration references it; no MCP call was made against it local integration (observation) receipt data.host_install, limitation 1
Cooldown not served and not waived; earliest cutover 2026-10-05T12:23Z source review (workstation-refresh record) receipt limitation 2

Local commands run

$ python3 scripts/validate.py
{"components": 69, "hashed_files": 8243, "profiles": 4, "receipts": 174, "status": "passed"}   exit 0
$ python3 scripts/evidence_manifest.py --check
{"files": 8243, "status": "passed"}                                                            exit 0
$ REQUIRE_QDRANT=1 QDRANT_URL=http://127.0.0.1:1 npx vitest run tests/integration/partial-upsert-skip.test.ts   (upstream tag v1.16.0)
Error: REQUIRE_QDRANT=1 but Qdrant is not reachable. This regression must fail rather than skip   exit 1 (the control)

Decision record

docs/decisions/2026-09-25-workstation-sota-refresh.md, dated update 2026-09-30 (alternatives: move the pin now, skip 1.16.0; overturn conditions: the cooldown decision or an early window close, a native call erroring on 1.16.0, an upstream regression).

Host evidence

This PR adds a receipt under evidence/receipts/, not evidence/hosts/.

  • scripts/validate.py (exit 0 above) validates the receipt and its registration.
  • Independent review: an Opus evidence-reviewer read the full 26-file branch on 2026-09-30 (1 high, 1 medium, 6 low). The high finding (moving the pin starts a cutover) is why this PR carries no pin change; the medium and the receipt-level lows (evidence kind now compatibility_attempt, dates, failing control, CI-coverage wording, Mac lag statement) are fixed here; the doc-level lows belong to the unmerged pin branch.
  • No platform_status change.

Residuals

  • The pin-move branch (26 files, receipt excluded) stays unmerged until the cutover; it must be rebased, re-registered through the hot-file protocol and re-reviewed then. It also needs the Gate A owner's approval (its freeze includes the pinned socraticode version).
  • The four Qdrant regressions ran on the lockfile tree only; upstream's git-refresh-integration job ran only in upstream CI (it needs Docker and Ollama); 11 of the 18 tests/integration files ran neither upstream nor here.
  • No MCP initialize, tools/list or codebase_health call was made against the installed 1.16.0 prefix.

Checklist

  • No workflows or Actions are changed.
  • No secrets are printed, logged or committed.
  • No new paid hosting, subscription or billing surface.
  • Peer-owned untracked files and worktrees were preserved.

🤖 Generated with Claude Code

@seathatflowsinourveins seathatflowsinourveins added the lane:foundation Foundation lane: Claude/Codex setup, hosts, memory, RAG, research, workers label Sep 30, 2026
Scout and others added 2 commits September 30, 2026 00:01
…the Linux pin stays 1.15.0

The upstream suite passes at v1.16.0 and on the tree the recipe resolves with --before=2026-09-28T12:30:00Z; the tarball equals
the tag's build; the prefix is installed and unreferenced. An independent review found that moving the pin starts cutover
paths (bootstrap re-run, render_config, scratch Codex homes), so the pin stays behind the cutover, which waits for Gate A's
window W and the 7-day cooldown (earliest 2026-10-05T12:23Z). The receipt records a failing control (REQUIRE_QDRANT=1 with
no Qdrant exits 1) and corrects the CI-coverage, evidence-kind and date wording the review flagged.

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
…cision record

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

lane:foundation Foundation lane: Claude/Codex setup, hosts, memory, RAG, research, workers

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant