Repository navigation
fix(kiro): preserve GPT-5.6 Max reasoning via Responses - #9163
Merged
diegosouzapw merged 1 commit intoAug 6, 2026
Merged
diegosouzapw merged 1 commit into
diegosouzapw merged 1 commit into
Conversation
VXNCXNX
marked this pull request as ready for review
August 1, 2026 19:15
diegosouzapw
merged commit Aug 6, 2026
a5a78fd
into
cheaper-inference:release/v3.8.50
13 checks passed
diegosouzapw
added a commit
that referenced
this pull request
Aug 6, 2026
…9554) * fix(quality): reconcile inherited file-size drift on the release tip 13 files sit above their frozen LOC on the clean tip 8180b49 (measured by the gate itself). The PR-mode base-relative check (#8522) correctly lets innocent PRs pass, but per-PR rebaselines were lost across successive conflict resolutions of this hot file during the 08-05/06 merge batch — so the absolute mode (nightly, local runs) is permanently red and stops distinguishing real growth from inherited drift. Frozen values updated to the measured tip, each annotated with the merged PR that grew the file (#9024 #9324 #9329 #9193 #9332 #9228 #9236 #9314 #9260 #8934 #9196 #9163); executors default.ts and kiro.ts (above the 1000 cap with no frozen entry) join the frozen set. * fix(quality): prune orphaned ESLint suppressions and clear the 5 unsuppressed errors The 'No new ESLint warnings' job reds the whole queue with exit 2: 'There are suppressions left that do not occur anymore' — the 08-05 merge batch removed code whose violations were frozen in eslint-suppressions.json, leaving orphaned entries (673->670 files, 4338->4333 violations after eslint --prune-suppressions). The full-tree run also surfaced 5 real unsuppressed errors merged with the batch, fixed here instead of suppressed (new violations must be fixed, per policy): 4x no-explicit-any in tests/unit/catalog-order-contract.test.ts ((conn as any).id -> typed cast) and 1x react/no-unescaped-entities in the agent-bridge SetupWizard (#9095). Also restores the _comment policy header the successive hot-file conflict resolutions had dropped (TS7 debt freeze provenance + prune policy). * fix(quality): absorb the two file-size growths merged while this PR was in CI The base kept moving during the reconcile cycle: #9184 grew src/sse/handlers/chat.ts 1857->1877 and #9005 grew open-sse/executors/default.ts 1027->1042. Re-measured on the merged tree; gate back to 0 violations. * fix(tests): move the orphaned RTL ratchet test to a collected path as node:test #8828 added tests/unit/scripts/check-rtl-ratchet.test.ts — a path no runner collects (the node:test globs enumerate an explicit subdir list without scripts/, and vitest.config.ts never included it), so the file NEVER ran and the test-discovery orphan gate reds the queue. Moved to tests/unit/ (collected by node:test) and converted from vitest describe/it/expect to node:test+assert to match the runner and the sibling check-*.test.ts files. 5/5 green under the real runner. --------- Co-authored-by: diegosouzapw <diegosouzapw@users.noreply.github.com>
muhamadgalihsaputra
pushed a commit
to niyatna/NiyatnaRoute
that referenced
this pull request
Sep 27, 2026
…rence#9163) Validated in local merge-train (devbox-vm-06-dev002) @ combined-tip (FAST gates green: static + changed tests + vitest — only pre-existing audit.test.ts flake). Evidence: /home/diegosouzapw/dev/proxys/OmniRoute/.claude/worktrees/merge-train-20260805-213228-suite.log
muhamadgalihsaputra
pushed a commit
to niyatna/NiyatnaRoute
that referenced
this pull request
Sep 27, 2026
…heaper-inference#9554) * fix(quality): reconcile inherited file-size drift on the release tip 13 files sit above their frozen LOC on the clean tip 30be158 (measured by the gate itself). The PR-mode base-relative check (cheaper-inference#8522) correctly lets innocent PRs pass, but per-PR rebaselines were lost across successive conflict resolutions of this hot file during the 08-05/06 merge batch — so the absolute mode (nightly, local runs) is permanently red and stops distinguishing real growth from inherited drift. Frozen values updated to the measured tip, each annotated with the merged PR that grew the file (cheaper-inference#9024 cheaper-inference#9324 cheaper-inference#9329 cheaper-inference#9193 cheaper-inference#9332 cheaper-inference#9228 cheaper-inference#9236 cheaper-inference#9314 cheaper-inference#9260 cheaper-inference#8934 cheaper-inference#9196 cheaper-inference#9163); executors default.ts and kiro.ts (above the 1000 cap with no frozen entry) join the frozen set. * fix(quality): prune orphaned ESLint suppressions and clear the 5 unsuppressed errors The 'No new ESLint warnings' job reds the whole queue with exit 2: 'There are suppressions left that do not occur anymore' — the 08-05 merge batch removed code whose violations were frozen in eslint-suppressions.json, leaving orphaned entries (673->670 files, 4338->4333 violations after eslint --prune-suppressions). The full-tree run also surfaced 5 real unsuppressed errors merged with the batch, fixed here instead of suppressed (new violations must be fixed, per policy): 4x no-explicit-any in tests/unit/catalog-order-contract.test.ts ((conn as any).id -> typed cast) and 1x react/no-unescaped-entities in the agent-bridge SetupWizard (cheaper-inference#9095). Also restores the _comment policy header the successive hot-file conflict resolutions had dropped (TS7 debt freeze provenance + prune policy). * fix(quality): absorb the two file-size growths merged while this PR was in CI The base kept moving during the reconcile cycle: cheaper-inference#9184 grew src/sse/handlers/chat.ts 1857->1877 and cheaper-inference#9005 grew open-sse/executors/default.ts 1027->1042. Re-measured on the merged tree; gate back to 0 violations. * fix(tests): move the orphaned RTL ratchet test to a collected path as node:test cheaper-inference#8828 added tests/unit/scripts/check-rtl-ratchet.test.ts — a path no runner collects (the node:test globs enumerate an explicit subdir list without scripts/, and vitest.config.ts never included it), so the file NEVER ran and the test-discovery orphan gate reds the queue. Moved to tests/unit/ (collected by node:test) and converted from vitest describe/it/expect to node:test+assert to match the runner and the sibling check-*.test.ts files. 5/5 green under the real runner. --------- Co-authored-by: diegosouzapw <diegosouzapw@users.noreply.github.com>
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
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
reasoning.effort=maxfor Kiro GPT-5.6 Sol, Terra, and Luna through/v1/responses.additionalModelRequestFields.reasoning.effortpayload for GPT-5.6.maxeffort tier in the Kiro model catalog.Root cause
Responses normalization did not receive the routed provider/model, so Kiro GPT-5.6
maxcould be downgraded toxhigh. The Kiro GPT-5.6 family also rejects Claude-shapedoutput_config,thinking, andmax_tokensfields. Kiro documentsreasoning.effortwithnone,low,medium,high,xhigh, andmaxfor these models: https://kiro.dev/docs/cli/chat/effort/Validation
npm run typecheck:core: passed.git diff --check: passed.npm run test:coverage: coverage thresholds passed, with 10 unrelated failures in the dirty local environment. They include local untracked documentation drift, macOSreadarrayincompatibility, and existing Kiro/MCP/machine-id tests. No failure was in the changed GPT files.kr/gpt-5.6-lunawithreasoning.effort=maxreturned HTTP 200 withpongand usage, and/v1/modelsexposed all six Kiro effort tiers. Existing Grok plain,x_search, andweb_searchsmokes remained green.Scope
This branch is based on
release/v3.8.50and contains only the GPT/Kiro corrections plus tests. The existing xAI Agent Tools patch and local deployment/runbook changes are intentionally excluded.