refactor: consolidate verified modularization train and closeout records

[lidge-jun]

Summary

• Consolidate the contents of 14 already-reviewed modularization PRs into one integration PR against dev, as requested by the maintainer. Original branches/checkpoint refs are preserved; original PRs will be closed as superseded only after verified landing.
• Preserve current provider fields, Combo cooldown defaults, Reserve admission/policy, image/file-ID handling, public exports and single-owner caches while extracting 31 focused leaves.
• Include the accumulated split-train records with explicit historical/deferred notices. Further debt layers are deferred; this does not resolve all 68 original rows.
• Repair one HTTP/SSE test fixture so native upstream WebSocket fallback is deterministic. Production transport, authentication assertions and test deadlines are unchanged.
• Delivery order is explicit: merge this aggregate after its exact-head CI passes, then perform at least two complete main-to-merged-dev regression PABCD cycles. Those post-merge cycles are not claimed complete here.

Supersedes (after landing): #3557, #3559, #3566, #3567, #3570, #3574, #3577, #3580, #3583, #3585, #3590, #3594, #3599, #3611.

Verification

• Isolated remote verification on source ef7914d4a51899f49baa141990f79750b4c75cf9: bun run test 19,236 pass / 16 skip / 0 fail; focused server-auth/native-WS/provider fixture 161 pass / 1 skip / 0 fail; bun run typecheck and bun run privacy:scan passed.
• All 14 rebased staging tips passed their focused remote checks. Pinned-main 14-module / 244-export snapshot passed 15 runtime checks on main and the aggregate.
• Main/dev baselines, dashboard component/build/lint checks and isolated rendered comparisons are recorded. Independent source/security, fixture, documentation and receipt reviews passed.
• No local suites, release or live-service changes. Our later changes contain closeout records only. The branch also incorporates newly landed dev changes (feat(gui): unify quota activation in Advanced settings #3662 and feat(responses): bounded canonical upstream WS reuse #3664); this PR's hosted checks must validate the exact updated final head/current-dev integration before merge.
• Earlier interrupted discovery-worker run remains an unexplained observation, not a claimed fix. The separate WebSocket fixture dependency was reproduced and repaired with a RED/GREEN no-egress oracle. The unchanged opt-in Fast/cache ambiguity is disclosed in the outcome record.
• Hosted CI: pending this PR. No missing, failed or cancelled check is counted as passing.

Checklist

• Scope stays focused and avoids unrelated cleanup.
• Docs or release notes were updated when needed.
• Security-sensitive changes were reviewed for secrets, auth, and unsafe defaults.

Summary by CodeRabbit

• New Features

  • Improved image normalization with tiered resizing, validation, caching, and memory-aware eviction.
  • Added broader client configuration export support, including model metadata, provider settings, and loopback configurations.
  • Expanded vision planning, image handling, and routed-backend support.
  • Added desktop execution contracts and enhanced tool naming, schemas, and guidance.
  • Improved Claude inbound request translation and response parser coverage.

• Documentation

  • Added comprehensive historical planning, verification, and closeout records for the modularization work.

• Tests

  • Added compatibility, export-boundary, dependency, schema, and fixture-isolation coverage.

[chatgpt-codex-connector]

Codex Review Summary

This comment shows the latest Codex review activity on this pull request.

Review	Status	Commit	Review trigger
📝 Code Review	✅ Completed 2026-09-05T14:15:51.997815Z	0d071d8	PR opened

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review" or "@codex security review".

Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings.

[github-actions]

✅ Deterministic PR hygiene checks passed.

[coderabbitai]

Caution

Review failed

The pull request is closed.

ℹ️ Recent review info

⚙️ Run configuration

Configuration used: Path: .coderabbit.yaml

Review profile: ASSERTIVE

Plan: Team

Run ID: eb928c2b-b709-4531-8fce-33991f6150f7

📥 Commits

Reviewing files that changed from the base of the PR and between ef9c538 and c923cb2.

📒 Files selected for processing (159)

• devlog/_plan/260905_now_split_train/000_3_main_export_baseline.json
• devlog/_plan/260905_now_split_train/000_plan.md
• devlog/_plan/260905_now_split_train/001_stale_check.md
• devlog/_plan/260905_now_split_train/002_layer_map.md
• devlog/_plan/260905_now_split_train/003_parent_decisions.md
• devlog/_plan/260905_now_split_train/004_roadmap_lock.md
• devlog/_plan/260905_now_split_train/005_delivery_evidence_refresh.md
• devlog/_plan/260905_now_split_train/006_macos_recovery_verification_debt.md
• devlog/_plan/260905_now_split_train/007_wp450_check_progress.md
• devlog/_plan/260905_now_split_train/008_serial_ci_coordination.md
• devlog/_plan/260905_now_split_train/009_pending_preflight_findings.md
• devlog/_plan/260905_now_split_train/010_lib_redact.md
• devlog/_plan/260905_now_split_train/020_lib_errors.md
• devlog/_plan/260905_now_split_train/030_lib_upstream_retry.md
• devlog/_plan/260905_now_split_train/040_providers_openai_tiers.md
• devlog/_plan/260905_now_split_train/050_providers_registry_a.md
• devlog/_plan/260905_now_split_train/060_providers_registry_b.md
• devlog/_plan/260905_now_split_train/070_providers_registry_c.md
• devlog/_plan/260905_now_split_train/080_adapters_anthropic_image_normalize.md
• devlog/_plan/260905_now_split_train/090_adapters_anthropic_a.md
• devlog/_plan/260905_now_split_train/100_adapters_anthropic_b.md
• devlog/_plan/260905_now_split_train/105_cursor_desktop_executor_contract.md
• devlog/_plan/260905_now_split_train/110_adapters_cursor_tool_definitions.md
• devlog/_plan/260905_now_split_train/120_adapters_cursor_catalog.md
• devlog/_plan/260905_now_split_train/130_adapters_cursor_images.md
• devlog/_plan/260905_now_split_train/140_adapters_cursor_request_builder.md
• devlog/_plan/260905_now_split_train/150_adapters_cursor_protobuf_events.md
• devlog/_plan/260905_now_split_train/160_adapters_xai_tool_schema.md
• devlog/_plan/260905_now_split_train/170_adapters_command_code.md
• devlog/_plan/260905_now_split_train/180_adapters_ollama_native.md
• devlog/_plan/260905_now_split_train/190_vision_index.md
• devlog/_plan/260905_now_split_train/200_images_artifacts.md
• devlog/_plan/260905_now_split_train/210_responses_parser.md
• devlog/_plan/260905_now_split_train/220_responses_namespace_tool_compat.md
• devlog/_plan/260905_now_split_train/230_server_responses_agent_task_recovery.md
• devlog/_plan/260905_now_split_train/240_server_responses_collaboration.md
• devlog/_plan/260905_now_split_train/250_claude_inbound.md
• devlog/_plan/260905_now_split_train/260_server_claude_messages.md
• devlog/_plan/260905_now_split_train/270_server_system_env.md
• devlog/_plan/260905_now_split_train/280_server_management_logs_usage_routes.md
• devlog/_plan/260905_now_split_train/290_server_management_lab_routes.md
• devlog/_plan/260905_now_split_train/300_codex_prompt_layers_a.md
• devlog/_plan/260905_now_split_train/310_codex_prompt_layers_b.md
• devlog/_plan/260905_now_split_train/320_combos_types.md
• devlog/_plan/260905_now_split_train/330_codex_subagent_defaults.md
• devlog/_plan/260905_now_split_train/340_codex_cli_install_provenance.md
• devlog/_plan/260905_now_split_train/350_routing_trace.md
• devlog/_plan/260905_now_split_train/360_oauth_github_copilot.md
• devlog/_plan/260905_now_split_train/370_codex_log_guard_inspect.md
• devlog/_plan/260905_now_split_train/380_codex_log_guard_protection.md
• devlog/_plan/260905_now_split_train/390_codex_log_guard_maintenance.md
• devlog/_plan/260905_now_split_train/400_clients_config_export_a.md
• devlog/_plan/260905_now_split_train/410_clients_config_export_b.md
• devlog/_plan/260905_now_split_train/420_cli_opencode.md
• devlog/_plan/260905_now_split_train/430_cli_minimax.md
• devlog/_plan/260905_now_split_train/440_integrations_state.md
• devlog/_plan/260905_now_split_train/445_server_port_probe_disposal.md
• devlog/_plan/260905_now_split_train/447_port_probe_verification_progress.md
• devlog/_plan/260905_now_split_train/450_cli_status.md
• devlog/_plan/260905_now_split_train/460_cli_provider.md
• devlog/_plan/260905_now_split_train/470_client_hub_client.md
• devlog/_plan/260905_now_split_train/480_lab_events_validate.md
• devlog/_plan/260905_now_split_train/490_lab_ledger_store.md
• devlog/_plan/260905_now_split_train/500_lab_artifacts_sanitize.md
• devlog/_plan/260905_now_split_train/510_lab_fabric_observe.md
• devlog/_plan/260905_now_split_train/520_lab_fabric_scratch.md
• devlog/_plan/260905_now_split_train/530_lab_conformance_executor.md
• devlog/_plan/260905_now_split_train/540_lab_automation_persistence.md
• devlog/_plan/260905_now_split_train/550_lab_public_community.md
• devlog/_plan/260905_now_split_train/560_lab_projection_verification.md
• devlog/_plan/260905_now_split_train/570_lab_projection_verdicts.md
• devlog/_plan/260905_now_split_train/580_components_storage_workspace_StorageWorkspace.md
• devlog/_plan/260905_now_split_train/590_pages_Storage_a.md
• devlog/_plan/260905_now_split_train/600_pages_Storage_b.md
• devlog/_plan/260905_now_split_train/610_pages_integrations_overview_clients.md
• devlog/_plan/260905_now_split_train/620_pages_integrations_IntegrationsOverview.md
• devlog/_plan/260905_now_split_train/625_pages_integrations_IntegrationsOverview_b.md
• devlog/_plan/260905_now_split_train/630_pages_compatibility_matrix_api.md
• devlog/_plan/260905_now_split_train/640_pages_CompatibilityMatrix.md
• devlog/_plan/260905_now_split_train/650_combo_workspace_data.md
• devlog/_plan/260905_now_split_train/660_components_combo_workspace_detail_panel.md
• devlog/_plan/260905_now_split_train/670_pages_ClaudeDesktop.md
• devlog/_plan/260905_now_split_train/680_components_MemoryObservabilityCard.md
• devlog/_plan/260905_now_split_train/690_components_provider_workspace_ProviderSettings.md
• devlog/_plan/260905_now_split_train/700_pages_dashboard_shared.md
• devlog/_plan/260905_now_split_train/710_components_QuotaBars.md
• devlog/_plan/260905_now_split_train/720_release_notes_a.md
• devlog/_plan/260905_now_split_train/730_release_notes_b.md
• devlog/_plan/260905_now_split_train/740_test.md
• devlog/_plan/260905_now_split_train/750_disposable_host_codex_service_composed_acceptance.md
• devlog/_plan/260905_now_split_train/800_closeout.md
• devlog/_plan/260905_now_split_train/801_closeout_regression_matrix.md
• devlog/_plan/260905_now_split_train/810_first_rebase_regression.md
• devlog/_plan/260905_now_split_train/811_first_execution.md
• devlog/_plan/260905_now_split_train/812_first_cycle_outcome.md
• devlog/_plan/260905_now_split_train/813_http_fixture_isolation.md
• devlog/_plan/260905_now_split_train/820_second_regression_delivery.md
• devlog/_plan/260905_now_split_train/830_delivery_first.md
• src/adapters/anthropic-image-codec.ts
• src/adapters/anthropic-image-normalize.ts
• src/adapters/cursor/desktop-executor-contract.ts
• src/adapters/cursor/native-exec-desktop.ts
• src/adapters/cursor/tool-definitions.ts
• src/adapters/cursor/tool-guidance.ts
• src/adapters/cursor/tool-naming.ts
• src/adapters/cursor/tool-schemas.ts
• src/adapters/xai-schema-analysis.ts
• src/adapters/xai-tool-schema.ts
• src/claude/inbound-content-options.ts
• src/claude/inbound-model-options.ts
• src/claude/inbound-records.ts
• src/claude/inbound.ts
• src/clients/config-export.ts
• src/clients/config-export/constants.ts
• src/clients/config-export/contracts.ts
• src/clients/config-export/dsh.ts
• src/clients/config-export/mcode.ts
• src/clients/config-export/model-metadata.ts
• src/clients/config-export/omp.ts
• src/clients/config-export/zcode.ts
• src/codex/log-guard/inspect-schema.ts
• src/codex/log-guard/inspect.ts
• src/codex/prompt-layers.ts
• src/codex/prompt-layers/encoding.ts
• src/codex/prompt-layers/paths.ts
• src/codex/prompt-layers/revision.ts
• src/codex/prompt-layers/toml-edit.ts
• src/codex/prompt-layers/toml-read.ts
• src/combos/identifiers.ts
• src/combos/types.ts
• src/lib/redact-folding.ts
• src/lib/redact.ts
• src/providers/openai-tiers-destination.ts
• src/providers/openai-tiers.ts
• src/responses/parser-content.ts
• src/responses/parser-text-format.ts
• src/responses/parser-tools.ts
• src/responses/parser.ts
• src/server/system-env-shell.ts
• src/server/system-env.ts
• src/types/provider.ts
• src/vision/image-rewrite.ts
• src/vision/index.ts
• src/vision/plan.ts
• tests/adapters/anthropic/anthropic-image-normalize.test.ts
• tests/adapters/openai/openai-provider-option.test.ts
• tests/claude-integration/claude-inbound.test.ts
• tests/codex-integration/codex-log-guard-inspect.test.ts
• tests/codex-integration/codex-prompt-layers.test.ts
• tests/codex-integration/combos.test.ts
• tests/config/client-config-export.test.ts
• tests/lib/redact.test.ts
• tests/providers/cursor/cursor-desktop-exec.test.ts
• tests/providers/cursor/cursor-tool-definitions.test.ts
• tests/providers/xai/xai-tool-schema.test.ts
• tests/responses/responses-parser.test.ts
• tests/server/server-auth.test.ts
• tests/server/system-env.test.ts
• tests/vision/vision-cache.test.ts

---

📝 Walkthrough

Walkthrough

This PR delivers a large "split train": pure-move refactors that split 14 large source modules (image codec, Cursor tool definitions, xai schema, Claude inbound, config-export, log-guard, prompt-layers, combos, redact, openai-tiers, responses parser, system-env, vision, provider type) into smaller sibling files while preserving public exports and behavior. It also adds a WebSocket isolation fix to a test fixture and roughly 90 devlog planning/history documents.

Changes

Split-Train Planning Documentation

Layer / File(s)	Summary
Baseline manifest and history records (000-009) devlog/_plan/260905_now_split_train/000_3_main_export_baseline.json, .../000_plan.md, .../001-009*.md	Records the pinned export baseline and closed-out process history for the split train.
Core lib split plans (010-030) devlog/_plan/260905_now_split_train/010-030*.md	Plans and execution records for splitting redact.ts, errors.ts, and upstream-retry.ts.
Provider registry split plans (040-070) devlog/_plan/260905_now_split_train/040-070*.md	Plans and execution records for splitting openai-tiers.ts and registry.ts into four PR layers.
Adapter, cursor, and vision split plans (080-190) devlog/_plan/260905_now_split_train/080-190*.md	Plans for anthropic image normalize/adapter, cursor tool/catalog/images/request-builder/protobuf files, xai tool schema, ollama native, and vision index.
Images, responses, claude, server split plans (200-260) devlog/_plan/260905_now_split_train/200-260*.md	Plans for images/artifacts.ts, responses/parser.ts, namespace-tool-compat, agent-task-recovery, collaboration, claude/inbound.ts, and server/claude-messages.ts.
System-env, management, codex, combos plans (270-340) devlog/_plan/260905_now_split_train/270-340*.md	Plans for server/system-env.ts, management routes, codex/prompt-layers.ts, combos/types.ts, subagent-defaults.ts, cli-install-provenance.ts.
Routing, oauth, log-guard, config-export plans (350-450) devlog/_plan/260905_now_split_train/350-450*.md	Plans for routing/trace.ts, github-copilot.ts, log-guard files, clients/config-export.ts, plus delivery-status records.
CLI, client, lab split plans (460-570) devlog/_plan/260905_now_split_train/460-570*.md	Plans for cli/provider.ts, client/hub-client.ts, and lab module files.
GUI pages/components/scripts split plans (580-750) devlog/_plan/260905_now_split_train/580-750*.md	Plans for GUI storage/integrations/compatibility pages, combo workspace components, release-notes and test scripts.
Closeout and regression plans (800-830) devlog/_plan/260905_now_split_train/800-830*.md	Defines the two-cycle regression matrix and delivery-first procedure for publishing the aggregate PR.

Estimated code review effort: 4 (Complex) | ~75 minutes

Module Split Implementation

Layer / File(s)	Summary
Anthropic image normalize/codec split src/adapters/anthropic-image-codec.ts, src/adapters/anthropic-image-normalize.ts, tests/adapters/anthropic/anthropic-image-normalize.test.ts	Moves the tiered normalization pipeline, byte-weighted cache, and encode/validate hooks into a new codec module; the original file re-exports them.
Cursor desktop-executor and tool-definitions split src/adapters/cursor/desktop-executor-contract.ts, native-exec-desktop.ts, tool-definitions.ts, tool-naming.ts, tool-schemas.ts, tool-guidance.ts, src/types/provider.ts, tests	Extracts DesktopExecutorConfig into a dependency-free contract, then splits tool naming, schemas, and guidance logic out of tool-definitions.ts.
XAI schema analysis split src/adapters/xai-tool-schema.ts, src/adapters/xai-schema-analysis.ts, tests	Moves JSON Pointer resolution and schema disjointness checks into a new analysis module.
Claude inbound split src/claude/inbound.ts, inbound-records.ts, inbound-model-options.ts, inbound-content-options.ts, tests	Splits model resolution, content conversion, and shared error types out of inbound.ts.
Client config-export split src/clients/config-export.ts, src/clients/config-export/*.ts, tests/config/client-config-export.test.ts	Splits contracts, constants, model metadata, and per-client (omp, dsh, mcode, zcode) builders into a config-export/ subdirectory.
Codex log-guard schema split src/codex/log-guard/inspect.ts, inspect-schema.ts, test	Extracts SQLite schema-compatibility checks into a dedicated module.
Codex prompt-layers split src/codex/prompt-layers.ts, src/codex/prompt-layers/*.ts, test	Splits path resolution, revision hashing, encoding, and TOML read/edit helpers into submodules.
Combos identifiers split src/combos/types.ts, src/combos/identifiers.ts, test	Moves combo identifier constants and helper functions into a new module.
Redact folding split src/lib/redact.ts, src/lib/redact-folding.ts, test	Moves the lexical matching-fold helper and its lookup tables into a sibling file.
OpenAI tiers destination split src/providers/openai-tiers.ts, openai-tiers-destination.ts, test	Moves provider identity and destination-classification helpers into a new module.
Responses parser split src/responses/parser.ts, parser-content.ts, parser-tools.ts, parser-text-format.ts, test	Splits content parsing, tool-choice mapping, and text-format parsing out of the request parser.
Server system-env shell split src/server/system-env.ts, system-env-shell.ts, test	Moves shell-hook environment file writing and hook install/uninstall logic into a new module.
Vision planning and image rewrite split src/vision/index.ts, plan.ts, image-rewrite.ts, test	Splits vision-sidecar planning and image-stripping/rewrite logic into two new modules.
Server-auth WebSocket transport test isolation tests/server/server-auth.test.ts	Wraps globalThis.WebSocket in a Proxy to refuse native upstream dials, forcing the SSE fallback path in the HTTP fixture test.

Estimated code review effort: 4 (Complex) | ~90 minutes

Sequence Diagram(s)

No sequence diagram is generated. The changes are a large batch of file-level pure-move refactors and documentation additions rather than a new feature or multi-component interaction flow.

✨ Finishing Touches

📝 Generate docstrings

• Create stacked PR
• Commit on current branch

🧪 Generate unit tests (beta)

• Create PR with unit tests
• Commit unit tests in branch codex/closeout-split-train

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands.}

[lidge-jun]

리뷰 · 우선순위 74 / 80

이 PR은 이미 따로 리뷰된 모듈화(파일 쪼개기) PR 14개(#3557, #3559, #3566, #3567, #3570, #3574, #3577, #3580, #3583, #3585, #3590, #3594, #3599, #3611)의 내용을 한 번에 dev에 넣는 통합 PR입니다. 지금 dev HEAD는 c4701938c(#3662 Advanced 설정에 Codex 쿼터 켜기 UI 통합)이고, 이 브랜치 codex/closeout-split-train(머리 0d071d822)는 그 바로 직전 기준점 bf58ef182 위에 62커밋이 쌓인 뒤 #3662 한 커밋만 뒤처진 상태입니다. GitHub 비교로는 diverged(ahead 62 / behind 1)이고 mergeable은 맞지만, 호스트 CI가 아직 돌아가는 중이라 mergeStateStatus는 BLOCKED입니다.

하는 일은 크게 세 덩어리입니다. 첫째, 큰 파일을 손대지 않고 옆 잎(leaf)으로 옮깁니다. 예: src/lib/redact.ts → redact-folding.ts, src/providers/openai-tiers.ts → openai-tiers-destination.ts, src/adapters/cursor/tool-definitions.ts → tool-naming/tool-schemas/tool-guidance, src/vision/index.ts → plan.ts/image-rewrite.ts, src/responses/parser.ts → parser-content/parser-tools/parser-text-format, src/claude/inbound.ts → records/model/content options, src/server/system-env.ts → system-env-shell.ts, src/codex/prompt-layers.ts → encoding/paths/revision/toml 잎, src/combos/types.ts → identifiers.ts, src/codex/log-guard/inspect.ts → inspect-schema.ts, src/clients/config-export.ts → config-export/ 아래 포맷별 모듈, src/adapters/anthropic-image-normalize.ts → anthropic-image-codec.ts, src/adapters/xai-tool-schema.ts → xai-schema-analysis.ts, 그리고 Cursor desktopExecutor 타입을 desktop-executor-contract.ts로 빼서 src/types/provider.ts의 import 경로만 바꿉니다. 공개 export 이름은 옛 경로에서 다시 내보내서(re-export) 호출부가 깨지지 않게 유지하는 설계입니다.

둘째, devlog/_plan/260905_now_split_train/에 이 열차의 기록·마감·회귀 순서 문서를 한꺼번에 넣습니다. 800/830 문서 기준으로는 "이 통합 PR을 먼저 dev에 넣고, 그 다음에 main→merged-dev 회귀 PABCD를 두 바퀴"가 최신 순서입니다. 원래 68개 행을 전부 끝낸 것이 아니며, 남은 부채 층은 미룬다고 명시합니다. 셋째, 프로덕션 동작이 아니라 테스트만 고칩니다. tests/server/server-auth.test.ts의 redirectCanonicalCodexTo가 예전에는 fetch만 가로채서 네이티브 업스트림 WebSocket이 붙으면 1초 워치독에 흔들렸습니다. 지금은 chatgpt.com + /backend-api/codex 경로의 wss 생성만 Proxy로 거부하고, 루프백·다른 호스트는 그대로 두며 afterEach에서 WebSocket을 복구합니다. 인증 단언·타임아웃·프로덕션 전송은 건드리지 않는다고 적혀 있습니다.

현재 dev가 최적화하는 방향(쿼터 활성화 UI, 릴리스 버전 라인, Reserve/WS 주변)과 파일 충돌면은 작습니다. #3662는 GUI Advanced 쪽이고 이 PR은 서버/어댑터 잎 분리라 겹침이 거의 없습니다. 다만 열린 #3664(canonical upstream WS reuse)는 src/server/responses/ws-upstream*.ts 쪽 실사용 WS 풀을 다루므로, 이 PR의 픽스처 Proxy와는 파일이 다르지만 "업스트림 WS가 언제 붙는가"라는 이야기의 이웃입니다. 둘 다 넣을 때는 #3668 픽스처가 HTTP-only 단언을 지키는지, #3664 실전송 테스트가 서로 안 깨지는지만 보면 됩니다. 변경 규모는 +23326 / −3893, 파일 159개로 커 보이지만 상당수는 plan 문서이고, 코드는 이미 개별 PR에서 본 추출을 한 줄기로 모은 형태입니다.

왜 지금이냐면, 메인테이너가 "14개를 하나씩 머지하지 말고 검증된 통합본으로 넣고 원본은 landing 뒤에 superseded로 닫아라"고 이미 정해 둔 마감입니다. 개별 14개는 아직 OPEN이고, 이 PR이 실제로 dev에 들어간 뒤에 Landed via #3668 at <commit> + landed-via-maintainer로 닫는 흐름이 맞습니다. 호스트 CI(test 1–4, macos 등)가 이 정확한 머리 0d071d822에서 초록이 되기 전에는 머지하면 안 됩니다. 로컬에서 예전에 통과했다는 기록(약 19236 pass)은 참고일 뿐, 최종 머지 기준은 이 PR의 호스트 체크입니다.

tests/server/server-auth.test.ts redirectCanonicalCodexTo - 프로덕션이 아니라 픽스처만 고친 변경인데, chatgpt.com + /backend-api/codex wss만 거부한다. 경로·호스트가 조금만 달라도 예전처럼 네이티브 dial이 살아나 워치독 플레이크로 돌아갈 수 있다. 회귀 때 이 경계를 다시 확인해야 한다.
브랜치 vs 현재 dev - #3662(c4701938c) 한 커밋 뒤처진 diverged 상태다. mergeable이지만 머지 커밋 트리가 로컬에서 검증한 소스와 같다고 단정하면 안 되고, 호스트 merge-ref CI 결과를 최종으로 봐야 한다.
원본 PR 14개(#3557…#3611) - 아직 모두 OPEN이다. 이 통합본이 랜딩되기 전에 개별 머지하거나, 랜딩 후에 닫지 않으면 오픈 PR 숫자가 다시 부풀어 기여자를 헷갈리게 한다.
devlog 800/830 vs 실제 체크 - 문서에 '누락·실패·취소 체크는 PASS로 세지 말 것', '두 번의 main→dev 회귀는 머지 후'라고 적혀 있다. CI가 도는 지금 단계에서 초록을 가정한 머지 예약은 문서 스스로와 어긋난다.
src/combos/types.ts / identifiers.ts - 식별자 헬퍼만 옮기고 COMBO_DEFAULT_WAIT_FOR_COOLDOWN_MS·cooldownMs·waitForCooldownMs 검증은 types에 남아 있다. 추출은 맞아 보이지만, cooldown 쪽 동작 회귀는 통합 머지 후 combos 테스트로 한 번 더 확인하는 편이 안전하다.
#3664와의 이웃 관계 - 이 PR은 HTTP 픽스처에서 업스트림 WS를 끄고, #3664는 실전송에서 업스트림 WS를 재사용한다. 파일은 안 겹치지만 머지 순서에 따라 server-auth·ws-upstream 테스트가 서로 영향을 줄 수 있으니 둘 다 넣을 때 교차 확인이 필요하다.
68행 전체 해소 아님 - closeout이 명시하듯 남은 debt layer는 미루었다. 이 PR을 '모듈화 열차 완료'로 읽으면 안 되고, 검증된 14개 cutoff 배달로 읽어야 한다.

메인테이너의 판단이 필요한 지점

• 호스트 CI(특히 test 샤드·macos)가 이 머리에서 전부 통과한 뒤에 바로 admin merge 할지, 아니면 behind 1(feat(gui): unify quota activation in Advanced settings #3662)을 브랜치에 먼저 rebase/merge 한 뒤 다시 CI를 볼지
• 랜딩 직후 원본 14개를 일괄 superseded 종료할지(권장), 일부만 남길지
• feat(responses): bounded canonical upstream WS reuse #3664(upstream WS reuse)를 이 통합본 앞·뒤 어느쪽에 둘지. 픽스처 격리와 실전송 재사용이 같은 주에 겹친다
• 머지 후 약속된 두 바퀴 PABCD(840/850)를 누가·언제 돌릴지. 830은 머지를 회귀 완료까지 미루지 말라고 한다

너의 추천
호스트 CI가 머리 0d071d822(또는 CI를 다시 돌린 최종 머리)에서 전부 초록이면 이 PR을 dev에 머지하고, 즉시 원본 14개에 Landed via #3668 at <commit> 댓글·landed-via-maintainer 라벨을 붙인 뒤 completed/superseded로 닫아라. CI가 하나라도 빨갛거나 취소면 머지하지 말고 그 체크만 고친 뒤 같은 최종 머리로 다시 검증해라. #3662와의 1커밋 behind는 충돌면이 작아 보이니, 머지 커밋 CI로 흡수해도 되고 불안하면 짧게 rebase 한 번이 더 깔끔하다. 머지 직후 840/850 회귀 두 바퀴를 스케줄하고, #3664는 이 랜딩 뒤에 교차 테스트하며 이어가라.

이 댓글은 grok-bot이 작성했습니다

[lidge-jun]

Maintainer-directed admin landing: exact head c923cb2 passed Cross-platform CI 33971699404 and governing hygiene/target checks. Independent source/security and current-base integration reviews passed. Checked integration commit 45f1923, tree 34314ca70e706d0deb8fb6532a3675513aca8b8b, against current dev ef9c538. No unresolved review threads at the landing check. CodeRabbit remains pending and is not represented as a completed review; the superseded cancelled label run is not counted as a pass. This uses the owner-requested admin approval bypass, not a self-approval. Two full main-to-merged-dev regression cycles follow landing; earlier limitations remain recorded.