You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
Other locales, phase 2: ja to full coverage (the remaining 79 pages, seven serial PRs) #388
Phase 2 of the locale plan the maintainer approved on 2026-10-29 (plan recorded on #384). Phase 1 (#386, landed as #387) leaves ja at 0 stale and 79 missing on maina3e5657. This card brings ja to full coverage, so that ja is a fully covered, auto-maintained locale.
Scope: 79 pages, about 14,800 English lines, in seven serial PRs
Each PR stays under 3,000 lines and contains locale files only. The PRs land one at a time, in this order. The next PR starts only after the previous one has merged and the main CI and smoke check are green.
PR
Pages
English lines
1
build/ top level (index, agents, ai-builder, ai-skills, extend-existing-systems, marketplace, packages, reports, templates) plus build/data/* (4)
~2,480
2
build/automation/* (4) plus build/interface/* (7)
~2,560
3
configure/ top level (14 pages)
~2,470
4
configure/permissions/* (7) plus cloud/* (8) and a new cloud/meta.ja.json
~2,160
5
deploy/* (4) plus operate/* (7)
~1,940
6
reference/* (9) plus root architecture and ontology
~1,740
7
resources/* (4: changelog, faq, glossary, support). This PR carries Fixes for this card.
~1,430
Every other section already has its meta.ja.json. Check that each one's pages equals the English, and fix it in the PR for that section if it does not.
Rules
These are the same as phase 1 (#386) and the zh-Hans pass (#347), under docs/TRANSLATION.md:
Product nouns stay English.
Use the glossary's ja column, including action = アクション and operation = 操作.
UI labels stay in English, as in phase 1.
Every heading keeps its English id.
Fences match the English byte for byte.
Stamp each page with check-translations.mjs --stamp.
Locale files only.
Routine
As each PR lands, the seat's translation routine (#379) watches its pages for ja as well as zh-Hans. Once PR 7 lands, every English change dispatches a ja re-sync next to the zh-Hans one. Then a two-week watch starts: ja must stay at 0 stale before phase 3 adds the next locale.
Acceptance
After each PR, check-translations shows ja at 0 stale, and missing falls by that PR's page count. After PR 7, ja shows 0 missing.
check-translation-output --files passes on each PR.
A rendered check of one page per section in the PR, under /ja/docs/..., shows ja content, the ja sidebar and no page errors.
Claim: locale plan phase 2 · 2026-10-29T10:37Z
Session: session_01QsR2jTaZ2GeyfjANMwstiw (seat thread; the dev is card thread session_01QsR2jRsjbApbf83iyTDDXm, reused from phase 1, identity by branch)
Account: zhuangjianguo
Branch: claude/project-thread-u7eoxi (platform-assigned; restarted from main for each PR, because #387 merged)
Worktree: the card thread's own docs.objectos.ai checkout
Domain: domain:docs (single-lane repository)
Seat: repo:docs.objectos.ai#1
File surface: 79 new *.ja.mdx files across build/, configure/, cloud/, deploy/, operate/, reference/, resources/, and root architecture and ontology. Also a new cloud/meta.ja.json, and fixes to existing meta.ja.jsonpages arrays where they differ from the English. Locale files only. If the work goes beyond this surface, stop and explain why in the report.
Container & model: S, mode:thread, model: default tier (TIER_DEFAULT — sister repository, hand-derived: no packages/spec, no protocol-semantic surface), effort: project default
Clause-②: no
Metadata-protocol: none
Responsibility: n/a — phase 2 of the maintainer-approved locale plan (2026-10-29T09:44Z)
Thread-read: none
Premise: confirmed. On maina3e5657, after phase 1 (#387), ja shows 0 stale and 79 missing.
Tier: none.
Serial constraints cleared: the lane is serial, and this is the only card in flight. Its seven PRs land one at a time, in the card's order. No open PR touches these files; only dependabot PRs are open.
The seat's ACCEPTs on #387, #389, #390, #396 and #398 each quoted check-translation-output --files: passed, 0 findings from a seat re-run, and those re-runs checked nothing. --files takes a list file with one path per line. The seat passed the paths themselves, so the script read the first .mdx as the list and blocked on 0 files. CI's own run of the gate was unaffected.
The seat re-ran it correctly on #398's head 8688a48, with list files covering every translated page in the tree:
ja. It blocked on all 57 ja pages: No findings.
zh-Hans. It blocked on all 98 zh-Hans pages: No findings.
The earlier conclusions therefore hold. The seat's review script now writes the list file and quotes the blocking on N line.
Phase 2 is done: ja is fully covered. #402, PR 7 of 7, merged as 384e74b and closed this card. Main CI run 40890084196 is green: Publish success, the new version is serving, and the smoke check passed.
On fresh main, check-translations shows ja at 0 stale, 0 missing, and zh-Hans at 0/0.
The first landing pushed /api/search past the Worker memory limit, and the deploy rolled back on its own. #392 reverted it, #394 fixed search (#393), and #396 re-landed the pages byte for byte.
With full ja, the Worker upload is about 43,950 KiB, with about 17,490 KiB of headroom under the budget, and one more full locale fits by the CI projection. Search heap stays under its 100 MB gate.
From now on, the seat's translation routine (#379) dispatches a ja re-sync next to the zh-Hans one whenever an English page changes. The two-week watch at ja 0 stale runs until 2026-11-12. Phase 3, the next locale, starts after that.
Activity
objectstack-fleet commentedon Oct 29, 2026
Claim: locale plan phase 2 · 2026-10-29T10:37Z
Session:
session_01QsR2jTaZ2GeyfjANMwstiw(seat thread; the dev is card threadsession_01QsR2jRsjbApbf83iyTDDXm, reused from phase 1, identity by branch)Account:
zhuangjianguoBranch:
claude/project-thread-u7eoxi(platform-assigned; restarted frommainfor each PR, because #387 merged)Worktree: the card thread's own
docs.objectos.aicheckoutDomain:
domain:docs(single-lane repository)Seat:
repo:docs.objectos.ai#1File surface: 79 new
*.ja.mdxfiles acrossbuild/,configure/,cloud/,deploy/,operate/,reference/,resources/, and rootarchitectureandontology. Also a newcloud/meta.ja.json, and fixes to existingmeta.ja.jsonpagesarrays where they differ from the English. Locale files only. If the work goes beyond this surface, stop and explain why in the report.Container & model: S,
mode:thread,model: default tier (TIER_DEFAULT — sister repository, hand-derived: no packages/spec, no protocol-semantic surface),effort: project defaultClause-②: no
Metadata-protocol: none
Responsibility: n/a — phase 2 of the maintainer-approved locale plan (2026-10-29T09:44Z)
Thread-read: none
Premise: confirmed. On
maina3e5657, after phase 1 (#387), ja shows 0 stale and 79 missing.Tier: none.
Serial constraints cleared: the lane is serial, and this is the only card in flight. Its seven PRs land one at a time, in the card's order. No open PR touches these files; only dependabot PRs are open.
docs(i18n): translate build/ and build/data/ into ja (#388, PR 1 of 7)
docs(i18n): translate build/automation/ and build/interface/ into ja (#…
docs(i18n): re-land build/automation/ and build/interface/ ja pages (#…
fix(worker): stop bundling every docs page twice (#395)
docs(i18n): translate configure/ into ja (#388, PR 3 of 7)
objectstack-fleet commentedon Oct 29, 2026
Seat correction · 2026-10-29T13:35Z ·
session_01QsR2jTaZ2GeyfjANMwstiw· seatrepo:docs.objectos.ai#1The seat's ACCEPTs on #387, #389, #390, #396 and #398 each quoted
check-translation-output --files: passed, 0 findingsfrom a seat re-run, and those re-runs checked nothing.--filestakes a list file with one path per line. The seat passed the paths themselves, so the script read the first.mdxas the list and blocked on 0 files. CI's own run of the gate was unaffected.The seat re-ran it correctly on #398's head
8688a48, with list files covering every translated page in the tree:The earlier conclusions therefore hold. The seat's review script now writes the list file and quotes the
blocking on Nline.docs(i18n): translate configure/permissions/ and cloud/ into ja (#388,…
docs(i18n): translate deploy/ and operate/ into ja (#388, PR 5 of 7)
docs(i18n): translate reference/, architecture and ontology into ja (#…
docs(i18n): translate resources/ into ja, completing ja coverage (#388,…
objectstack-fleet commentedon Oct 29, 2026
Landing · 2026-10-29T14:24Z ·
session_01QsR2jTaZ2GeyfjANMwstiw· seatrepo:docs.objectos.ai#1Phase 2 is done: ja is fully covered. #402, PR 7 of 7, merged as
384e74band closed this card. Main CI run 40890084196 is green: Publish success, the new version is serving, and the smoke check passed.On fresh
main,check-translationsshows ja at 0 stale, 0 missing, and zh-Hans at 0/0.build/top level,build/data/*(13)build/automation/*,build/interface/*(11)/api/searchpast the Worker memory limit, and the deploy rolled back on its own. #392 reverted it, #394 fixed search (#393), and #396 re-landed the pages byte for byte.configure/top level (14)configure/permissions/*,cloud/*(15), pluscloud/meta.ja.jsondeploy/*,operate/*(11)reference/*,architecture,ontology(11)resources/*(4)With full ja, the Worker upload is about 43,950 KiB, with about 17,490 KiB of headroom under the budget, and one more full locale fits by the CI projection. Search heap stays under its 100 MB gate.
From now on, the seat's translation routine (#379) dispatches a ja re-sync next to the zh-Hans one whenever an English page changes. The two-week watch at ja 0 stale runs until 2026-11-12. Phase 3, the next locale, starts after that.