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
[fix] Translate the remaining /m English in providers, lists and tool rows - #8363
After v0.129.0, a French, German or Japanese user still saw English in a few places:
Settings > AI providers. The Test button tooltip ("Test checks the key with OpenAI..."), the "Add a model ID the API doesn't list" field, "41 models fetched", "3 of 41", "Fetched 2 min ago", "Key configured", "Defaults" and "3 of 38 active" in the rows, "This provider offers no free credential check...", and the subscription status lines ("Not signed in", "Sign in needed. The ChatGPT sign-in is no longer valid.").
Provider errors. "This connection no longer exists. Reload and try again.", "Agenta could not reach OpenAI to test this credential.", the sign-in attempt errors, and "The sign-in started without a code. Try again."
Lists in Japanese and Chinese. Harness names, project names and schedule days were joined with ", " instead of "、" ("Pi, Claude Code と Codex で有効").
Chat tool rows. Our own nouns stayed English in every language ("Sucht nach triggers", "schedule erstellt", "channel messages gelesen"). German query rows read "nach X gefunden" ("found after X"), which is wrong for every verb except "search".
Custom skins. A skin that overrides only the row text lost its text: the built-in frame filled in the file name instead. A skin verb in Japanese was moved after the object, so the row did not match the skin's English sentence.
French provider dialog. The help text under "Modèles actifs" read "les modèles Point de terminaison compatible OpenAI proposés par cette connexion".
Schedule builder. The "3h" chips and the cron placeholder "minute hour day month weekday (UTC)".
The cause is the same in most places: @agenta/entities built these sentences in English, and the UI showed them as they were.
Changes
The pattern from #8354 is kept: @agenta/entities keeps the rules and returns data or a code, and @agenta/entity-ui gives the words from the catalog.
New providerCardText.ts and subscriptionText.ts in @agenta/entity-ui/secretProvider hold the card text. The English copy helpers in entities/secret/core/cardCopy.ts and connectedRowSubtitle are removed. Their English tests moved to entity-ui/tests/unit/providerCardText.test.ts, byte for byte.
Signatures in @agenta/entities/secret that changed:
probeFailureMessage(error, {gone, unreachable}). A 4xx still shows the server's own message, as before.
doneState(...).note is now "savedUntested" | "testFirst" | null instead of an English sentence.
subscriptionStatusLine, subscriptionAttemptErrorSentence and SUBSCRIPTION_SIGN_IN_HINT are replaced by subscriptionStatus() and subscriptionAttemptFailure(), which return a code.
Relative fetch time uses Intl.RelativeTimeFormat in the active language ("il y a 2 min", "2 分前"). English keeps "2 min ago".
New formatList() in @agenta/i18n (Intl.ListFormat, no new dependency). English output does not change ("a, b, c"; long form "a, b and c"). ScheduleText gets enumerate() for the capped day lists and cron day sets.
Chat rows: 11 platform nouns join the glossary (trace, schedule(s), tool(s), triggers, connections, deliveries, channel destinations/messages). A search verb uses the new forSearchQuery key. German now reads "nach X gesucht" and "für X gefunden".
Skin overrides: a registered sentence without a verb drops the built-in verb and frame. A registered verb gets a verb-first frame in every language.
Catalogs: 37 new keys (plural families count once), each in all 7 languages, written by hand (Chinese terms follow the glossary). enabledInHarnesses is removed; the harness list is now one formatList value. Changed values: de forQuery ("für {{query}}"), fr enterCredential, zh fetched ("{{time}}获取"). .meta hashes are recorded for every new key.
Display only: the saved default names ("OpenAI-compatible endpoint", "OpenAI (Organization)") stay English, because they are written to the API.
Tests
pnpm lint-fix in web: pass, with no new warnings. prettier --check on the changed files: pass.
tsc --noEmit: pass for @agenta/i18n, entities, entity-ui, settings-ui, chat, automation-ui and web/mobile.
node scripts/check.ts (i18n): pass.
Unit tests (vitest run): @agenta/i18n 66/66, entities 3034/3034, entity-ui 1771/1771, settings-ui 330/330, chat 1481/1481. automation-ui has no unit tests.
New tests pin the English text byte for byte. They also check Japanese and French text, plural forms, Intl relative time, and the "、" separator (providerCardText.test.ts, scheduleText.test.ts, translate.test.ts). German nouns and query rows and the skin override frames are checked in toolActivityLanguage.test.ts.
Demo is outstanding. I did not run the app for this PR, so there are no screenshots yet.
What to QA
Set the language to Français. Open Settings > AI providers > OpenAI, type a key and hover Test. The tooltip is French. Click Test. The line ends with "N modèles récupérés", the header shows "3 sur 41", and the footer shows "Récupéré il y a ... ·".
In French, open the OpenAI-compatible endpoint card with no credential. The help text ends with "...proposés par cette connexion (Point de terminaison compatible OpenAI)."
In 日本語, enable two harnesses on a connection. The summary reads "Pi、Claude Code で有効". A weekly schedule with five days shows "月、火、水 +2".
In Deutsch, run an agent that calls discover_triggers and list_schedules. The rows read "Sucht nach Auslösern" and "Zeitpläne geprüft".
Regression: in English, every text above reads exactly as before.
Follow-ups (not in this PR)
The API returns the probe verdicts in English ("OpenAI accepted this key.", "... rejected this key (401)."). It should return a code that the UI can translate.
Cron validation errors in entities/gatewayTrigger/core/cron.ts ("Invalid minute: bad range ...") are English.
Other ", " joins outside these screens: settings-ui/shared/accountOwners.ts ("Used by"), the observability analytics filter chips, QueueEvaluatorField, AgentConfigCard skill names, and commitDiff/summaryMessage.ts.
The approval card sentences in agenta-chat/src/model/approvalDescribers are still English.
Three existing providerConnection keys (anyHarness, noHarnessSelected, enabledInHarness) have no .meta hash in the six languages. translate.ts will adopt them on its next run.
… leftovers
Provider card text, subscription status lines, probe errors, list separators,
German tool rows and skin override frames now follow the active language.
@agenta/entities returns codes or takes the words; @agenta/entity-ui gives them
from the catalog. English text is unchanged.
Provider settings now show localized credential guidance, connection status, model counts, and subscription sign-in outcomes.
Schedule controls include localized interval and cron-field guidance, with list formatting adapted to each language.
Chat activity labels now cover tools, traces, schedules, triggers, connections, deliveries, and channel messages.
Improvements
Search activity wording and registered tool displays now use more context-appropriate phrasing.
Provider and schedule lists display with locale-appropriate separators across supported languages.
The PR updates chat activity wording, schedule formatting, provider connection text, and subscription sign-in text. It adds locale-aware list formatting and translations across English, German, Spanish, French, Japanese, Brazilian Portuguese, and Chinese.
Changes: Chat activity wording
Layer / File(s)
Summary
Registered tool wording and localized terms web/packages/agenta-chat/src/skin/registry.ts, web/packages/agenta-chat/tests/unit/skin/*, web/packages/agenta-i18n/locales/*/chat.toolActivity.json, web/packages/agenta-i18n/locales/.meta/*
Registered tool frames and query wording now use updated localized terms. The platform glossary and chat activity tests include terms and phrasing for tools, triggers, traces, connections, schedules, deliveries, and channel content.
Schedule weekday and capped-item lists use locale-aware enumeration. Schedule controls use translated hour-chip and cron-placeholder text. Japanese tests cover localized separators.
Provider connection helpers now accept localized text or return status tokens instead of embedding display sentences. Removed card-copy and subtitle helpers are replaced by UI-localized helpers. Tests now pass display text and assert status tokens.
Localized provider text and UI wiring web/packages/agenta-entity-ui/src/secretProvider/*, web/packages/agenta-settings-ui/src/providers/*, web/packages/agenta-i18n/src/index.ts, web/packages/agenta-i18n/tests/unit/translate.test.ts, web/packages/agenta-entity-ui/tests/unit/providerCardText.test.ts, web/packages/agenta-i18n/locales/*/entityUi.providerConnection.json, web/packages/agenta-i18n/locales/*/settings.providers.json, web/packages/agenta-i18n/locales/.meta/*
Provider card text is localized in entity UI. Provider summaries and settings rows use localized copy, and formatList formats harness and project names. Tests cover provider text in English, Japanese, and French.
Subscription status classification web/packages/agenta-entities/src/secret/core/subscriptionConnections.ts, web/packages/agenta-entities/tests/unit/subscription-connections.test.ts
Subscription status and attempt outcomes are represented by semantic values. Tests cover mapped and fallback classifications.
Entity UI maps subscription classifications to translated messages. The sign-in flow uses a translated missing-code error and refreshes status text when the text version changes.
Agenta-AI/agenta#8347: It updates the same chat activity catalogs, tool-row wording, registry, and related tests.
Merge Risk: Low · up to 44f32
This PR mostly localizes provider, schedule and tool text. A few minor display issues remain: some German wording is wrong, narrow screens can clip hour chips, and some text keeps the previous language after a switch. None of these should block merge, but fix them soon.
Pre-merge checks: 4
Check name
Status
Explanation
Title check
The title names the main change: translating the remaining English text in provider settings, lists, and chat tool rows. The changeset covers each of these areas: provider-card and subscription copy m…
Description check
The description explains the problem, the changes by file area, the tests run, QA steps, and follow-ups. It matches the changeset, including provider text, subscription helpers, formatList, schedule t…
Linked Issues check
Check skipped because no linked issues were found for this pull request.
Out of Scope Changes check
Check skipped because no linked issues were found for this pull request.
Full details: Title check
Explanation
The title names the main change: translating the remaining English text in provider settings, lists, and chat tool rows. The changeset covers each of these areas: provider-card and subscription copy moved to entity-ui with localized text, formatList and Intl-based list joins, and chat tool-row glossary and verb frames. The title is specific and concise, and it fits the PR objectives.
Full details: Description check
Explanation
The description explains the problem, the changes by file area, the tests run, QA steps, and follow-ups. It matches the changeset, including provider text, subscription helpers, formatList, schedule text, chat tool rows, skin overrides, and the locale catalogs.
Finishing Touches
Generate unit tests (beta)
Commit to this branch
Create a new PR
Autofix · Keep fixing CodeRabbit findings and required CI, and resolving merge conflicts
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.
Included review availability: This review used your included allowance. Your plan provides up to 8 included reviews per hour; 7 remain after this review.
The reason will be displayed to describe this comment to others. Learn more.
Minor · Functional Correctness · Quick win
Rebuild the picker rows when the display language changes.
The rows memo in ModelPickerControl.tsx does not depend on the active language. The disabled subscription row stores its hint as a translated string at build time. After a language change, the row can keep the previous-language hint until one of the memo inputs changes. The same memo also produces other translated row text, so the fix should cover the row build as a whole, not only this hint.
Add the active locale, or a text-version value derived from it, to the rows dependency list. Alternatively, resolve the hint at render time instead of storing it in the row.
The reason will be displayed to describe this comment to others. Learn more.
Minor · Functional Correctness · Quick win
Let the hourly interval chips wrap when the popover is narrow.
The hourly "every" row sets its chips to shrink-0 and whitespace-nowrap, and its container does not wrap. At narrow widths, the seven chips can need more width than the popover provides. The German {{hours}} Std. labels make this more likely. Chips past the popover edge can be unreachable. Allowing the row to wrap keeps every interval reachable.
The reason will be displayed to describe this comment to others. Learn more.
Minor · Functional Correctness · Quick win
Translate retained probe failures during render.
If a network probe fails in Japanese, this stores the Japanese fallback in probeFailure. Switching to French leaves that sentence unchanged because the language update does not reset this state. statusLine continues to use the stored sentence.
Retain a failure descriptor and translate it during render. Preserve provider-supplied messages and the existing retry/reset behavior.
The reason will be displayed to describe this comment to others. Learn more.
Minor · Functional Correctness · Quick win
Use the "nach" search frame for query-led German find rows.
When find_traces has a query and no app name, the row uses forQuery. The German value is now "für {{query}}", so the running label reads "Sucht für timeout". German search text uses "nach" before the searched object, and the previous value gave "Sucht nach timeout". Route query-led find rows through the German search frame, or otherwise keep "nach" before the query, so the label reads "Sucht nach timeout".
The reason will be displayed to describe this comment to others. Learn more.
Minor · Functional Correctness · Quick win
Use a case-correct object for search tools.
The glossary key termTools is fixed as “Werkzeugen”, the dative plural. search_tools with no query reaches the direct-object frame {{verb}} {{object}}. It renders “Sucht Werkzeugen” and “Werkzeugen gesucht”, but the direct-object form is “Werkzeuge”.
The same key also feeds the discover frame, which needs the dative after “nach”. Changing termTools to “Werkzeuge” alone would fix search tools and break discover tools (“Sucht nach Werkzeuge”). Provide an accusative form for the direct-object frames, and keep the dative form for the “nach” frames.
Dependent tests finished. Post /preview for manual review. Environment absence verified; retained storage requires separate provider accounting.
Post /preview as a new comment to request a one-hour preview. Repeating it while active does not extend the hour.
Preview data is disposable. Expiry is enforced by the next successful cleanup sweep, not an exact timer.
Environment deletion stops its compute. Provider-retained storage and permanent services may still be billed.
This branch has not been deployed
No deployments
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
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.
Context
After v0.129.0, a French, German or Japanese user still saw English in a few places:
The cause is the same in most places:
@agenta/entitiesbuilt these sentences in English, and the UI showed them as they were.Changes
The pattern from #8354 is kept:
@agenta/entitieskeeps the rules and returns data or a code, and@agenta/entity-uigives the words from the catalog.providerCardText.tsandsubscriptionText.tsin@agenta/entity-ui/secretProviderhold the card text. The English copy helpers inentities/secret/core/cardCopy.tsandconnectedRowSubtitleare removed. Their English tests moved toentity-ui/tests/unit/providerCardText.test.ts, byte for byte.@agenta/entities/secretthat changed:credentialSummary(connection, keyConfigured)activeModelsSummary(connection, capabilities, {defaults, active})probeFailureMessage(error, {gone, unreachable}). A 4xx still shows the server's own message, as before.doneState(...).noteis now"savedUntested" | "testFirst" | nullinstead of an English sentence.subscriptionStatusLine,subscriptionAttemptErrorSentenceandSUBSCRIPTION_SIGN_IN_HINTare replaced bysubscriptionStatus()andsubscriptionAttemptFailure(), which return a code.Intl.RelativeTimeFormatin the active language ("il y a 2 min", "2 分前"). English keeps "2 min ago".formatList()in@agenta/i18n(Intl.ListFormat, no new dependency). English output does not change ("a, b, c"; long form "a, b and c").ScheduleTextgetsenumerate()for the capped day lists and cron day sets.forSearchQuerykey. German now reads "nach X gesucht" and "für X gefunden".enabledInHarnessesis removed; the harness list is now oneformatListvalue. Changed values: deforQuery("für {{query}}"), frenterCredential, zhfetched("{{time}}获取")..metahashes are recorded for every new key.Tests
pnpm lint-fixinweb: pass, with no new warnings.prettier --checkon the changed files: pass.tsc --noEmit: pass for@agenta/i18n,entities,entity-ui,settings-ui,chat,automation-uiandweb/mobile.node scripts/check.ts(i18n): pass.vitest run):@agenta/i18n66/66,entities3034/3034,entity-ui1771/1771,settings-ui330/330,chat1481/1481.automation-uihas no unit tests.Intlrelative time, and the "、" separator (providerCardText.test.ts,scheduleText.test.ts,translate.test.ts). German nouns and query rows and the skin override frames are checked intoolActivityLanguage.test.ts.What to QA
discover_triggersandlist_schedules. The rows read "Sucht nach Auslösern" and "Zeitpläne geprüft".Follow-ups (not in this PR)
entities/gatewayTrigger/core/cron.ts("Invalid minute: bad range ...") are English.settings-ui/shared/accountOwners.ts("Used by"), the observability analytics filter chips,QueueEvaluatorField,AgentConfigCardskill names, andcommitDiff/summaryMessage.ts.agenta-chat/src/model/approvalDescribersare still English.providerConnectionkeys (anyHarness,noHarnessSelected,enabledInHarness) have no.metahash in the six languages.translate.tswill adopt them on its next run.