[UPDATE][router][1.0.87] Router 2.7.59 fixes FlowStudio workflow synchronization errors in the console.

router/Chart.yaml

@@ -1,6 +1,6 @@
 apiVersion: v2
-appVersion: 2.7.58
+appVersion: 2.7.59
 description: Olares AI gateway with OpenAI-compatible model and media routing.
 name: router
 type: application
-version: 1.0.86
+version: 1.0.87

router/OlaresManifest.yaml

@@ -9,7 +9,7 @@ metadata:
   appid: router
   description: The native AI gateway for Olares.
   icon: https://app.cdn.olares.com/appstore/router/icon2.png
-  version: "1.0.86"
+  version: "1.0.87"
   title: Router
   categories:
     - Utilities_v112
@@ -22,183 +22,13 @@ permission:
   appCache: true
 spec:
   onlyAdmin: true
-  versionName: "2.7.58"
+  versionName: "2.7.59"
   upgradeDescription: |
-    Router 2.7.53 unifies image generation and editing across FlowStudio and
-    cloud providers. Applications can discover each model's declared media
-    operations and parameters, submit canonical image edits, and receive a
-    clear rejection before dispatch when editing is unsupported. Media audit
-    records now retain the requested operation on failures, and idempotency is
-    preserved through the Olares application gateway so a retried request does
-    not create a duplicate paid generation.
+    Router 2.7.59 fixes FlowStudio workflow synchronization errors in the console.
+    Workflow sync now accepts newer workflow types and audio and 3D outputs.
+    FlowStudio workflows no longer show a model details button that could not open; the connect button remains available.
 
-    Router 2.7.52 accepts TypeSafe System One noul answers that include optional
-    confidence and auxiliary probability distributions. Models such as Lev can
-    now preserve their calibrated response details while Router continues to
-    validate probability ranges and normalization before forwarding results.
-
-    Router adds System One as a first-class model mode under Tools. It exposes
-    a typed /v1/systemone endpoint for choice, score, and noul decisions,
-    validates each model's declared context and option limits, and keeps
-    default routing restricted to models that explicitly opt in. The console
-    includes typed Quick Start examples and makes Laya's explicit-only status
-    visible instead of silently using it as a fallback.
-
-    Router 2.7.49 tells a caller where a generated file already is. When the
-    model application stores its output on your own Olares drive, the result
-    now carries that address, so an app showing you the picture reads it from
-    your disk instead of asking Router for a second copy of it. A generated
-    clip also reports its duration again, whichever way the application spells
-    the field. Model lists gained two things a FlowStudio workflow needs: the
-    name you gave it, in place of the identifier it is filed under, and the
-    settings that one workflow accepts, so an app can offer you those and not
-    the ones it would be turned away for.
-    Router 2.7.48 shows each model's engine concurrency in the capability
-    lists. The value is the number of requests that engine can execute at
-    once, not a quota or a live request count; models whose width cannot be
-    established show a dash instead of zero or unlimited.
-    Router 2.7.47 attaches a provider's models when you add it, so the Models
-    tab opens as a list to review rather than an empty form to fill in. The
-    vendors that publish more models than anyone should inherit unseen still
-    start empty and wait for you to choose: OpenAI, OpenRouter, xAI, Fireworks
-    AI, Anthropic, Moonshot, DeepSeek and Black Forest Labs.
-    Router 2.7.45 offers one Olares entry in the Usage provider filter instead
-    of one per installed model application, which all carry the same name and
-    could only be told apart by trying them; FlowStudio is still listed on its
-    own. A call refused because its model is still loading now says which
-    model that was, and a translation that failed before any model answered is
-    filed under default-translate rather than showing no model at all.
-    Router 2.7.44 correctly applies the Windows sidebar scrollbar fix to the
-    Quasar drawer content element itself. The drawer now clips its one-pixel
-    horizontal overflow while the navigation keeps vertical scrolling.
-    Router 2.7.43 removes the horizontal scrollbar that could remain at the
-    bottom of the console sidebar on Windows. The navigation still scrolls
-    vertically when the window is too short to show every item.
-    Router 2.7.42 counts more of your calls toward the generation speed in
-    Usage. A chat that Router had timed from start to finish itself could be
-    mistaken for one of the long-running jobs that are submitted first and
-    collected later, which left it out of that figure and reported its
-    duration a few milliseconds long. The speed shown was right; it was
-    averaged over fewer calls than it should have been.
-    Router 2.7.41 reports generation speed in Usage instead of the speed of a
-    whole request. The figure divided the tokens a model wrote by how long the
-    request took, and that time also covered reading the prompt and waiting in
-    the engine's queue, so a long prompt with a short answer read as a slow
-    model. It now counts only the time spent writing, across successful chat
-    and response calls, and shows a dash where that time cannot be established
-    rather than filling it in with the wall clock.
-    Router 2.7.40 shows an installed application's current display title in
-    Usage, while falling back to the stable manifest name stored with usage
-    after uninstall or when the application directory is unavailable.
-    Router 2.7.39 keeps an application's stable identity with every usage
-    record, so calls remain attributable after the application is uninstalled
-    or Router restarts. Caller summaries now honor exact hourly ranges,
-    application-owner summaries no longer fail while decoding owners, and the
-    initial Application CRD snapshot is applied before requests are served.
-    Router 2.7.37 lets trusted Olares applications fall back to their app
-    identity when an SDK-required Bearer credential is rejected. Valid Router
-    API keys still take precedence and keep user attribution, allowlists, and
-    quotas; internal authentication failures still surface instead of being
-    hidden by the fallback.
-    Router 2.7.36 names the model that answered in your usage list. A call
-    that does not name a model asks for a category instead, and the record
-    kept that category, so a list of work done by several engines read as
-    a column of "default-tts-clone" and "default-stt". Each row now leads
-    with the model that actually ran, and keeps the category underneath it
-    as what was asked for.
-    Router 2.7.35 finishes the usage record for a voice you keep. Audio is
-    charged by the second, and a stored voice has no seconds to count, so
-    the record for making one stayed at "in progress" until it was written
-    off hours later as abandoned. Making a voice now completes as soon as
-    the model is done with it, and the record says the work was not
-    measured instead of saying nothing at all.
-    Router 2.7.34 labels a usage record with the kind of work its model
-    does. When a call names a model that cannot do what the endpoint asks
-    for, the record was filed under the endpoint's kind instead of the
-    model's, so an OCR model called on a speech endpoint showed up as
-    Audio. The one record that explains a misrouted call was the one you
-    could not find by filtering for the kind of work you were looking for.
-    Router 2.7.33 names a usage record's model the way you addressed it, so
-    one call is one model in your reports instead of two. Audio, speech,
-    OCR and translation recorded a shortened name when the call succeeded
-    and the full name when it failed. A streaming audio session is also
-    billed now: the duration the engine reports as the stream closes is
-    what you are charged for, so an hour of speech costs the same over a
-    streaming connection as it does over a plain request.
-    Router 2.7.31 charges a collected audio result once, bills a succeeded
-    audio task whose result nobody collects, counts every song a music
-    generation delivered rather than one of them, and answers a failed task
-    lookup as retryable instead of reporting the task as missing.
-    Router 2.7.30 counts the images an asynchronous workflow delivered, so a
-    priced image workflow is billed instead of recording every call at zero
-    cost, which was indistinguishable from a free one.
-    Router 2.7.29 preserves canonical audio task routing through polling and
-    keeps opaque Router task IDs stable in task status and task board results.
-    Router 2.7.28 surfaces a shared KV pool as kv_pool_tokens and passes
-    kv_budget_exhausted through with Retry-After instead of retrying the
-    same backend. Two requests that no longer fit the pool fail as
-    capacity, not as an upstream 500.
-    Router 2.7.27 preserves optional hidden lyric conditioning across Draft,
-    Format, polling, and restart-safe media snapshots.
-    Router 2.7.26 adds a versioned audio operation catalog sourced from
-    Model Console, validates authoritative HTTP and WebSocket operations,
-    and keeps asynchronous audio task routing and ownership across restarts.
-    Existing models without an operation spec remain compatible.
-    Router 2.7.25 adds a durable lyrics-alignment endpoint for completed
-    music generations. Music applications can now fetch a model-provided
-    sentence timeline while preserving the original generation binding.
-    Router 2.7.24 closes the model card as soon as it has asked for a
-    restart. Restarting an engine, and saving engine flags you changed,
-    are both a request rather than a result — the engine takes seconds to
-    come back — so the card no longer holds you in front of a panel that
-    answers long after you have stopped reading it. The model's own row
-    carries the restart through to "model ready", and you are only
-    interrupted for the endings that need something from you.
-    Router 2.7.23 lets a model row notice its engine came back. A restart
-    is asked for from a dialog, and closing that dialog to watch the list
-    behind it used to take away the only thing still checking, so the row
-    stayed on "restarting the engine" until the console gave up waiting —
-    a minute and a half after an engine that was already serving again.
-    The console now keeps checking wherever you are in it.
-    Router 2.7.22 stops a model row from contradicting itself while its
-    engine restarts. The badge and the line under it were worked out
-    separately, so for the first seconds of a relaunch the row showed a
-    green "callable" over "restarting the engine" — and once the app
-    caught up, a greyed badge over a line still saying "restarting". Both
-    now come from one reading of the row. A load that fails after a
-    restart also shows as failed straight away instead of sitting on
-    "restarting the engine" until the console gives up waiting.
-    Router 2.7.21 offers image embedding models the same try-it-out panel
-    the text ones have, so a model that embeds pictures can be given one
-    without leaving the console.
-    Router 2.7.20 shows a restart the moment you ask for one. The status
-    on a model row used to keep reading "ready" for around ten seconds
-    after the button was pressed — by which point a quick relaunch had
-    already finished — because every hop between the click and that row
-    is a poll. The console now says so from the click, on the model list
-    as well as in the dialog the button lives in, and the list keeps
-    refreshing through the relaunch instead of going quiet the moment
-    every row looks settled.
-    Router 2.7.19 keeps watching a model closely for as long as a restart
-    it asked for can take, so the loading state is actually on screen
-    while the engine is away rather than starting and ending between two
-    checks.
-    Router 2.7.18 reports what actually became of an engine restart. The
-    console used to warn about a restart that had worked, because it read
-    a field recording whether the application had ever been seen relaunching
-    an engine rather than whether this relaunch succeeded. Restarting an
-    engine, or saving engine arguments that require one, now moves the model
-    through a visible loading state and ends on the application's own verdict.
-    Router 2.7.17 extends the ElevenLabs-compatible speech surface, and
-    2.7.16 fixes a horizontal overflow in the console sidebar.
-    Router 2.7.15 proxies durable asynchronous song-draft tasks. A music
-    model's own language model can now write a caption, lyrics and the
-    tempo, key and time signature it intends to render, so an application
-    no longer needs a separate chat model to reach a first draft.
-    Router 2.7.14 adds durable asynchronous music-format tasks used by
-    Music Studio to preview ACE-Step Caption and Lyrics optimization before
-    starting expensive audio generation. It also fixes dispatch and database
-    support for declared music repaint and format operations.
+    Related changes: [#134](https://github.com/beclab/router/pull/134), [#136](https://github.com/beclab/router/pull/136).
   fullDescription: |
     Router gives Olares a single OpenAI-style API for local and cloud AI models. Use one base URL and API key to access all your models—no extra SDKs or keys needed. Admins can easily connect providers and manage models in one place, with simple setup and control from Olares Market.
 

router/i18n/de-DE/OlaresManifest.yaml

@@ -2,6 +2,12 @@ metadata:
   title: Router
   description: Das native KI-Gateway für Olares.
 spec:
+  upgradeDescription: |
+    Router 2.7.59 behebt Fehler bei der Synchronisierung von FlowStudio-Workflows in der Konsole.
+    Die Synchronisierung akzeptiert jetzt neuere Workflow-Typen sowie Audio- und 3D-Ausgaben.
+    Bei FlowStudio-Workflows wird die Schaltfläche für Modelldetails, die sich nicht öffnen ließen, nicht mehr angezeigt. Die Verbindungsschaltfläche bleibt verfügbar.
+
+    Zugehörige Änderungen: [#134](https://github.com/beclab/router/pull/134), [#136](https://github.com/beclab/router/pull/136).
   fullDescription: |
     Router stellt Olares eine einheitliche OpenAI-kompatible API für lokale und Cloud-KI-Modelle bereit. Mit einer Base-URL und einem API-Key greifen Sie auf alle Modelle zu – ohne zusätzliche SDKs oder Schlüssel. Administratoren können Anbieter verbinden und Modelle an einem Ort verwalten, mit einfacher Einrichtung und Steuerung über den Olares-Markt.
 

router/i18n/en-US/OlaresManifest.yaml

@@ -2,6 +2,12 @@ metadata:
   title: Router
   description: The native AI gateway for Olares.
 spec:
+  upgradeDescription: |
+    Router 2.7.59 fixes FlowStudio workflow synchronization errors in the console.
+    Workflow sync now accepts newer workflow types and audio and 3D outputs.
+    FlowStudio workflows no longer show a model details button that could not open; the connect button remains available.
+
+    Related changes: [#134](https://github.com/beclab/router/pull/134), [#136](https://github.com/beclab/router/pull/136).
   fullDescription: |
     Router gives Olares a single OpenAI-style API for local and cloud AI models. Use one base URL and API key to access all your models—no extra SDKs or keys needed. Admins can easily connect providers and manage models in one place, with simple setup and control from Olares Market.
 

router/i18n/es-ES/OlaresManifest.yaml

@@ -2,6 +2,12 @@ metadata:
   title: Router
   description: La pasarela de IA nativa de Olares.
 spec:
+  upgradeDescription: |
+    Router 2.7.59 corrige errores al sincronizar flujos de trabajo de FlowStudio en la consola.
+    La sincronización ahora admite nuevos tipos de flujos de trabajo y salidas de audio y 3D.
+    Los flujos de trabajo de FlowStudio ya no muestran el botón de detalles del modelo que no se podía abrir; el botón de conexión sigue disponible.
+
+    Cambios relacionados: [#134](https://github.com/beclab/router/pull/134), [#136](https://github.com/beclab/router/pull/136).
   fullDescription: |
     Router ofrece a Olares una API unificada al estilo de OpenAI para modelos de IA locales y en la nube. Use una sola URL base y una clave de API para acceder a todos sus modelos, sin SDKs ni claves adicionales. Los administradores pueden conectar proveedores y gestionar modelos en un solo lugar, con una configuración y un control sencillos desde Mercado de Olares.
 

router/i18n/fr-FR/OlaresManifest.yaml

@@ -2,6 +2,12 @@ metadata:
   title: Router
   description: La passerelle IA native d'Olares.
 spec:
+  upgradeDescription: |
+    Router 2.7.59 corrige les erreurs de synchronisation des workflows FlowStudio dans la console.
+    La synchronisation accepte désormais de nouveaux types de workflows ainsi que les sorties audio et 3D.
+    Les workflows FlowStudio n’affichent plus le bouton de détails du modèle qui ne pouvait pas ouvrir ces détails ; le bouton de connexion reste disponible.
+
+    Modifications associées: [#134](https://github.com/beclab/router/pull/134), [#136](https://github.com/beclab/router/pull/136).
   fullDescription: |
     Router offre à Olares une API unifiée de type OpenAI pour les modèles d'IA locaux et cloud. Utilisez une seule URL de base et une clé API pour accéder à tous vos modèles, sans SDK ni clés supplémentaires. Les administrateurs peuvent connecter les fournisseurs et gérer les modèles en un seul endroit, avec une configuration et un contrôle simples depuis Marché Olares.
 

router/i18n/it-IT/OlaresManifest.yaml

@@ -2,6 +2,12 @@ metadata:
   title: Router
   description: Il gateway IA nativo per Olares.
 spec:
+  upgradeDescription: |
+    Router 2.7.59 corregge gli errori di sincronizzazione dei flussi di lavoro di FlowStudio nella console.
+    La sincronizzazione ora accetta nuovi tipi di flussi di lavoro e output audio e 3D.
+    I flussi di lavoro di FlowStudio non mostrano più il pulsante dei dettagli del modello che non si potevano aprire; il pulsante di connessione resta disponibile.
+
+    Modifiche correlate: [#134](https://github.com/beclab/router/pull/134), [#136](https://github.com/beclab/router/pull/136).
   fullDescription: |
     Router offre a Olares un'unica API in stile OpenAI per i modelli IA locali e cloud. Usa un solo URL di base e una chiave API per accedere a tutti i tuoi modelli, senza SDK o chiavi aggiuntive. Gli amministratori possono collegare i provider e gestire i modelli in un unico punto, con configurazione e controllo semplici da Mercato Olares.
 

router/i18n/ja-JP/OlaresManifest.yaml

@@ -2,6 +2,12 @@ metadata:
   title: Router
   description: Olares ネイティブの AI ゲートウェイ。
 spec:
+  upgradeDescription: |
+    Router 2.7.59では、コンソールでのFlowStudioワークフローの同期エラーを修正しました。
+    同期時に新しいワークフローの種類と音声・3D出力を受け付けるようになりました。
+    FlowStudioワークフローでは、詳細を開けないモデル詳細ボタンを非表示にしました。接続ボタンは引き続き利用できます。
+
+    関連する変更: [#134](https://github.com/beclab/router/pull/134), [#136](https://github.com/beclab/router/pull/136).
   fullDescription: |
     Router は、Olares にローカルおよびクラウド AI モデル向けの統一された OpenAI 互換 API を提供します。1 つのベース URL と API キーだけで、追加の SDK やキーなしにすべてのモデルへアクセスできます。管理者は Olares マーケット から、プロバイダーの接続とモデル管理を一か所で簡単に行えます。