Skip to content
Open
Show file tree
Hide file tree
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
2 changes: 1 addition & 1 deletion startos/manifest/index.ts
Original file line number Diff line number Diff line change
Expand Up @@ -6,7 +6,7 @@ const variant = process.env.VARIANT || 'generic'
type Mutable<T> = { -readonly [K in keyof T]: Mutable<T[K]> }
const mutable = <T>(value: T): Mutable<T> => value as Mutable<T>

const upstreamBuild = 'b10450'
const upstreamBuild = 'b10573'

const imageConfigs = {
generic: {
Expand Down
152 changes: 66 additions & 86 deletions startos/versions/current.ts
Original file line number Diff line number Diff line change
@@ -1,93 +1,73 @@
import { IMPOSSIBLE, VersionInfo } from '@start9labs/start-sdk'

export const current = VersionInfo.of({
version: '1.0.10450:0',
version: '1.0.10573:0',
releaseNotes: {
en_US: `Updated llama.cpp to build b10450, and the **Set Model** action now remembers what you picked.

**This package**

- The **Set Model** action now opens pre-filled with your current selection instead of resetting to the defaults, so you can change one setting — or just check which custom model is configured — without re-entering everything.

**llama.cpp**

- A small maintenance bump — 12 builds since b10438.
- Adds support for the Kimi-K3 text model family.
- The server's request queue was reworked, improving behaviour under concurrent requests.
- Vulkan: better performance on Intel Xe graphics, and a workaround for a problematic Intel driver version.
- The built-in chat UI now masks the API key field so browsers stop offering to save it.
- More robust GGUF parsing — malformed model metadata is rejected instead of misread.
- The \`--mmap\`/\`--no-mmap\`/\`--mlock\` flags are now deprecated in favour of \`--load-mode\`. They still work; if you pass them through **Set Model**'s extra arguments, switch when convenient.

llama.cpp publishes one build per merged commit and does not provide a per-build changelog. Full commit range: https://github.com/ggml-org/llama.cpp/compare/b10438...b10450`,
es_ES: `Actualiza llama.cpp a la compilación b10450, y la acción **Establecer modelo** ahora recuerda lo que eligió.

**Este paquete**

- La acción **Establecer modelo** ahora se abre con su selección actual ya rellenada, en lugar de volver a los valores predeterminados, de modo que puede cambiar un solo ajuste —o simplemente comprobar qué modelo personalizado está configurado— sin volver a introducirlo todo.

**llama.cpp**

- Una pequeña actualización de mantenimiento: 12 compilaciones desde la b10438.
- Añade compatibilidad con la familia de modelos de texto Kimi-K3.
- Se ha rediseñado la cola de peticiones del servidor, mejorando su comportamiento con peticiones simultáneas.
- Vulkan: mejor rendimiento en gráficos Intel Xe y una solución alternativa para una versión problemática del controlador de Intel.
- La interfaz de chat integrada ahora enmascara el campo de la clave de API para que el navegador no ofrezca guardarla.
- Análisis de GGUF más robusto: los metadatos de modelo mal formados se rechazan en lugar de interpretarse mal.
- Las opciones \`--mmap\`/\`--no-mmap\`/\`--mlock\` quedan obsoletas en favor de \`--load-mode\`. Siguen funcionando; si las pasa como argumentos adicionales en **Establecer modelo**, cámbielas cuando le resulte cómodo.

llama.cpp publica una compilación por cada commit fusionado y no ofrece un registro de cambios por compilación. Rango completo de commits: https://github.com/ggml-org/llama.cpp/compare/b10438...b10450`,
de_DE: `Aktualisiert llama.cpp auf Build b10450, und die Aktion **Modell festlegen** merkt sich jetzt Ihre Auswahl.

**Dieses Paket**

- Die Aktion **Modell festlegen** öffnet sich jetzt mit Ihrer aktuellen Auswahl vorausgefüllt, statt auf die Standardwerte zurückzuspringen. So können Sie eine einzelne Einstellung ändern – oder einfach nachsehen, welches benutzerdefinierte Modell konfiguriert ist – ohne alles neu einzugeben.

**llama.cpp**

- Ein kleines Wartungsupdate – 12 Builds seit b10438.
- Ergänzt Unterstützung für die Textmodellfamilie Kimi-K3.
- Die Anfrage-Warteschlange des Servers wurde überarbeitet, was das Verhalten bei gleichzeitigen Anfragen verbessert.
- Vulkan: bessere Leistung auf Intel-Xe-Grafik sowie eine Umgehung für eine problematische Intel-Treiberversion.
- Die integrierte Chat-Oberfläche maskiert das Feld für den API-Schlüssel, damit Browser dessen Speicherung nicht mehr anbieten.
- Robustere GGUF-Auswertung: fehlerhafte Modell-Metadaten werden abgelehnt statt falsch gelesen.
- Die Optionen \`--mmap\`/\`--no-mmap\`/\`--mlock\` gelten zugunsten von \`--load-mode\` als veraltet. Sie funktionieren weiterhin; wenn Sie sie über die zusätzlichen Argumente von **Modell festlegen** übergeben, stellen Sie bei Gelegenheit um.

llama.cpp veröffentlicht einen Build pro zusammengeführtem Commit und stellt kein Änderungsprotokoll je Build bereit. Vollständiger Commit-Bereich: https://github.com/ggml-org/llama.cpp/compare/b10438...b10450`,
pl_PL: `Aktualizuje llama.cpp do kompilacji b10450, a akcja **Ustaw model** zapamiętuje teraz Twój wybór.

**Ten pakiet**

- Akcja **Ustaw model** otwiera się teraz wypełniona bieżącym wyborem, zamiast wracać do wartości domyślnych — możesz zmienić pojedyncze ustawienie lub po prostu sprawdzić, który model niestandardowy jest skonfigurowany, bez wpisywania wszystkiego od nowa.

**llama.cpp**

- Niewielka aktualizacja konserwacyjna — 12 kompilacji od b10438.
- Dodaje obsługę rodziny modeli tekstowych Kimi-K3.
- Przeprojektowano kolejkę żądań serwera, co poprawia zachowanie przy równoczesnych żądaniach.
- Vulkan: lepsza wydajność na układach graficznych Intel Xe oraz obejście problematycznej wersji sterownika Intela.
- Wbudowany interfejs czatu maskuje teraz pole klucza API, dzięki czemu przeglądarka nie proponuje jego zapisania.
- Solidniejsze przetwarzanie plików GGUF — błędne metadane modelu są odrzucane zamiast błędnie odczytywane.
- Opcje \`--mmap\`/\`--no-mmap\`/\`--mlock\` są przestarzałe na rzecz \`--load-mode\`. Nadal działają; jeśli przekazujesz je w dodatkowych argumentach akcji **Ustaw model**, zmień je przy okazji.

llama.cpp publikuje jedną kompilację na scalony commit i nie udostępnia listy zmian dla poszczególnych kompilacji. Pełny zakres commitów: https://github.com/ggml-org/llama.cpp/compare/b10438...b10450`,
fr_FR: `Met à jour llama.cpp vers la version b10450, et l'action **Définir le modèle** mémorise désormais votre sélection.

**Ce paquet**

- L'action **Définir le modèle** s'ouvre désormais pré-remplie avec votre sélection actuelle au lieu de revenir aux valeurs par défaut : vous pouvez ainsi modifier un seul paramètre — ou simplement vérifier quel modèle personnalisé est configuré — sans tout ressaisir.

**llama.cpp**

- Une petite mise à jour de maintenance — 12 versions depuis la b10438.
- Ajoute la prise en charge de la famille de modèles de texte Kimi-K3.
- La file d'attente des requêtes du serveur a été repensée, ce qui améliore le comportement en cas de requêtes simultanées.
- Vulkan : meilleures performances sur les cartes graphiques Intel Xe et contournement d'une version problématique du pilote Intel.
- L'interface de discussion intégrée masque désormais le champ de la clé d'API, afin que le navigateur ne propose plus de l'enregistrer.
- Analyse des fichiers GGUF plus robuste : des métadonnées de modèle malformées sont rejetées au lieu d'être mal interprétées.
- Les options \`--mmap\`/\`--no-mmap\`/\`--mlock\` sont désormais obsolètes au profit de \`--load-mode\`. Elles fonctionnent toujours ; si vous les passez dans les arguments supplémentaires de **Définir le modèle**, changez-les à votre convenance.

llama.cpp publie une version par commit fusionné et ne fournit pas de journal des modifications par version. Plage complète des commits : https://github.com/ggml-org/llama.cpp/compare/b10438...b10450`,
en_US: `Updated llama.cpp to build b10573.

A maintenance bump — 123 builds since b10450.

- Adds support for the BailingMoE3 architecture, which runs the Ling 3.0 Flash models, and for the Granite SWA and Granite MoE SWA architectures. Speculative decoding also accepts DSpark draft models for LFM2.
- Fixes a scheduler race that could produce wrong results when a model was split across more than one backend.
- Multimodal requests use less memory on the server, accept WebP images, gain a \`--mmproj-device\` flag to place the vision projector on a chosen device, and read images more accurately (LFM2 tiling, DeepSeek-OCR).
- Truncated or corrupted LoRA adapters are rejected with an error instead of loading silently with zero-padded weights, and grammar-constrained output falls back gracefully when a JSON schema uses a regular expression llama.cpp cannot compile.
- Faster inference on the CUDA and Vulkan builds.
- The built-in chat UI groups its built-in tools into Server and Browser, adds a browser page-info tool, and tidies up the settings navigation.
- The date-and-time tool moved out of the server into the chat UI. If you list tools explicitly in **Set Model**'s extra arguments (\`--tools ...\`), drop \`get_datetime\` from the list or the server will refuse to start.

llama.cpp publishes one build per merged commit and does not provide a per-build changelog. Full commit range: https://github.com/ggml-org/llama.cpp/compare/b10450...b10573`,
es_ES: `Actualiza llama.cpp a la compilación b10573.

Una actualización de mantenimiento: 123 compilaciones desde la b10450.

- Añade compatibilidad con la arquitectura BailingMoE3, que permite ejecutar los modelos Ling 3.0 Flash, y con las arquitecturas Granite SWA y Granite MoE SWA. La decodificación especulativa también admite modelos borrador DSpark para LFM2.
- Corrige una condición de carrera en el planificador que podía dar resultados incorrectos cuando un modelo se repartía entre más de un backend.
- Las peticiones multimodales consumen menos memoria en el servidor, aceptan imágenes WebP, incorporan la opción \`--mmproj-device\` para situar el proyector de visión en un dispositivo concreto y leen las imágenes con más precisión (mosaico de LFM2, DeepSeek-OCR).
- Los adaptadores LoRA truncados o dañados se rechazan con un error en lugar de cargarse en silencio con pesos rellenados de ceros, y la salida restringida por gramática recurre a una alternativa cuando un esquema JSON usa una expresión regular que llama.cpp no puede compilar.
- Inferencia más rápida en las compilaciones CUDA y Vulkan.
- La interfaz de chat integrada agrupa sus herramientas integradas en Servidor y Navegador, añade una herramienta de información de página del navegador y reorganiza la navegación de los ajustes.
- La herramienta de fecha y hora se ha trasladado del servidor a la interfaz de chat. Si enumera herramientas explícitamente en los argumentos adicionales de **Establecer modelo** (\`--tools ...\`), quite \`get_datetime\` de la lista o el servidor no arrancará.

llama.cpp publica una compilación por cada commit fusionado y no ofrece un registro de cambios por compilación. Rango completo de commits: https://github.com/ggml-org/llama.cpp/compare/b10450...b10573`,
de_DE: `Aktualisiert llama.cpp auf Build b10573.

Ein Wartungsupdate – 123 Builds seit b10450.

- Ergänzt Unterstützung für die Architektur BailingMoE3, mit der sich die Modelle Ling 3.0 Flash ausführen lassen, sowie für die Architekturen Granite SWA und Granite MoE SWA. Die spekulative Dekodierung akzeptiert außerdem DSpark-Entwurfsmodelle für LFM2.
- Behebt ein Wettlaufproblem im Planer, das zu falschen Ergebnissen führen konnte, wenn ein Modell auf mehr als ein Backend verteilt wurde.
- Multimodale Anfragen benötigen auf dem Server weniger Speicher, akzeptieren WebP-Bilder, erhalten die Option \`--mmproj-device\`, um den Vision-Projektor auf einem bestimmten Gerät auszuführen, und lesen Bilder genauer (LFM2-Kachelung, DeepSeek-OCR).
- Abgeschnittene oder beschädigte LoRA-Adapter werden mit einem Fehler abgelehnt, statt still mit nullgefüllten Gewichten geladen zu werden, und grammatikgebundene Ausgaben weichen auf eine Alternative aus, wenn ein JSON-Schema einen regulären Ausdruck enthält, den llama.cpp nicht übersetzen kann.
- Schnellere Inferenz in den CUDA- und Vulkan-Builds.
- Die integrierte Chat-Oberfläche gruppiert ihre eingebauten Werkzeuge in Server und Browser, ergänzt ein Browser-Werkzeug für Seiteninformationen und räumt die Einstellungsnavigation auf.
- Das Werkzeug für Datum und Uhrzeit ist vom Server in die Chat-Oberfläche umgezogen. Wenn Sie in den zusätzlichen Argumenten von **Modell festlegen** Werkzeuge ausdrücklich auflisten (\`--tools ...\`), entfernen Sie \`get_datetime\` aus der Liste, sonst startet der Server nicht.

llama.cpp veröffentlicht einen Build pro zusammengeführtem Commit und stellt kein Änderungsprotokoll je Build bereit. Vollständiger Commit-Bereich: https://github.com/ggml-org/llama.cpp/compare/b10450...b10573`,
pl_PL: `Aktualizuje llama.cpp do kompilacji b10573.

Aktualizacja konserwacyjna — 123 kompilacje od b10450.

- Dodaje obsługę architektury BailingMoE3, która pozwala uruchamiać modele Ling 3.0 Flash, oraz architektur Granite SWA i Granite MoE SWA. Dekodowanie spekulatywne przyjmuje też modele robocze DSpark dla LFM2.
- Naprawia sytuację wyścigu w harmonogramie, która mogła dawać błędne wyniki, gdy model był rozdzielony na więcej niż jeden backend.
- Żądania multimodalne zużywają mniej pamięci na serwerze, przyjmują obrazy WebP, zyskują opcję \`--mmproj-device\` pozwalającą umieścić projektor wizyjny na wybranym urządzeniu i dokładniej odczytują obrazy (kafelkowanie LFM2, DeepSeek-OCR).
- Obcięte lub uszkodzone adaptery LoRA są odrzucane z błędem, zamiast być po cichu wczytywane z wagami uzupełnionymi zerami, a wyjście ograniczone gramatyką korzysta z rozwiązania zastępczego, gdy schemat JSON zawiera wyrażenie regularne, którego llama.cpp nie potrafi skompilować.
- Szybsze wnioskowanie w kompilacjach CUDA i Vulkan.
- Wbudowany interfejs czatu grupuje wbudowane narzędzia w kategorie Serwer i Przeglądarka, dodaje przeglądarkowe narzędzie z informacjami o stronie i porządkuje nawigację w ustawieniach.
- Narzędzie daty i godziny zostało przeniesione z serwera do interfejsu czatu. Jeśli wymieniasz narzędzia jawnie w dodatkowych argumentach akcji **Ustaw model** (\`--tools ...\`), usuń \`get_datetime\` z listy, inaczej serwer się nie uruchomi.

llama.cpp publikuje jedną kompilację na scalony commit i nie udostępnia listy zmian dla poszczególnych kompilacji. Pełny zakres commitów: https://github.com/ggml-org/llama.cpp/compare/b10450...b10573`,
fr_FR: `Met à jour llama.cpp vers la version b10573.

Une mise à jour de maintenance — 123 versions depuis la b10450.

- Ajoute la prise en charge de l'architecture BailingMoE3, qui permet d'exécuter les modèles Ling 3.0 Flash, ainsi que des architectures Granite SWA et Granite MoE SWA. Le décodage spéculatif accepte également les modèles brouillons DSpark pour LFM2.
- Corrige une situation de compétition dans l'ordonnanceur qui pouvait produire des résultats erronés lorsqu'un modèle était réparti sur plus d'un backend.
- Les requêtes multimodales consomment moins de mémoire côté serveur, acceptent les images WebP, gagnent l'option \`--mmproj-device\` pour placer le projecteur visuel sur l'appareil choisi et lisent les images plus précisément (tuilage LFM2, DeepSeek-OCR).
- Les adaptateurs LoRA tronqués ou corrompus sont rejetés avec une erreur au lieu d'être chargés silencieusement avec des poids complétés par des zéros, et la sortie contrainte par grammaire bascule sur une solution de repli lorsqu'un schéma JSON utilise une expression régulière que llama.cpp ne sait pas compiler.
- Inférence plus rapide sur les versions CUDA et Vulkan.
- L'interface de discussion intégrée regroupe ses outils intégrés en Serveur et Navigateur, ajoute un outil d'informations de page du navigateur et réorganise la navigation des paramètres.
- L'outil de date et d'heure a été déplacé du serveur vers l'interface de discussion. Si vous listez explicitement des outils dans les arguments supplémentaires de **Définir le modèle** (\`--tools ...\`), retirez \`get_datetime\` de la liste, sinon le serveur ne démarrera pas.

llama.cpp publie une version par commit fusionné et ne fournit pas de journal des modifications par version. Plage complète des commits : https://github.com/ggml-org/llama.cpp/compare/b10450...b10573`,
},
migrations: {
up: async ({ effects }) => {},
Expand Down