diff --git a/content/clis/de/muse-code.md b/content/clis/de/muse-code.md new file mode 100644 index 00000000..140afeb0 --- /dev/null +++ b/content/clis/de/muse-code.md @@ -0,0 +1,71 @@ +## answer + +Muse Code ist Metas Terminal-Coding-Agent auf Basis von Muse Spark 1.2 für komplexe Softwareentwicklungsaufgaben in großen Codebeständen. + +## introduction + +Muse Code verbindet einen Hauptagenten mit persistenten Hintergrund-Agenten. Es plant Änderungen, bearbeitet Code, validiert Ergebnisse und protokolliert Modellaufrufe, Werkzeugläufe, Freigaben und Änderungen in einem lokalen, nur ergänzten Ereignisprotokoll, damit Sitzungen nach Unterbrechungen fortgesetzt werden können. + +## capabilities.items.0.title + +Persistente Hintergrund-Agenten + +## capabilities.items.0.description + +Spezialisierte Hintergrund-Agenten bleiben während der gesamten Sitzung aktiv und melden nützliche Erkenntnisse oder nächste Schritte an den Hauptagenten. + +## capabilities.items.1.title + +Ausfallsichere Fortsetzung + +## capabilities.items.1.description + +Ein lokales, nur ergänztes Ereignisprotokoll erfasst Modellaufrufe, Werkzeugläufe, Freigaben und Änderungen, sodass die Laufzeit Arbeit nach einem Absturz wiedergeben und fortsetzen kann. + +## capabilities.items.2.title + +Integrierte Planungs-Skills + +## capabilities.items.2.description + +Die integrierten Skills /plan, /grill und /goal unterstützen freigabepflichtige Planung, Planprüfung und ausdauernde Arbeit an einem festgelegten Ziel. + +## faq.items.0.question + +Was ist Muse Code? + +## faq.items.0.answer + +Muse Code ist Metas Terminal-Coding-Agent zum Planen, Implementieren und Validieren von Softwareänderungen in großen Codebeständen. + +## faq.items.1.question + +Welches Modell treibt Muse Code an? + +## faq.items.1.answer + +Muse Code wird von Muse Spark 1.2 angetrieben, einem auf Coding ausgerichteten Update der Muse-Spark-Modellfamilie von Meta. + +## faq.items.2.question + +Wie installiere und starte ich Muse Code? + +## faq.items.2.answer + +Führe Metas offiziellen Installationsbefehl unter macOS oder Linux aus und starte den Agenten anschließend mit muse im Terminal. + +## faq.items.3.question + +Ist Muse Code Open Source? + +## faq.items.3.answer + +Für Muse Code werden kein öffentliches Quellcode-Repository und keine Open-Source-Lizenz bereitgestellt. Die verteilte CLI ist proprietäre Software. + +## faq.items.4.question + +Wie setzt Muse Code unterbrochene Arbeit fort? + +## faq.items.4.answer + +Muse Code führt ein lokales, nur ergänztes Ereignisprotokoll über Modellaufrufe, Werkzeugläufe, Freigaben und Änderungen, damit die Laufzeit eine Sitzung nach einem Absturz wiedergeben und fortsetzen kann. diff --git a/content/clis/en/muse-code.md b/content/clis/en/muse-code.md new file mode 100644 index 00000000..88f79905 --- /dev/null +++ b/content/clis/en/muse-code.md @@ -0,0 +1,71 @@ +## answer + +Muse Code is Meta's terminal coding agent, powered by Muse Spark 1.2, for completing complex software engineering tasks across large repositories. + +## introduction + +Muse Code combines a main agent with persistent background agents. It plans changes, edits code, validates results, and records model calls, tool runs, approvals, and edits in a local append-only event log so sessions can resume after interruption. + +## capabilities.items.0.title + +Persistent background agents + +## capabilities.items.0.description + +Specialized background agents remain active throughout a session and report useful findings or next steps to the main agent. + +## capabilities.items.1.title + +Restart-safe execution + +## capabilities.items.1.description + +A local append-only event log records every model call, tool run, approval, and edit, allowing the runtime to replay and resume work after a crash. + +## capabilities.items.2.title + +Bundled planning skills + +## capabilities.items.2.description + +Built-in /plan, /grill, and /goal skills support approval-gated planning, plan review, and persistent work toward a specified objective. + +## faq.items.0.question + +What is Muse Code? + +## faq.items.0.answer + +Muse Code is Meta's terminal coding agent for planning, implementing, and validating software changes across large repositories. + +## faq.items.1.question + +Which model powers Muse Code? + +## faq.items.1.answer + +Muse Code is powered by Muse Spark 1.2, a coding-focused update to Meta's Muse Spark model family. + +## faq.items.2.question + +How do I install and launch Muse Code? + +## faq.items.2.answer + +Run Meta's official installation command on macOS or Linux, then launch the agent with muse from a terminal. + +## faq.items.3.question + +Is Muse Code open source? + +## faq.items.3.answer + +No public source repository or open-source license is provided for Muse Code. The distributed CLI is proprietary software. + +## faq.items.4.question + +How does Muse Code resume interrupted work? + +## faq.items.4.answer + +Muse Code keeps a local append-only event log of model calls, tool runs, approvals, and edits, allowing the runtime to replay and resume the session after a crash. diff --git a/content/clis/es/muse-code.md b/content/clis/es/muse-code.md new file mode 100644 index 00000000..8ad7f9b9 --- /dev/null +++ b/content/clis/es/muse-code.md @@ -0,0 +1,71 @@ +## answer + +Muse Code es el agente de programación para terminal de Meta, impulsado por Muse Spark 1.2, para completar tareas complejas de ingeniería de software en repositorios grandes. + +## introduction + +Muse Code combina un agente principal con agentes persistentes en segundo plano. Planifica cambios, edita código, valida resultados y registra llamadas al modelo, ejecuciones de herramientas, aprobaciones y ediciones en un historial local de solo anexado para reanudar sesiones tras una interrupción. + +## capabilities.items.0.title + +Agentes persistentes en segundo plano + +## capabilities.items.0.description + +Los agentes especializados permanecen activos durante toda la sesión y comunican al agente principal hallazgos útiles o los siguientes pasos. + +## capabilities.items.1.title + +Ejecución recuperable + +## capabilities.items.1.description + +Un historial local de solo anexado registra cada llamada al modelo, herramienta, aprobación y edición para reproducir y reanudar el trabajo después de un fallo. + +## capabilities.items.2.title + +Habilidades de planificación incluidas + +## capabilities.items.2.description + +Las habilidades /plan, /grill y /goal permiten crear planes sujetos a aprobación, someterlos a revisión y avanzar de forma persistente hacia un objetivo. + +## faq.items.0.question + +¿Qué es Muse Code? + +## faq.items.0.answer + +Muse Code es el agente de programación para terminal de Meta para planificar, implementar y validar cambios de software en repositorios grandes. + +## faq.items.1.question + +¿Qué modelo impulsa Muse Code? + +## faq.items.1.answer + +Muse Code funciona con Muse Spark 1.2, una actualización centrada en programación de la familia de modelos Muse Spark de Meta. + +## faq.items.2.question + +¿Cómo instalo e inicio Muse Code? + +## faq.items.2.answer + +Ejecuta el comando de instalación oficial de Meta en macOS o Linux y luego inicia el agente con muse desde un terminal. + +## faq.items.3.question + +¿Muse Code es de código abierto? + +## faq.items.3.answer + +Meta no proporciona un repositorio público de código fuente ni una licencia de código abierto para Muse Code. La CLI distribuida es software propietario. + +## faq.items.4.question + +¿Cómo reanuda Muse Code el trabajo interrumpido? + +## faq.items.4.answer + +Muse Code mantiene un historial local de solo anexado con llamadas al modelo, herramientas, aprobaciones y ediciones para reproducir y reanudar la sesión después de un fallo. diff --git a/content/clis/fr/muse-code.md b/content/clis/fr/muse-code.md new file mode 100644 index 00000000..dff7d57f --- /dev/null +++ b/content/clis/fr/muse-code.md @@ -0,0 +1,71 @@ +## answer + +Muse Code est l'agent de codage en terminal de Meta, propulsé par Muse Spark 1.2, pour accomplir des tâches complexes d'ingénierie logicielle dans de grands dépôts. + +## introduction + +Muse Code associe un agent principal à des agents persistants en arrière-plan. Il planifie les changements, modifie le code, valide les résultats et consigne les appels au modèle, les outils, les approbations et les modifications dans un journal local en ajout seul afin de reprendre une session après une interruption. + +## capabilities.items.0.title + +Agents persistants en arrière-plan + +## capabilities.items.0.description + +Des agents spécialisés restent actifs pendant toute la session et transmettent à l'agent principal les conclusions utiles ou les prochaines étapes. + +## capabilities.items.1.title + +Exécution résistante aux interruptions + +## capabilities.items.1.description + +Un journal local en ajout seul enregistre chaque appel au modèle, outil, approbation et modification pour permettre de rejouer et reprendre le travail après un arrêt. + +## capabilities.items.2.title + +Compétences de planification intégrées + +## capabilities.items.2.description + +Les compétences /plan, /grill et /goal prennent en charge la planification soumise à approbation, la vérification des plans et le travail persistant vers un objectif défini. + +## faq.items.0.question + +Qu'est-ce que Muse Code ? + +## faq.items.0.answer + +Muse Code est l'agent de codage en terminal de Meta pour planifier, implémenter et valider des changements logiciels dans de grands dépôts. + +## faq.items.1.question + +Quel modèle propulse Muse Code ? + +## faq.items.1.answer + +Muse Code repose sur Muse Spark 1.2, une mise à jour orientée codage de la famille de modèles Muse Spark de Meta. + +## faq.items.2.question + +Comment installer et lancer Muse Code ? + +## faq.items.2.answer + +Exécutez la commande d'installation officielle de Meta sur macOS ou Linux, puis lancez l'agent avec muse dans un terminal. + +## faq.items.3.question + +Muse Code est-il open source ? + +## faq.items.3.answer + +Aucun dépôt public de code source ni aucune licence open source ne sont fournis pour Muse Code. La CLI distribuée est un logiciel propriétaire. + +## faq.items.4.question + +Comment Muse Code reprend-il un travail interrompu ? + +## faq.items.4.answer + +Muse Code conserve un journal local en ajout seul des appels au modèle, des outils, des approbations et des modifications afin de rejouer et reprendre la session après un arrêt. diff --git a/content/clis/id/muse-code.md b/content/clis/id/muse-code.md new file mode 100644 index 00000000..1b500877 --- /dev/null +++ b/content/clis/id/muse-code.md @@ -0,0 +1,71 @@ +## answer + +Muse Code adalah agen coding terminal Meta yang didukung Muse Spark 1.2 untuk menyelesaikan tugas rekayasa perangkat lunak kompleks pada repositori besar. + +## introduction + +Muse Code menggabungkan agen utama dengan agen latar belakang persisten. Alat ini merencanakan perubahan, mengedit kode, memvalidasi hasil, serta mencatat panggilan model, alat, persetujuan, dan edit dalam log peristiwa lokal yang hanya ditambahkan agar sesi dapat dilanjutkan setelah terhenti. + +## capabilities.items.0.title + +Agen latar belakang persisten + +## capabilities.items.0.description + +Agen khusus tetap aktif sepanjang sesi dan menyampaikan temuan yang berguna atau langkah berikutnya kepada agen utama. + +## capabilities.items.1.title + +Eksekusi yang dapat dipulihkan + +## capabilities.items.1.description + +Log peristiwa lokal yang hanya ditambahkan mencatat setiap panggilan model, alat, persetujuan, dan edit sehingga pekerjaan dapat diputar ulang dan dilanjutkan setelah kegagalan. + +## capabilities.items.2.title + +Skill perencanaan bawaan + +## capabilities.items.2.description + +Skill /plan, /grill, dan /goal mendukung perencanaan dengan persetujuan, pengujian rencana, dan pekerjaan berkelanjutan menuju tujuan tertentu. + +## faq.items.0.question + +Apa itu Muse Code? + +## faq.items.0.answer + +Muse Code adalah agen coding terminal Meta untuk merencanakan, menerapkan, dan memvalidasi perubahan perangkat lunak pada repositori besar. + +## faq.items.1.question + +Model apa yang mendukung Muse Code? + +## faq.items.1.answer + +Muse Code didukung Muse Spark 1.2, pembaruan berfokus coding untuk keluarga model Muse Spark dari Meta. + +## faq.items.2.question + +Bagaimana cara memasang dan menjalankan Muse Code? + +## faq.items.2.answer + +Jalankan perintah instalasi resmi Meta di macOS atau Linux, lalu mulai agen dengan muse dari terminal. + +## faq.items.3.question + +Apakah Muse Code bersifat sumber terbuka? + +## faq.items.3.answer + +Tidak ada repositori sumber publik atau lisensi sumber terbuka yang disediakan untuk Muse Code. CLI yang didistribusikan adalah perangkat lunak proprietari. + +## faq.items.4.question + +Bagaimana Muse Code melanjutkan pekerjaan yang terhenti? + +## faq.items.4.answer + +Muse Code menyimpan log peristiwa lokal yang hanya ditambahkan untuk panggilan model, alat, persetujuan, dan edit agar runtime dapat memutar ulang serta melanjutkan sesi setelah kegagalan. diff --git a/content/clis/ja/muse-code.md b/content/clis/ja/muse-code.md new file mode 100644 index 00000000..2485fa2b --- /dev/null +++ b/content/clis/ja/muse-code.md @@ -0,0 +1,71 @@ +## answer + +Muse CodeはMuse Spark 1.2を搭載したMetaのターミナル向けコーディングエージェントで、大規模リポジトリの複雑なソフトウェア開発タスクを実行します。 + +## introduction + +Muse Codeはメインエージェントと常駐バックグラウンドエージェントを組み合わせます。変更の計画、コード編集、結果検証を行い、モデル呼び出し、ツール実行、承認、編集をローカルの追記専用イベントログに記録するため、中断後もセッションを再開できます。 + +## capabilities.items.0.title + +常駐バックグラウンドエージェント + +## capabilities.items.0.description + +専門化されたバックグラウンドエージェントがセッション中ずっと稼働し、有用な調査結果や次の手順をメインエージェントへ報告します。 + +## capabilities.items.1.title + +再開可能な実行 + +## capabilities.items.1.description + +ローカルの追記専用イベントログにモデル呼び出し、ツール実行、承認、編集を記録し、障害後に処理を再現して再開できます。 + +## capabilities.items.2.title + +組み込みの計画スキル + +## capabilities.items.2.description + +/plan、/grill、/goalスキルが、承認付きの計画作成、計画の検証、指定目標に向けた継続的な作業を支援します。 + +## faq.items.0.question + +Muse Codeとは何ですか？ + +## faq.items.0.answer + +Muse Codeは、大規模リポジトリのソフトウェア変更を計画、実装、検証するMetaのターミナル向けコーディングエージェントです。 + +## faq.items.1.question + +Muse Codeを動かすモデルは何ですか？ + +## faq.items.1.answer + +Muse Codeは、MetaのMuse Sparkモデルファミリーをコーディング向けに更新したMuse Spark 1.2を搭載しています。 + +## faq.items.2.question + +Muse Codeをインストールして起動する方法は？ + +## faq.items.2.answer + +macOSまたはLinuxでMeta公式のインストールコマンドを実行し、ターミナルからmuseでエージェントを起動します。 + +## faq.items.3.question + +Muse Codeはオープンソースですか？ + +## faq.items.3.answer + +Muse Codeには公開ソースリポジトリやオープンソースライセンスが提供されていません。配布されるCLIはプロプライエタリソフトウェアです。 + +## faq.items.4.question + +Muse Codeは中断した作業をどのように再開しますか？ + +## faq.items.4.answer + +Muse Codeはモデル呼び出し、ツール実行、承認、編集をローカルの追記専用イベントログに保存し、障害後にセッションを再現して再開できるようにします。 diff --git a/content/clis/ko/muse-code.md b/content/clis/ko/muse-code.md new file mode 100644 index 00000000..d811c09a --- /dev/null +++ b/content/clis/ko/muse-code.md @@ -0,0 +1,71 @@ +## answer + +Muse Code는 Muse Spark 1.2 기반의 Meta 터미널 코딩 에이전트로, 대규모 저장소에서 복잡한 소프트웨어 엔지니어링 작업을 수행합니다. + +## introduction + +Muse Code는 주 에이전트와 지속형 백그라운드 에이전트를 결합합니다. 변경을 계획하고 코드를 편집하며 결과를 검증하고, 모델 호출·도구 실행·승인·편집을 로컬 추가 전용 이벤트 로그에 기록해 중단 후에도 세션을 재개할 수 있습니다. + +## capabilities.items.0.title + +지속형 백그라운드 에이전트 + +## capabilities.items.0.description + +전문화된 백그라운드 에이전트가 세션 내내 활성 상태를 유지하며 유용한 조사 결과나 다음 단계를 주 에이전트에 전달합니다. + +## capabilities.items.1.title + +재개 가능한 실행 + +## capabilities.items.1.description + +로컬 추가 전용 이벤트 로그가 모든 모델 호출, 도구 실행, 승인, 편집을 기록하여 장애 후 작업을 재현하고 재개할 수 있습니다. + +## capabilities.items.2.title + +내장 계획 스킬 + +## capabilities.items.2.description + +/plan, /grill, /goal 스킬이 승인 기반 계획, 계획 검토, 지정된 목표를 향한 지속적인 작업을 지원합니다. + +## faq.items.0.question + +Muse Code란 무엇인가요? + +## faq.items.0.answer + +Muse Code는 대규모 저장소의 소프트웨어 변경을 계획하고 구현하며 검증하는 Meta의 터미널 코딩 에이전트입니다. + +## faq.items.1.question + +Muse Code는 어떤 모델을 사용하나요? + +## faq.items.1.answer + +Muse Code는 Meta의 Muse Spark 모델 계열을 코딩 중심으로 업데이트한 Muse Spark 1.2를 사용합니다. + +## faq.items.2.question + +Muse Code를 어떻게 설치하고 실행하나요? + +## faq.items.2.answer + +macOS 또는 Linux에서 Meta의 공식 설치 명령을 실행한 뒤 터미널에서 muse로 에이전트를 시작합니다. + +## faq.items.3.question + +Muse Code는 오픈 소스인가요? + +## faq.items.3.answer + +Muse Code에는 공개 소스 저장소나 오픈 소스 라이선스가 제공되지 않습니다. 배포되는 CLI는 독점 소프트웨어입니다. + +## faq.items.4.question + +Muse Code는 중단된 작업을 어떻게 재개하나요? + +## faq.items.4.answer + +Muse Code는 모델 호출, 도구 실행, 승인, 편집을 로컬 추가 전용 이벤트 로그에 저장하여 장애 후 런타임이 세션을 재현하고 재개할 수 있게 합니다. diff --git a/content/clis/pt/muse-code.md b/content/clis/pt/muse-code.md new file mode 100644 index 00000000..884e7398 --- /dev/null +++ b/content/clis/pt/muse-code.md @@ -0,0 +1,71 @@ +## answer + +Muse Code é o agente de programação para terminal da Meta, com Muse Spark 1.2, para concluir tarefas complexas de engenharia de software em grandes repositórios. + +## introduction + +Muse Code combina um agente principal com agentes persistentes em segundo plano. Ele planeja alterações, edita código, valida resultados e registra chamadas ao modelo, ferramentas, aprovações e edições em um log local somente de acréscimo para retomar sessões após uma interrupção. + +## capabilities.items.0.title + +Agentes persistentes em segundo plano + +## capabilities.items.0.description + +Agentes especializados permanecem ativos durante toda a sessão e comunicam descobertas úteis ou próximos passos ao agente principal. + +## capabilities.items.1.title + +Execução recuperável + +## capabilities.items.1.description + +Um log local somente de acréscimo registra cada chamada ao modelo, ferramenta, aprovação e edição, permitindo reproduzir e retomar o trabalho após uma falha. + +## capabilities.items.2.title + +Habilidades de planejamento integradas + +## capabilities.items.2.description + +As habilidades /plan, /grill e /goal oferecem planejamento sujeito a aprovação, revisão de planos e trabalho persistente em direção a um objetivo definido. + +## faq.items.0.question + +O que é Muse Code? + +## faq.items.0.answer + +Muse Code é o agente de programação para terminal da Meta para planejar, implementar e validar alterações de software em grandes repositórios. + +## faq.items.1.question + +Qual modelo alimenta o Muse Code? + +## faq.items.1.answer + +Muse Code usa o Muse Spark 1.2, uma atualização voltada à programação da família de modelos Muse Spark da Meta. + +## faq.items.2.question + +Como instalar e iniciar o Muse Code? + +## faq.items.2.answer + +Execute o comando oficial de instalação da Meta no macOS ou Linux e depois inicie o agente com muse em um terminal. + +## faq.items.3.question + +Muse Code é open source? + +## faq.items.3.answer + +Não há repositório público de código-fonte nem licença open source para o Muse Code. A CLI distribuída é um software proprietário. + +## faq.items.4.question + +Como o Muse Code retoma um trabalho interrompido? + +## faq.items.4.answer + +Muse Code mantém um log local somente de acréscimo de chamadas ao modelo, ferramentas, aprovações e edições para reproduzir e retomar a sessão após uma falha. diff --git a/content/clis/ru/muse-code.md b/content/clis/ru/muse-code.md new file mode 100644 index 00000000..2b006475 --- /dev/null +++ b/content/clis/ru/muse-code.md @@ -0,0 +1,71 @@ +## answer + +Muse Code — терминальный агент программирования Meta на базе Muse Spark 1.2 для выполнения сложных задач разработки в крупных репозиториях. + +## introduction + +Muse Code объединяет основного агента с постоянными фоновыми агентами. Он планирует изменения, редактирует код, проверяет результаты и записывает вызовы модели, запуски инструментов, согласования и правки в локальный журнал только для добавления, чтобы продолжать сеансы после прерывания. + +## capabilities.items.0.title + +Постоянные фоновые агенты + +## capabilities.items.0.description + +Специализированные фоновые агенты работают на протяжении всего сеанса и сообщают основному агенту полезные результаты или следующие шаги. + +## capabilities.items.1.title + +Возобновляемое выполнение + +## capabilities.items.1.description + +Локальный журнал только для добавления фиксирует каждый вызов модели, запуск инструмента, согласование и правку, позволяя воспроизвести и продолжить работу после сбоя. + +## capabilities.items.2.title + +Встроенные навыки планирования + +## capabilities.items.2.description + +Навыки /plan, /grill и /goal поддерживают планирование с согласованием, проверку плана и длительную работу над заданной целью. + +## faq.items.0.question + +Что такое Muse Code? + +## faq.items.0.answer + +Muse Code — терминальный агент программирования Meta для планирования, реализации и проверки изменений в крупных репозиториях. + +## faq.items.1.question + +Какая модель используется в Muse Code? + +## faq.items.1.answer + +Muse Code работает на Muse Spark 1.2 — ориентированном на программирование обновлении семейства моделей Muse Spark от Meta. + +## faq.items.2.question + +Как установить и запустить Muse Code? + +## faq.items.2.answer + +Выполните официальную команду установки Meta в macOS или Linux, затем запустите агента командой muse в терминале. + +## faq.items.3.question + +Muse Code имеет открытый исходный код? + +## faq.items.3.answer + +Для Muse Code не опубликованы репозиторий исходного кода и лицензия с открытым исходным кодом. Распространяемая CLI является проприетарным ПО. + +## faq.items.4.question + +Как Muse Code продолжает прерванную работу? + +## faq.items.4.answer + +Muse Code хранит локальный журнал только для добавления с вызовами модели, инструментами, согласованиями и правками, позволяя среде воспроизвести и продолжить сеанс после сбоя. diff --git a/content/clis/tr/muse-code.md b/content/clis/tr/muse-code.md new file mode 100644 index 00000000..5efdea16 --- /dev/null +++ b/content/clis/tr/muse-code.md @@ -0,0 +1,71 @@ +## answer + +Muse Code, büyük depolarda karmaşık yazılım mühendisliği görevlerini tamamlamak için Muse Spark 1.2 destekli Meta terminal kodlama ajanıdır. + +## introduction + +Muse Code, ana ajanı kalıcı arka plan ajanlarıyla birleştirir. Değişiklikleri planlar, kodu düzenler, sonuçları doğrular ve kesintiden sonra oturumların sürdürülebilmesi için model çağrılarını, araç çalıştırmalarını, onayları ve düzenlemeleri yerel bir salt eklemeli olay günlüğüne kaydeder. + +## capabilities.items.0.title + +Kalıcı arka plan ajanları + +## capabilities.items.0.description + +Uzmanlaşmış arka plan ajanları oturum boyunca etkin kalır ve yararlı bulguları veya sonraki adımları ana ajana bildirir. + +## capabilities.items.1.title + +Sürdürülebilir çalışma + +## capabilities.items.1.description + +Yerel salt eklemeli olay günlüğü her model çağrısını, araç çalıştırmasını, onayı ve düzenlemeyi kaydederek çalışmanın bir arızadan sonra yeniden oynatılmasını ve sürdürülmesini sağlar. + +## capabilities.items.2.title + +Yerleşik planlama becerileri + +## capabilities.items.2.description + +/plan, /grill ve /goal becerileri onaya bağlı planlama, plan inceleme ve belirli bir hedefe yönelik kalıcı çalışmayı destekler. + +## faq.items.0.question + +Muse Code nedir? + +## faq.items.0.answer + +Muse Code, büyük depolardaki yazılım değişikliklerini planlamak, uygulamak ve doğrulamak için Meta'nın terminal kodlama ajanıdır. + +## faq.items.1.question + +Muse Code hangi modeli kullanır? + +## faq.items.1.answer + +Muse Code, Meta'nın Muse Spark model ailesinin kodlama odaklı güncellemesi Muse Spark 1.2 tarafından desteklenir. + +## faq.items.2.question + +Muse Code nasıl kurulur ve başlatılır? + +## faq.items.2.answer + +macOS veya Linux'ta Meta'nın resmi kurulum komutunu çalıştırın, ardından ajanı terminalden muse komutuyla başlatın. + +## faq.items.3.question + +Muse Code açık kaynaklı mı? + +## faq.items.3.answer + +Muse Code için herkese açık bir kaynak deposu veya açık kaynak lisansı sunulmuyor. Dağıtılan CLI özel mülk yazılımdır. + +## faq.items.4.question + +Muse Code kesintiye uğrayan çalışmayı nasıl sürdürür? + +## faq.items.4.answer + +Muse Code model çağrılarını, araç çalıştırmalarını, onayları ve düzenlemeleri yerel salt eklemeli olay günlüğünde tutarak çalışma zamanının bir arızadan sonra oturumu yeniden oynatıp sürdürmesini sağlar. diff --git a/content/clis/zh-Hans/muse-code.md b/content/clis/zh-Hans/muse-code.md new file mode 100644 index 00000000..0411f271 --- /dev/null +++ b/content/clis/zh-Hans/muse-code.md @@ -0,0 +1,71 @@ +## answer + +Muse Code 是 Meta 基于 Muse Spark 1.2 的终端编码 Agent，用于在大型代码库中完成复杂的软件工程任务。 + +## introduction + +Muse Code 将主 Agent 与持续运行的后台 Agent 结合起来。它可以规划变更、编辑代码、验证结果，并将模型调用、工具运行、审批和编辑写入本地只追加事件日志，以便在中断后准确恢复会话。 + +## capabilities.items.0.title + +持续运行的后台 Agent + +## capabilities.items.0.description + +专门的后台 Agent 会在整个会话期间保持活动，并把有用的发现或后续步骤反馈给主 Agent。 + +## capabilities.items.1.title + +可恢复执行 + +## capabilities.items.1.description + +本地只追加事件日志记录每次模型调用、工具运行、审批和编辑，使运行时能在故障后重放并继续工作。 + +## capabilities.items.2.title + +内置规划技能 + +## capabilities.items.2.description + +内置的 /plan、/grill 和 /goal 技能分别支持需审批的规划、计划审查，以及朝指定目标持续推进。 + +## faq.items.0.question + +Muse Code 是什么？ + +## faq.items.0.answer + +Muse Code 是 Meta 的终端编码 Agent，用于在大型代码库中规划、实施并验证软件变更。 + +## faq.items.1.question + +Muse Code 使用什么模型？ + +## faq.items.1.answer + +Muse Code 由 Muse Spark 1.2 驱动，这是 Meta Muse Spark 模型系列面向编码任务的更新版本。 + +## faq.items.2.question + +如何安装和启动 Muse Code？ + +## faq.items.2.answer + +在 macOS 或 Linux 上运行 Meta 官方安装命令，然后在终端中执行 muse 启动 Agent。 + +## faq.items.3.question + +Muse Code 是开源的吗？ + +## faq.items.3.answer + +Muse Code 没有公开源码仓库或开源许可证，分发的 CLI 属于专有软件。 + +## faq.items.4.question + +Muse Code 如何恢复中断的任务？ + +## faq.items.4.answer + +Muse Code 会把模型调用、工具运行、审批和编辑写入本地只追加事件日志，使运行时能够在故障后重放并继续会话。 diff --git a/content/clis/zh-Hant/muse-code.md b/content/clis/zh-Hant/muse-code.md new file mode 100644 index 00000000..404fdcf8 --- /dev/null +++ b/content/clis/zh-Hant/muse-code.md @@ -0,0 +1,71 @@ +## answer + +Muse Code 是 Meta 以 Muse Spark 1.2 驅動的終端機程式設計 Agent，用於在大型程式碼庫中完成複雜的軟體工程任務。 + +## introduction + +Muse Code 結合主 Agent 與持續運作的背景 Agent。它能規劃變更、編輯程式碼、驗證結果，並將模型呼叫、工具執行、核准與編輯寫入本機僅附加事件紀錄，以便在中斷後準確恢復工作階段。 + +## capabilities.items.0.title + +持續運作的背景 Agent + +## capabilities.items.0.description + +專門的背景 Agent 會在整個工作階段保持運作，並將有用的發現或後續步驟回報給主 Agent。 + +## capabilities.items.1.title + +可恢復執行 + +## capabilities.items.1.description + +本機僅附加事件紀錄會保存每次模型呼叫、工具執行、核准與編輯，讓執行環境能在故障後重播並繼續工作。 + +## capabilities.items.2.title + +內建規劃技能 + +## capabilities.items.2.description + +內建的 /plan、/grill 與 /goal 技能分別支援需核准的規劃、計畫檢視，以及朝指定目標持續推進。 + +## faq.items.0.question + +Muse Code 是什麼？ + +## faq.items.0.answer + +Muse Code 是 Meta 的終端機程式設計 Agent，用於在大型程式碼庫中規劃、實作並驗證軟體變更。 + +## faq.items.1.question + +Muse Code 使用什麼模型？ + +## faq.items.1.answer + +Muse Code 由 Muse Spark 1.2 驅動，這是 Meta Muse Spark 模型系列針對程式設計任務的更新版本。 + +## faq.items.2.question + +如何安裝並啟動 Muse Code？ + +## faq.items.2.answer + +在 macOS 或 Linux 上執行 Meta 官方安裝命令，然後在終端機中執行 muse 啟動 Agent。 + +## faq.items.3.question + +Muse Code 是開放原始碼嗎？ + +## faq.items.3.answer + +Muse Code 沒有公開原始碼儲存庫或開放原始碼授權，所發佈的 CLI 屬於專有軟體。 + +## faq.items.4.question + +Muse Code 如何恢復中斷的任務？ + +## faq.items.4.answer + +Muse Code 會將模型呼叫、工具執行、核准與編輯寫入本機僅附加事件紀錄，讓執行環境能在故障後重播並繼續工作階段。 diff --git a/cspell.json b/cspell.json index 3d4d2bc3..e979cf7e 100644 --- a/cspell.json +++ b/cspell.json @@ -61,6 +61,7 @@ "betikler", "betiklere", "comprenderlo", + "Cowork", "dikonfigurasi", "entegrasyonlarıyla", "gehosteter", @@ -107,6 +108,7 @@ "önizlemesi", "sağlayıcılı", "token’ların", + "trilhões", "ccstatusline", "API'lerle", "acli", diff --git a/data/artificial-analysis-index.json b/data/artificial-analysis-index.json index 2cfe0e05..35f44d6c 100644 --- a/data/artificial-analysis-index.json +++ b/data/artificial-analysis-index.json @@ -3,8 +3,8 @@ "source": "Artificial Analysis", "sourceUrl": "https://artificialanalysis.ai/leaderboards/models", "methodologyUrl": "https://artificialanalysis.ai/methodology/intelligence-benchmarking", - "indexVersion": "4.1", - "observedAt": "2026-07-31", + "indexVersion": "4.1.1", + "observedAt": "2026-08-13", "legacyMissingModelIds": ["cursor-composer-2", "cursor-composer-2-5"], "entries": [ { @@ -129,8 +129,8 @@ }, { "modelId": "deepseek-3-2", - "score": 32, - "estimated": false, + "score": 33, + "estimated": true, "configuration": "DeepSeek V3.2 (Reasoning)" }, { @@ -159,31 +159,37 @@ }, { "modelId": "deepseek-v3-2-exp", - "score": 21, + "score": 22, "estimated": true, "configuration": "DeepSeek V3.2 Exp (Non-reasoning)" }, { "modelId": "deepseek-v3-terminus", - "score": 21, + "score": 22, "estimated": true, "configuration": "DeepSeek V3.1 Terminus (Non-reasoning)" }, { "modelId": "deepseek-v4-flash", - "score": 50, + "score": 52, "estimated": false, "configuration": "DeepSeek V4 Flash 0731 (max)" }, { "modelId": "deepseek-v4-flash-preview", - "score": 40, + "score": 42, "estimated": false, "configuration": "DeepSeek V4 Flash (max)" }, { "modelId": "deepseek-v4-pro", - "score": 44, + "score": 53, + "estimated": false, + "configuration": "DeepSeek V4 Pro 0813 (max)" + }, + { + "modelId": "deepseek-v4-pro-preview", + "score": 45, "estimated": false, "configuration": "DeepSeek V4 Pro (max)" }, @@ -477,7 +483,7 @@ }, { "modelId": "grok-4", - "score": 33, + "score": 34, "estimated": true, "configuration": "Grok 4" }, @@ -489,7 +495,7 @@ }, { "modelId": "grok-4-20", - "score": 37, + "score": 38, "estimated": true, "configuration": "Grok 4.20 0309 v2" }, @@ -501,13 +507,19 @@ }, { "modelId": "grok-4-5", - "score": 54, + "score": 56, "estimated": false, "configuration": "Grok 4.5 (high)" }, + { + "modelId": "grok-4-6", + "score": 61, + "estimated": false, + "configuration": "Grok 4.6 (high)" + }, { "modelId": "grok-4-fast", - "score": 16, + "score": 17, "estimated": true, "configuration": "Grok 4 Fast" }, @@ -639,10 +651,16 @@ }, { "modelId": "muse-spark-1-1", - "score": 51, + "score": 53, "estimated": false, "configuration": "Muse Spark 1.1 (xhigh)" }, + { + "modelId": "muse-spark-1-2", + "score": 57, + "estimated": false, + "configuration": "Muse Spark 1.2 (xhigh)" + }, { "modelId": "o3", "score": 30, @@ -663,7 +681,7 @@ }, { "modelId": "qwen3-5-122b-a10b", - "score": 32, + "score": 33, "estimated": false, "configuration": "Qwen3.5 122B A10B (Reasoning)" }, @@ -699,7 +717,7 @@ }, { "modelId": "qwen3-6-27b", - "score": 37, + "score": 38, "estimated": false, "configuration": "Qwen3.6 27B (Reasoning)" }, @@ -711,13 +729,13 @@ }, { "modelId": "qwen3-6-max-preview", - "score": 40, + "score": 41, "estimated": true, - "configuration": "Qwen3.6 Max" + "configuration": "Qwen3.6 Max Preview" }, { "modelId": "qwen3-7-max", - "score": 46, + "score": 47, "estimated": false, "configuration": "Qwen3.7 Max" }, @@ -727,6 +745,12 @@ "estimated": false, "configuration": "Qwen3.7 Plus" }, + { + "modelId": "qwen3-8-max", + "score": 58, + "estimated": false, + "configuration": "Qwen3.8 Max" + }, { "modelId": "qwen3-coder-next", "score": 21, diff --git a/data/data-health.json b/data/data-health.json index eeb8ce26..7949ca0d 100644 --- a/data/data-health.json +++ b/data/data-health.json @@ -1,5 +1,5 @@ { - "asOf": "2026-08-03", + "asOf": "2026-08-13", "thresholds": { "models": 30, "providers": 30, @@ -10,16 +10,16 @@ "vendors": 90 }, "summary": { - "totalRecords": 253, - "recordsWithSources": 253, - "verifiedRecords": 253, - "provenanceComplete": 253, + "totalRecords": 258, + "recordsWithSources": 258, + "verifiedRecords": 258, + "provenanceComplete": 258, "staleVerifiedRecords": 0, "translationPlaceholderValues": 300, "danglingRelationships": 0, - "modelBenchmarkCoverage": 9.5, + "modelBenchmarkCoverage": 9.2, "productsWithPricing": 66, - "productRecords": 66, + "productRecords": 67, "communityUrlsPopulated": 331, "communityUrlsWithProvenance": 331, "duplicatedVendorCommunityUrls": 0, @@ -35,9 +35,9 @@ "stale": 0 }, "clis": { - "total": 27, - "verified": 27, - "provenanceComplete": 27, + "total": 28, + "verified": 28, + "provenanceComplete": 28, "stale": 0 }, "desktops": { @@ -53,9 +53,9 @@ "stale": 0 }, "models": { - "total": 123, - "verified": 123, - "provenanceComplete": 123, + "total": 127, + "verified": 127, + "provenanceComplete": 127, "stale": 0 }, "providers": { diff --git a/data/generated/cli-landing-pages.json b/data/generated/cli-landing-pages.json index 5af0407c..68d6dc95 100644 --- a/data/generated/cli-landing-pages.json +++ b/data/generated/cli-landing-pages.json @@ -924,6 +924,50 @@ ] } }, + "muse-code": { + "answer": "Muse Code is Meta's terminal coding agent, powered by Muse Spark 1.2, for completing complex software engineering tasks across large repositories.", + "introduction": "Muse Code combines a main agent with persistent background agents. It plans changes, edits code, validates results, and records model calls, tool runs, approvals, and edits in a local append-only event log so sessions can resume after interruption.", + "capabilities": { + "items": [ + { + "title": "Persistent background agents", + "description": "Specialized background agents remain active throughout a session and report useful findings or next steps to the main agent." + }, + { + "title": "Restart-safe execution", + "description": "A local append-only event log records every model call, tool run, approval, and edit, allowing the runtime to replay and resume work after a crash." + }, + { + "title": "Bundled planning skills", + "description": "Built-in /plan, /grill, and /goal skills support approval-gated planning, plan review, and persistent work toward a specified objective." + } + ] + }, + "faq": { + "items": [ + { + "question": "What is Muse Code?", + "answer": "Muse Code is Meta's terminal coding agent for planning, implementing, and validating software changes across large repositories." + }, + { + "question": "Which model powers Muse Code?", + "answer": "Muse Code is powered by Muse Spark 1.2, a coding-focused update to Meta's Muse Spark model family." + }, + { + "question": "How do I install and launch Muse Code?", + "answer": "Run Meta's official installation command on macOS or Linux, then launch the agent with muse from a terminal." + }, + { + "question": "Is Muse Code open source?", + "answer": "No public source repository or open-source license is provided for Muse Code. The distributed CLI is proprietary software." + }, + { + "question": "How does Muse Code resume interrupted work?", + "answer": "Muse Code keeps a local append-only event log of model calls, tool runs, approvals, and edits, allowing the runtime to replay and resume the session after a crash." + } + ] + } + }, "omp": { "answer": "omp is Oh My Pi's open-source terminal coding tool for hash-anchored edits, LSP and debugger context, subagents, persistent runtimes, and broad model-provider support.", "introduction": "omp targets development work that benefits from precise edit locations and richer execution context. Hash anchors identify where a change belongs, while language-server data, debugging tools, delegated subagent work, and reusable runtimes support longer or more technical tasks.", @@ -2114,6 +2158,50 @@ ] } }, + "muse-code": { + "answer": "Muse Code ist Metas Terminal-Coding-Agent auf Basis von Muse Spark 1.2 für komplexe Softwareentwicklungsaufgaben in großen Codebeständen.", + "introduction": "Muse Code verbindet einen Hauptagenten mit persistenten Hintergrund-Agenten. Es plant Änderungen, bearbeitet Code, validiert Ergebnisse und protokolliert Modellaufrufe, Werkzeugläufe, Freigaben und Änderungen in einem lokalen, nur ergänzten Ereignisprotokoll, damit Sitzungen nach Unterbrechungen fortgesetzt werden können.", + "capabilities": { + "items": [ + { + "title": "Persistente Hintergrund-Agenten", + "description": "Spezialisierte Hintergrund-Agenten bleiben während der gesamten Sitzung aktiv und melden nützliche Erkenntnisse oder nächste Schritte an den Hauptagenten." + }, + { + "title": "Ausfallsichere Fortsetzung", + "description": "Ein lokales, nur ergänztes Ereignisprotokoll erfasst Modellaufrufe, Werkzeugläufe, Freigaben und Änderungen, sodass die Laufzeit Arbeit nach einem Absturz wiedergeben und fortsetzen kann." + }, + { + "title": "Integrierte Planungs-Skills", + "description": "Die integrierten Skills /plan, /grill und /goal unterstützen freigabepflichtige Planung, Planprüfung und ausdauernde Arbeit an einem festgelegten Ziel." + } + ] + }, + "faq": { + "items": [ + { + "question": "Was ist Muse Code?", + "answer": "Muse Code ist Metas Terminal-Coding-Agent zum Planen, Implementieren und Validieren von Softwareänderungen in großen Codebeständen." + }, + { + "question": "Welches Modell treibt Muse Code an?", + "answer": "Muse Code wird von Muse Spark 1.2 angetrieben, einem auf Coding ausgerichteten Update der Muse-Spark-Modellfamilie von Meta." + }, + { + "question": "Wie installiere und starte ich Muse Code?", + "answer": "Führe Metas offiziellen Installationsbefehl unter macOS oder Linux aus und starte den Agenten anschließend mit muse im Terminal." + }, + { + "question": "Ist Muse Code Open Source?", + "answer": "Für Muse Code werden kein öffentliches Quellcode-Repository und keine Open-Source-Lizenz bereitgestellt. Die verteilte CLI ist proprietäre Software." + }, + { + "question": "Wie setzt Muse Code unterbrochene Arbeit fort?", + "answer": "Muse Code führt ein lokales, nur ergänztes Ereignisprotokoll über Modellaufrufe, Werkzeugläufe, Freigaben und Änderungen, damit die Laufzeit eine Sitzung nach einem Absturz wiedergeben und fortsetzen kann." + } + ] + } + }, "omp": { "answer": "omp ist das quelloffene Terminal-Coding-Werkzeug von Oh My Pi mit hash-verankerten Änderungen, LSP- und Debugger-Kontext, Subagents, persistenten Laufzeiten und Unterstützung für viele Modellanbieter.", "introduction": "omp bindet Änderungen über Hash-Anker an die vorgesehene Codestelle und verringert so Fehländerungen durch ungenaue Texttreffer. LSP, Debugger und Subagents können in Terminalaufgaben einbezogen werden; Laufzeiten bleiben für spätere Aufgaben verfügbar. Nach der Installation wird omp unter macOS, Windows oder Linux gestartet.", @@ -3304,6 +3392,50 @@ ] } }, + "muse-code": { + "answer": "Muse Code es el agente de programación para terminal de Meta, impulsado por Muse Spark 1.2, para completar tareas complejas de ingeniería de software en repositorios grandes.", + "introduction": "Muse Code combina un agente principal con agentes persistentes en segundo plano. Planifica cambios, edita código, valida resultados y registra llamadas al modelo, ejecuciones de herramientas, aprobaciones y ediciones en un historial local de solo anexado para reanudar sesiones tras una interrupción.", + "capabilities": { + "items": [ + { + "title": "Agentes persistentes en segundo plano", + "description": "Los agentes especializados permanecen activos durante toda la sesión y comunican al agente principal hallazgos útiles o los siguientes pasos." + }, + { + "title": "Ejecución recuperable", + "description": "Un historial local de solo anexado registra cada llamada al modelo, herramienta, aprobación y edición para reproducir y reanudar el trabajo después de un fallo." + }, + { + "title": "Habilidades de planificación incluidas", + "description": "Las habilidades /plan, /grill y /goal permiten crear planes sujetos a aprobación, someterlos a revisión y avanzar de forma persistente hacia un objetivo." + } + ] + }, + "faq": { + "items": [ + { + "question": "¿Qué es Muse Code?", + "answer": "Muse Code es el agente de programación para terminal de Meta para planificar, implementar y validar cambios de software en repositorios grandes." + }, + { + "question": "¿Qué modelo impulsa Muse Code?", + "answer": "Muse Code funciona con Muse Spark 1.2, una actualización centrada en programación de la familia de modelos Muse Spark de Meta." + }, + { + "question": "¿Cómo instalo e inicio Muse Code?", + "answer": "Ejecuta el comando de instalación oficial de Meta en macOS o Linux y luego inicia el agente con muse desde un terminal." + }, + { + "question": "¿Muse Code es de código abierto?", + "answer": "Meta no proporciona un repositorio público de código fuente ni una licencia de código abierto para Muse Code. La CLI distribuida es software propietario." + }, + { + "question": "¿Cómo reanuda Muse Code el trabajo interrumpido?", + "answer": "Muse Code mantiene un historial local de solo anexado con llamadas al modelo, herramientas, aprobaciones y ediciones para reproducir y reanudar la sesión después de un fallo." + } + ] + } + }, "omp": { "answer": "omp es la herramienta de programación para terminal de código abierto de Oh My Pi, con ediciones ancladas por hash, contexto LSP y del depurador, subagentes, entornos persistentes y numerosos proveedores de modelos.", "introduction": "omp vincula cada cambio con la ubicación prevista mediante un ancla hash, lo que reduce los errores de una coincidencia de texto imprecisa. Puede combinar LSP, depurador y subagentes en una tarea, y conservar los entornos de ejecución para trabajos posteriores. Tras la instalación, omp funciona en macOS, Windows o Linux.", @@ -4494,6 +4626,50 @@ ] } }, + "muse-code": { + "answer": "Muse Code est l'agent de codage en terminal de Meta, propulsé par Muse Spark 1.2, pour accomplir des tâches complexes d'ingénierie logicielle dans de grands dépôts.", + "introduction": "Muse Code associe un agent principal à des agents persistants en arrière-plan. Il planifie les changements, modifie le code, valide les résultats et consigne les appels au modèle, les outils, les approbations et les modifications dans un journal local en ajout seul afin de reprendre une session après une interruption.", + "capabilities": { + "items": [ + { + "title": "Agents persistants en arrière-plan", + "description": "Des agents spécialisés restent actifs pendant toute la session et transmettent à l'agent principal les conclusions utiles ou les prochaines étapes." + }, + { + "title": "Exécution résistante aux interruptions", + "description": "Un journal local en ajout seul enregistre chaque appel au modèle, outil, approbation et modification pour permettre de rejouer et reprendre le travail après un arrêt." + }, + { + "title": "Compétences de planification intégrées", + "description": "Les compétences /plan, /grill et /goal prennent en charge la planification soumise à approbation, la vérification des plans et le travail persistant vers un objectif défini." + } + ] + }, + "faq": { + "items": [ + { + "question": "Qu'est-ce que Muse Code ?", + "answer": "Muse Code est l'agent de codage en terminal de Meta pour planifier, implémenter et valider des changements logiciels dans de grands dépôts." + }, + { + "question": "Quel modèle propulse Muse Code ?", + "answer": "Muse Code repose sur Muse Spark 1.2, une mise à jour orientée codage de la famille de modèles Muse Spark de Meta." + }, + { + "question": "Comment installer et lancer Muse Code ?", + "answer": "Exécutez la commande d'installation officielle de Meta sur macOS ou Linux, puis lancez l'agent avec muse dans un terminal." + }, + { + "question": "Muse Code est-il open source ?", + "answer": "Aucun dépôt public de code source ni aucune licence open source ne sont fournis pour Muse Code. La CLI distribuée est un logiciel propriétaire." + }, + { + "question": "Comment Muse Code reprend-il un travail interrompu ?", + "answer": "Muse Code conserve un journal local en ajout seul des appels au modèle, des outils, des approbations et des modifications afin de rejouer et reprendre la session après un arrêt." + } + ] + } + }, "omp": { "answer": "omp est l’outil de programmation en terminal open source d’Oh My Pi, avec des modifications ancrées par hash, le contexte LSP et du débogueur, des sous-Agents, des environnements persistants et de nombreux fournisseurs de modèles.", "introduction": "omp associe chaque modification à l’emplacement prévu du code grâce à une ancre de hash, ce qui réduit les erreurs liées à une simple correspondance textuelle. Il peut combiner LSP, débogueur et sous-Agents dans une tâche, tout en conservant les environnements d’exécution pour la suite. Après installation, omp s’exécute sous macOS, Windows ou Linux.", @@ -5684,6 +5860,50 @@ ] } }, + "muse-code": { + "answer": "Muse Code adalah agen coding terminal Meta yang didukung Muse Spark 1.2 untuk menyelesaikan tugas rekayasa perangkat lunak kompleks pada repositori besar.", + "introduction": "Muse Code menggabungkan agen utama dengan agen latar belakang persisten. Alat ini merencanakan perubahan, mengedit kode, memvalidasi hasil, serta mencatat panggilan model, alat, persetujuan, dan edit dalam log peristiwa lokal yang hanya ditambahkan agar sesi dapat dilanjutkan setelah terhenti.", + "capabilities": { + "items": [ + { + "title": "Agen latar belakang persisten", + "description": "Agen khusus tetap aktif sepanjang sesi dan menyampaikan temuan yang berguna atau langkah berikutnya kepada agen utama." + }, + { + "title": "Eksekusi yang dapat dipulihkan", + "description": "Log peristiwa lokal yang hanya ditambahkan mencatat setiap panggilan model, alat, persetujuan, dan edit sehingga pekerjaan dapat diputar ulang dan dilanjutkan setelah kegagalan." + }, + { + "title": "Skill perencanaan bawaan", + "description": "Skill /plan, /grill, dan /goal mendukung perencanaan dengan persetujuan, pengujian rencana, dan pekerjaan berkelanjutan menuju tujuan tertentu." + } + ] + }, + "faq": { + "items": [ + { + "question": "Apa itu Muse Code?", + "answer": "Muse Code adalah agen coding terminal Meta untuk merencanakan, menerapkan, dan memvalidasi perubahan perangkat lunak pada repositori besar." + }, + { + "question": "Model apa yang mendukung Muse Code?", + "answer": "Muse Code didukung Muse Spark 1.2, pembaruan berfokus coding untuk keluarga model Muse Spark dari Meta." + }, + { + "question": "Bagaimana cara memasang dan menjalankan Muse Code?", + "answer": "Jalankan perintah instalasi resmi Meta di macOS atau Linux, lalu mulai agen dengan muse dari terminal." + }, + { + "question": "Apakah Muse Code bersifat sumber terbuka?", + "answer": "Tidak ada repositori sumber publik atau lisensi sumber terbuka yang disediakan untuk Muse Code. CLI yang didistribusikan adalah perangkat lunak proprietari." + }, + { + "question": "Bagaimana Muse Code melanjutkan pekerjaan yang terhenti?", + "answer": "Muse Code menyimpan log peristiwa lokal yang hanya ditambahkan untuk panggilan model, alat, persetujuan, dan edit agar runtime dapat memutar ulang serta melanjutkan sesi setelah kegagalan." + } + ] + } + }, "omp": { "answer": "omp adalah alat coding terminal sumber terbuka dari Oh My Pi dengan edit berjangkar hash, konteks LSP dan debugger, subagen, lingkungan persisten, serta dukungan banyak penyedia model.", "introduction": "omp mengikat setiap perubahan ke lokasi kode yang dimaksud dengan jangkar hash sehingga mengurangi kesalahan akibat pencocokan teks yang longgar. LSP, debugger, dan subagen dapat dipakai dalam satu tugas, sementara lingkungan eksekusi tetap tersedia untuk pekerjaan berikutnya. Setelah dipasang, omp berjalan di macOS, Windows, atau Linux.", @@ -6874,6 +7094,50 @@ ] } }, + "muse-code": { + "answer": "Muse CodeはMuse Spark 1.2を搭載したMetaのターミナル向けコーディングエージェントで、大規模リポジトリの複雑なソフトウェア開発タスクを実行します。", + "introduction": "Muse Codeはメインエージェントと常駐バックグラウンドエージェントを組み合わせます。変更の計画、コード編集、結果検証を行い、モデル呼び出し、ツール実行、承認、編集をローカルの追記専用イベントログに記録するため、中断後もセッションを再開できます。", + "capabilities": { + "items": [ + { + "title": "常駐バックグラウンドエージェント", + "description": "専門化されたバックグラウンドエージェントがセッション中ずっと稼働し、有用な調査結果や次の手順をメインエージェントへ報告します。" + }, + { + "title": "再開可能な実行", + "description": "ローカルの追記専用イベントログにモデル呼び出し、ツール実行、承認、編集を記録し、障害後に処理を再現して再開できます。" + }, + { + "title": "組み込みの計画スキル", + "description": "/plan、/grill、/goalスキルが、承認付きの計画作成、計画の検証、指定目標に向けた継続的な作業を支援します。" + } + ] + }, + "faq": { + "items": [ + { + "question": "Muse Codeとは何ですか？", + "answer": "Muse Codeは、大規模リポジトリのソフトウェア変更を計画、実装、検証するMetaのターミナル向けコーディングエージェントです。" + }, + { + "question": "Muse Codeを動かすモデルは何ですか？", + "answer": "Muse Codeは、MetaのMuse Sparkモデルファミリーをコーディング向けに更新したMuse Spark 1.2を搭載しています。" + }, + { + "question": "Muse Codeをインストールして起動する方法は？", + "answer": "macOSまたはLinuxでMeta公式のインストールコマンドを実行し、ターミナルからmuseでエージェントを起動します。" + }, + { + "question": "Muse Codeはオープンソースですか？", + "answer": "Muse Codeには公開ソースリポジトリやオープンソースライセンスが提供されていません。配布されるCLIはプロプライエタリソフトウェアです。" + }, + { + "question": "Muse Codeは中断した作業をどのように再開しますか？", + "answer": "Muse Codeはモデル呼び出し、ツール実行、承認、編集をローカルの追記専用イベントログに保存し、障害後にセッションを再現して再開できるようにします。" + } + ] + } + }, "omp": { "answer": "omp は Oh My Pi のオープンソースなターミナル向けコーディングツールで、ハッシュを基準にした編集、LSP とデバッガーの文脈、サブ Agent、永続ランタイム、多数のモデルプロバイダーに対応します。", "introduction": "omp はハッシュアンカーで変更先を特定し、曖昧な文字列一致による誤編集を減らします。ターミナルの作業に LSP、デバッガー、サブ Agent を組み込み、ランタイムを次のタスクでも使える状態に保てます。インストール後は macOS、Windows、Linux で omp を実行します。", @@ -8064,6 +8328,50 @@ ] } }, + "muse-code": { + "answer": "Muse Code는 Muse Spark 1.2 기반의 Meta 터미널 코딩 에이전트로, 대규모 저장소에서 복잡한 소프트웨어 엔지니어링 작업을 수행합니다.", + "introduction": "Muse Code는 주 에이전트와 지속형 백그라운드 에이전트를 결합합니다. 변경을 계획하고 코드를 편집하며 결과를 검증하고, 모델 호출·도구 실행·승인·편집을 로컬 추가 전용 이벤트 로그에 기록해 중단 후에도 세션을 재개할 수 있습니다.", + "capabilities": { + "items": [ + { + "title": "지속형 백그라운드 에이전트", + "description": "전문화된 백그라운드 에이전트가 세션 내내 활성 상태를 유지하며 유용한 조사 결과나 다음 단계를 주 에이전트에 전달합니다." + }, + { + "title": "재개 가능한 실행", + "description": "로컬 추가 전용 이벤트 로그가 모든 모델 호출, 도구 실행, 승인, 편집을 기록하여 장애 후 작업을 재현하고 재개할 수 있습니다." + }, + { + "title": "내장 계획 스킬", + "description": "/plan, /grill, /goal 스킬이 승인 기반 계획, 계획 검토, 지정된 목표를 향한 지속적인 작업을 지원합니다." + } + ] + }, + "faq": { + "items": [ + { + "question": "Muse Code란 무엇인가요?", + "answer": "Muse Code는 대규모 저장소의 소프트웨어 변경을 계획하고 구현하며 검증하는 Meta의 터미널 코딩 에이전트입니다." + }, + { + "question": "Muse Code는 어떤 모델을 사용하나요?", + "answer": "Muse Code는 Meta의 Muse Spark 모델 계열을 코딩 중심으로 업데이트한 Muse Spark 1.2를 사용합니다." + }, + { + "question": "Muse Code를 어떻게 설치하고 실행하나요?", + "answer": "macOS 또는 Linux에서 Meta의 공식 설치 명령을 실행한 뒤 터미널에서 muse로 에이전트를 시작합니다." + }, + { + "question": "Muse Code는 오픈 소스인가요?", + "answer": "Muse Code에는 공개 소스 저장소나 오픈 소스 라이선스가 제공되지 않습니다. 배포되는 CLI는 독점 소프트웨어입니다." + }, + { + "question": "Muse Code는 중단된 작업을 어떻게 재개하나요?", + "answer": "Muse Code는 모델 호출, 도구 실행, 승인, 편집을 로컬 추가 전용 이벤트 로그에 저장하여 장애 후 런타임이 세션을 재현하고 재개할 수 있게 합니다." + } + ] + } + }, "omp": { "answer": "omp는 해시 기반 편집, LSP와 디버거 맥락, 서브에이전트, 지속 실행 환경, 다양한 모델 제공자 지원을 갖춘 Oh My Pi의 오픈 소스 터미널 코딩 도구입니다.", "introduction": "omp는 해시 앵커로 변경 위치를 지정해 느슨한 문자열 일치 때문에 잘못된 곳이 수정될 위험을 줄입니다. 터미널 작업에 LSP, 디버거, 서브에이전트를 함께 사용하고 실행 환경을 다음 작업에서도 유지할 수 있습니다. 설치 후 macOS, Windows, Linux에서 omp를 실행합니다.", @@ -9254,6 +9562,50 @@ ] } }, + "muse-code": { + "answer": "Muse Code é o agente de programação para terminal da Meta, com Muse Spark 1.2, para concluir tarefas complexas de engenharia de software em grandes repositórios.", + "introduction": "Muse Code combina um agente principal com agentes persistentes em segundo plano. Ele planeja alterações, edita código, valida resultados e registra chamadas ao modelo, ferramentas, aprovações e edições em um log local somente de acréscimo para retomar sessões após uma interrupção.", + "capabilities": { + "items": [ + { + "title": "Agentes persistentes em segundo plano", + "description": "Agentes especializados permanecem ativos durante toda a sessão e comunicam descobertas úteis ou próximos passos ao agente principal." + }, + { + "title": "Execução recuperável", + "description": "Um log local somente de acréscimo registra cada chamada ao modelo, ferramenta, aprovação e edição, permitindo reproduzir e retomar o trabalho após uma falha." + }, + { + "title": "Habilidades de planejamento integradas", + "description": "As habilidades /plan, /grill e /goal oferecem planejamento sujeito a aprovação, revisão de planos e trabalho persistente em direção a um objetivo definido." + } + ] + }, + "faq": { + "items": [ + { + "question": "O que é Muse Code?", + "answer": "Muse Code é o agente de programação para terminal da Meta para planejar, implementar e validar alterações de software em grandes repositórios." + }, + { + "question": "Qual modelo alimenta o Muse Code?", + "answer": "Muse Code usa o Muse Spark 1.2, uma atualização voltada à programação da família de modelos Muse Spark da Meta." + }, + { + "question": "Como instalar e iniciar o Muse Code?", + "answer": "Execute o comando oficial de instalação da Meta no macOS ou Linux e depois inicie o agente com muse em um terminal." + }, + { + "question": "Muse Code é open source?", + "answer": "Não há repositório público de código-fonte nem licença open source para o Muse Code. A CLI distribuída é um software proprietário." + }, + { + "question": "Como o Muse Code retoma um trabalho interrompido?", + "answer": "Muse Code mantém um log local somente de acréscimo de chamadas ao modelo, ferramentas, aprovações e edições para reproduzir e retomar a sessão após uma falha." + } + ] + } + }, "omp": { "answer": "omp é a ferramenta de programação para terminal de código aberto do Oh My Pi, com edições ancoradas por hash, contexto de LSP e depurador, subagentes, ambientes persistentes e suporte a muitos provedores de modelos.", "introduction": "omp vincula cada alteração ao local pretendido por meio de uma âncora hash, reduzindo erros causados por correspondência de texto imprecisa. LSP, depurador e subagentes podem participar da tarefa, e os ambientes de execução permanecem disponíveis depois dela. Após a instalação, omp funciona no macOS, Windows ou Linux.", @@ -10444,6 +10796,50 @@ ] } }, + "muse-code": { + "answer": "Muse Code — терминальный агент программирования Meta на базе Muse Spark 1.2 для выполнения сложных задач разработки в крупных репозиториях.", + "introduction": "Muse Code объединяет основного агента с постоянными фоновыми агентами. Он планирует изменения, редактирует код, проверяет результаты и записывает вызовы модели, запуски инструментов, согласования и правки в локальный журнал только для добавления, чтобы продолжать сеансы после прерывания.", + "capabilities": { + "items": [ + { + "title": "Постоянные фоновые агенты", + "description": "Специализированные фоновые агенты работают на протяжении всего сеанса и сообщают основному агенту полезные результаты или следующие шаги." + }, + { + "title": "Возобновляемое выполнение", + "description": "Локальный журнал только для добавления фиксирует каждый вызов модели, запуск инструмента, согласование и правку, позволяя воспроизвести и продолжить работу после сбоя." + }, + { + "title": "Встроенные навыки планирования", + "description": "Навыки /plan, /grill и /goal поддерживают планирование с согласованием, проверку плана и длительную работу над заданной целью." + } + ] + }, + "faq": { + "items": [ + { + "question": "Что такое Muse Code?", + "answer": "Muse Code — терминальный агент программирования Meta для планирования, реализации и проверки изменений в крупных репозиториях." + }, + { + "question": "Какая модель используется в Muse Code?", + "answer": "Muse Code работает на Muse Spark 1.2 — ориентированном на программирование обновлении семейства моделей Muse Spark от Meta." + }, + { + "question": "Как установить и запустить Muse Code?", + "answer": "Выполните официальную команду установки Meta в macOS или Linux, затем запустите агента командой muse в терминале." + }, + { + "question": "Muse Code имеет открытый исходный код?", + "answer": "Для Muse Code не опубликованы репозиторий исходного кода и лицензия с открытым исходным кодом. Распространяемая CLI является проприетарным ПО." + }, + { + "question": "Как Muse Code продолжает прерванную работу?", + "answer": "Muse Code хранит локальный журнал только для добавления с вызовами модели, инструментами, согласованиями и правками, позволяя среде воспроизвести и продолжить сеанс после сбоя." + } + ] + } + }, "omp": { "answer": "omp — открытый терминальный инструмент Oh My Pi с привязанными к хешам изменениями, контекстом LSP и отладчика, субагентами, сохраняемыми средами выполнения и поддержкой многих поставщиков моделей.", "introduction": "omp с помощью хеш-якоря связывает изменение с нужным местом кода и уменьшает риск ошибки из-за неточного текстового совпадения. В задаче можно использовать LSP, отладчик и субагентов, а среда выполнения сохраняется для последующей работы. После установки omp работает в macOS, Windows или Linux.", @@ -11634,6 +12030,50 @@ ] } }, + "muse-code": { + "answer": "Muse Code, büyük depolarda karmaşık yazılım mühendisliği görevlerini tamamlamak için Muse Spark 1.2 destekli Meta terminal kodlama ajanıdır.", + "introduction": "Muse Code, ana ajanı kalıcı arka plan ajanlarıyla birleştirir. Değişiklikleri planlar, kodu düzenler, sonuçları doğrular ve kesintiden sonra oturumların sürdürülebilmesi için model çağrılarını, araç çalıştırmalarını, onayları ve düzenlemeleri yerel bir salt eklemeli olay günlüğüne kaydeder.", + "capabilities": { + "items": [ + { + "title": "Kalıcı arka plan ajanları", + "description": "Uzmanlaşmış arka plan ajanları oturum boyunca etkin kalır ve yararlı bulguları veya sonraki adımları ana ajana bildirir." + }, + { + "title": "Sürdürülebilir çalışma", + "description": "Yerel salt eklemeli olay günlüğü her model çağrısını, araç çalıştırmasını, onayı ve düzenlemeyi kaydederek çalışmanın bir arızadan sonra yeniden oynatılmasını ve sürdürülmesini sağlar." + }, + { + "title": "Yerleşik planlama becerileri", + "description": "/plan, /grill ve /goal becerileri onaya bağlı planlama, plan inceleme ve belirli bir hedefe yönelik kalıcı çalışmayı destekler." + } + ] + }, + "faq": { + "items": [ + { + "question": "Muse Code nedir?", + "answer": "Muse Code, büyük depolardaki yazılım değişikliklerini planlamak, uygulamak ve doğrulamak için Meta'nın terminal kodlama ajanıdır." + }, + { + "question": "Muse Code hangi modeli kullanır?", + "answer": "Muse Code, Meta'nın Muse Spark model ailesinin kodlama odaklı güncellemesi Muse Spark 1.2 tarafından desteklenir." + }, + { + "question": "Muse Code nasıl kurulur ve başlatılır?", + "answer": "macOS veya Linux'ta Meta'nın resmi kurulum komutunu çalıştırın, ardından ajanı terminalden muse komutuyla başlatın." + }, + { + "question": "Muse Code açık kaynaklı mı?", + "answer": "Muse Code için herkese açık bir kaynak deposu veya açık kaynak lisansı sunulmuyor. Dağıtılan CLI özel mülk yazılımdır." + }, + { + "question": "Muse Code kesintiye uğrayan çalışmayı nasıl sürdürür?", + "answer": "Muse Code model çağrılarını, araç çalıştırmalarını, onayları ve düzenlemeleri yerel salt eklemeli olay günlüğünde tutarak çalışma zamanının bir arızadan sonra oturumu yeniden oynatıp sürdürmesini sağlar." + } + ] + } + }, "omp": { "answer": "omp; hash ile sabitlenmiş düzenlemeler, LSP ve hata ayıklayıcı bağlamı, alt ajanlar, kalıcı çalışma ortamları ve çok sayıda model sağlayıcısı desteği sunan Oh My Pi açık kaynak terminal kodlama aracıdır.", "introduction": "omp her değişikliği hash çapasıyla amaçlanan kod konumuna bağlayarak gevşek metin eşleşmesinden doğan yanlış düzenleme riskini azaltır. LSP, hata ayıklayıcı ve alt ajanlar aynı görevde kullanılabilir; çalışma ortamları sonraki görevler için korunur. Kurulumdan sonra omp macOS, Windows veya Linux'ta çalışır.", @@ -12824,6 +13264,50 @@ ] } }, + "muse-code": { + "answer": "Muse Code 是 Meta 基于 Muse Spark 1.2 的终端编码 Agent，用于在大型代码库中完成复杂的软件工程任务。", + "introduction": "Muse Code 将主 Agent 与持续运行的后台 Agent 结合起来。它可以规划变更、编辑代码、验证结果，并将模型调用、工具运行、审批和编辑写入本地只追加事件日志，以便在中断后准确恢复会话。", + "capabilities": { + "items": [ + { + "title": "持续运行的后台 Agent", + "description": "专门的后台 Agent 会在整个会话期间保持活动，并把有用的发现或后续步骤反馈给主 Agent。" + }, + { + "title": "可恢复执行", + "description": "本地只追加事件日志记录每次模型调用、工具运行、审批和编辑，使运行时能在故障后重放并继续工作。" + }, + { + "title": "内置规划技能", + "description": "内置的 /plan、/grill 和 /goal 技能分别支持需审批的规划、计划审查，以及朝指定目标持续推进。" + } + ] + }, + "faq": { + "items": [ + { + "question": "Muse Code 是什么？", + "answer": "Muse Code 是 Meta 的终端编码 Agent，用于在大型代码库中规划、实施并验证软件变更。" + }, + { + "question": "Muse Code 使用什么模型？", + "answer": "Muse Code 由 Muse Spark 1.2 驱动，这是 Meta Muse Spark 模型系列面向编码任务的更新版本。" + }, + { + "question": "如何安装和启动 Muse Code？", + "answer": "在 macOS 或 Linux 上运行 Meta 官方安装命令，然后在终端中执行 muse 启动 Agent。" + }, + { + "question": "Muse Code 是开源的吗？", + "answer": "Muse Code 没有公开源码仓库或开源许可证，分发的 CLI 属于专有软件。" + }, + { + "question": "Muse Code 如何恢复中断的任务？", + "answer": "Muse Code 会把模型调用、工具运行、审批和编辑写入本地只追加事件日志，使运行时能够在故障后重放并继续会话。" + } + ] + } + }, "omp": { "answer": "omp 是 Oh My Pi 的开源终端编码工具，提供基于哈希定位的编辑、LSP 与调试器上下文、子 Agent、持久运行环境和多模型提供商支持。", "introduction": "omp 通过哈希锚点把修改绑定到预期代码位置，降低宽泛文本匹配造成误改的风险。它还能在终端任务中结合 LSP、调试器和子 Agent，并保留运行环境以便后续任务继续使用；安装后可在 macOS、Windows 或 Linux 中运行 omp。", @@ -14014,6 +14498,50 @@ ] } }, + "muse-code": { + "answer": "Muse Code 是 Meta 以 Muse Spark 1.2 驅動的終端機程式設計 Agent，用於在大型程式碼庫中完成複雜的軟體工程任務。", + "introduction": "Muse Code 結合主 Agent 與持續運作的背景 Agent。它能規劃變更、編輯程式碼、驗證結果，並將模型呼叫、工具執行、核准與編輯寫入本機僅附加事件紀錄，以便在中斷後準確恢復工作階段。", + "capabilities": { + "items": [ + { + "title": "持續運作的背景 Agent", + "description": "專門的背景 Agent 會在整個工作階段保持運作，並將有用的發現或後續步驟回報給主 Agent。" + }, + { + "title": "可恢復執行", + "description": "本機僅附加事件紀錄會保存每次模型呼叫、工具執行、核准與編輯，讓執行環境能在故障後重播並繼續工作。" + }, + { + "title": "內建規劃技能", + "description": "內建的 /plan、/grill 與 /goal 技能分別支援需核准的規劃、計畫檢視，以及朝指定目標持續推進。" + } + ] + }, + "faq": { + "items": [ + { + "question": "Muse Code 是什麼？", + "answer": "Muse Code 是 Meta 的終端機程式設計 Agent，用於在大型程式碼庫中規劃、實作並驗證軟體變更。" + }, + { + "question": "Muse Code 使用什麼模型？", + "answer": "Muse Code 由 Muse Spark 1.2 驅動，這是 Meta Muse Spark 模型系列針對程式設計任務的更新版本。" + }, + { + "question": "如何安裝並啟動 Muse Code？", + "answer": "在 macOS 或 Linux 上執行 Meta 官方安裝命令，然後在終端機中執行 muse 啟動 Agent。" + }, + { + "question": "Muse Code 是開放原始碼嗎？", + "answer": "Muse Code 沒有公開原始碼儲存庫或開放原始碼授權，所發佈的 CLI 屬於專有軟體。" + }, + { + "question": "Muse Code 如何恢復中斷的任務？", + "answer": "Muse Code 會將模型呼叫、工具執行、核准與編輯寫入本機僅附加事件紀錄，讓執行環境能在故障後重播並繼續工作階段。" + } + ] + } + }, "omp": { "answer": "omp 是 Oh My Pi 的開源終端機程式設計工具，提供雜湊定位編輯、LSP 與偵錯器上下文、子 Agent、持久執行環境和多模型供應商支援。", "introduction": "omp 透過雜湊錨點把修改綁定到預期的程式碼位置，降低寬鬆文字比對造成誤改的風險。它也能在終端機任務中結合 LSP、偵錯器和子 Agent，並保留執行環境供後續任務使用；安裝後可在 macOS、Windows 或 Linux 中執行 omp。", diff --git a/data/model-price-intelligence-index.json b/data/model-price-intelligence-index.json index f1574da6..96580aa7 100644 --- a/data/model-price-intelligence-index.json +++ b/data/model-price-intelligence-index.json @@ -82,7 +82,7 @@ "labelAnchor": "end" }, { - "modelId": "grok-4-5", + "modelId": "grok-4-6", "labelDx": 18, "labelDy": -24, "labelAnchor": "start" diff --git a/docs/DATA-HEALTH.md b/docs/DATA-HEALTH.md index 5a1d9fe7..4638bae5 100644 --- a/docs/DATA-HEALTH.md +++ b/docs/DATA-HEALTH.md @@ -1,20 +1,20 @@ # Data Health Report -Snapshot date: 2026-08-03. Regenerate with `pnpm data-health:report`. +Snapshot date: 2026-08-13. Regenerate with `pnpm data-health:report`. ## Scorecard | Metric | Value | | --- | ---: | -| Manifest records | 253 | -| Records with structured sources | 253 | -| Verified records | 253 | -| Verified with complete provenance | 253 | +| Manifest records | 258 | +| Records with structured sources | 258 | +| Verified records | 258 | +| Verified with complete provenance | 258 | | Stale verified records | 0 | | Non-English values identical to English | 300 | | Dangling product relationships | 0 | -| Model benchmark coverage | 9.5% | -| Products with pricing | 66/66 | +| Model benchmark coverage | 9.2% | +| Products with pricing | 66/67 | | Community URLs with provenance | 331/331 | | Duplicated vendor community URLs | 0 | | Errors / warnings / info | 0 / 0 / 0 | @@ -24,10 +24,10 @@ Snapshot date: 2026-08-03. Regenerate with `pnpm data-health:report`. | Category | Total | Verified | Provenance complete | Stale | | --- | ---: | ---: | ---: | ---: | | ides | 9 | 9 | 9 | 0 | -| clis | 27 | 27 | 27 | 0 | +| clis | 28 | 28 | 28 | 0 | | desktops | 12 | 12 | 12 | 0 | | extensions | 18 | 18 | 18 | 0 | -| models | 123 | 123 | 123 | 0 | +| models | 127 | 127 | 127 | 0 | | providers | 17 | 17 | 17 | 0 | | vendors | 47 | 47 | 47 | 0 | diff --git a/manifests/clis/muse-code.json b/manifests/clis/muse-code.json new file mode 100644 index 00000000..0a9bd3c2 --- /dev/null +++ b/manifests/clis/muse-code.json @@ -0,0 +1,100 @@ +{ + "$schema": "../$schemas/cli.schema.json", + "id": "muse-code", + "name": "Muse Code", + "familyId": "muse-code", + "description": "Meta's terminal coding agent, powered by Muse Spark 1.2, for planning, implementing, and validating changes across large repositories with persistent background agents.", + "translations": { + "de": { + "description": "Metas Terminal-Coding-Agent auf Basis von Muse Spark 1.2 plant, implementiert und validiert Änderungen in großen Codebeständen mit persistenten Hintergrund-Agenten." + }, + "es": { + "description": "El agente de programación para terminal de Meta, impulsado por Muse Spark 1.2, planifica, implementa y valida cambios en repositorios grandes con agentes persistentes en segundo plano." + }, + "fr": { + "description": "L'agent de codage en terminal de Meta, propulsé par Muse Spark 1.2, planifie, implémente et valide des changements dans de grands dépôts avec des agents persistants en arrière-plan." + }, + "id": { + "description": "Agen coding terminal Meta yang didukung Muse Spark 1.2 untuk merencanakan, menerapkan, dan memvalidasi perubahan pada repositori besar dengan agen latar belakang persisten." + }, + "ja": { + "description": "Muse Spark 1.2を搭載したMetaのターミナル向けコーディングエージェント。常駐バックグラウンドエージェントを使い、大規模リポジトリの変更を計画、実装、検証します。" + }, + "ko": { + "description": "Muse Spark 1.2 기반의 Meta 터미널 코딩 에이전트로, 지속형 백그라운드 에이전트를 활용해 대규모 저장소의 변경을 계획하고 구현하며 검증합니다." + }, + "pt": { + "description": "O agente de programação para terminal da Meta, com Muse Spark 1.2, planeja, implementa e valida alterações em grandes repositórios com agentes persistentes em segundo plano." + }, + "ru": { + "description": "Терминальный агент программирования Meta на базе Muse Spark 1.2 планирует, реализует и проверяет изменения в крупных репозиториях с помощью постоянных фоновых агентов." + }, + "tr": { + "description": "Muse Spark 1.2 destekli Meta terminal kodlama ajanı, kalıcı arka plan ajanlarıyla büyük depolardaki değişiklikleri planlar, uygular ve doğrular." + }, + "zh-Hans": { + "description": "Muse Code 是 Meta 基于 Muse Spark 1.2 的终端编码 Agent，可借助持续运行的后台 Agent 在大型代码库中规划、实施并验证变更。" + }, + "zh-Hant": { + "description": "Muse Code 是 Meta 以 Muse Spark 1.2 驅動的終端機程式設計 Agent，可透過持續運作的背景 Agent 在大型程式碼庫中規劃、實作並驗證變更。" + } + }, + "verified": true, + "sources": [ + { + "url": "https://research.meta.ai/blog/introducing-muse-code-and-muse-spark-1-2", + "title": "Introducing Muse Code and Muse Spark 1.2", + "fields": ["name", "description", "websiteUrl", "vendor", "platforms"] + }, + { + "url": "https://dev.meta.ai/install.sh", + "title": "Muse Code installer", + "fields": ["resourceUrls.download", "platforms"] + }, + { + "url": "https://api.meta.ai/muse-code/channels/muse-stable", + "title": "Muse Code stable channel", + "fields": ["latestVersion"] + } + ], + "lastVerifiedAt": "2026-08-11", + "verifiedBy": "codex-agent", + "confidence": "high", + "websiteUrl": "https://dev.meta.ai", + "docsUrl": "https://dev.meta.ai", + "vendor": "Meta", + "latestVersion": "0.1.0-R708.1", + "githubUrl": null, + "license": "Proprietary", + "pricing": [], + "resourceUrls": { + "download": "https://dev.meta.ai/install.sh", + "changelog": null, + "pricing": null, + "issue": null + }, + "communityUrls": { + "linkedin": null, + "twitter": null, + "github": null, + "youtube": null, + "discord": null, + "reddit": null, + "blog": null + }, + "relatedProducts": [], + "platforms": [ + { + "os": "macOS", + "installCommand": "curl -fsSL https://dev.meta.ai/install.sh | bash", + "launchCommand": "muse", + "installPath": "~/.local/bin/muse" + }, + { + "os": "Linux", + "installCommand": "curl -fsSL https://dev.meta.ai/install.sh | bash", + "launchCommand": "muse", + "installPath": "~/.local/bin/muse" + } + ] +} diff --git a/manifests/collections.json b/manifests/collections.json index cacd3472..ed0c32be 100644 --- a/manifests/collections.json +++ b/manifests/collections.json @@ -2304,6 +2304,63 @@ "name": "AGENTS.md", "url": "https://agents.md", "description": "Open-format markdown file providing coding agents with project-specific instructions and context" + }, + { + "translations": { + "en": { + "name": "Agent Plugins", + "description": "Open, vendor-neutral v1.0.0 working-draft standard for packaging Agent Skills and MCP servers as portable plugins with a shared manifest and directory layout" + }, + "de": { + "name": "Agent Plugins", + "description": "Offener, herstellerneutraler Arbeitsentwurf v1.0.0 zum Verpacken von Agent Skills und MCP-Servern als portable Plugins mit gemeinsamem Manifest und Verzeichnislayout" + }, + "es": { + "name": "Agent Plugins", + "description": "Borrador abierto y neutral respecto a proveedores v1.0.0 para empaquetar Agent Skills y servidores MCP como plugins portátiles con un manifiesto y una estructura de directorios comunes" + }, + "fr": { + "name": "Agent Plugins", + "description": "Projet de norme ouvert et neutre vis-à-vis des fournisseurs v1.0.0 pour regrouper Agent Skills et serveurs MCP dans des plugins portables avec un manifeste et une arborescence communs" + }, + "id": { + "name": "Agent Plugins", + "description": "Draf kerja v1.0.0 yang terbuka dan netral vendor untuk mengemas Agent Skills dan server MCP sebagai plugin portabel dengan manifes dan tata letak direktori bersama" + }, + "ja": { + "name": "Agent Plugins", + "description": "Agent SkillsとMCPサーバーを共通のマニフェストとディレクトリ構成を持つポータブルなプラグインとしてパッケージ化する、オープンでベンダー中立なv1.0.0作業草案" + }, + "ko": { + "name": "Agent Plugins", + "description": "Agent Skills와 MCP 서버를 공통 매니페스트 및 디렉터리 구조를 갖춘 이식 가능한 플러그인으로 패키징하기 위한 개방형 벤더 중립 v1.0.0 작업 초안" + }, + "pt": { + "name": "Agent Plugins", + "description": "Rascunho aberto e neutro em relação a fornecedores v1.0.0 para empacotar Agent Skills e servidores MCP como plugins portáteis com manifesto e estrutura de diretórios compartilhados" + }, + "ru": { + "name": "Agent Plugins", + "description": "Открытый независимый от поставщиков рабочий проект v1.0.0 для упаковки Agent Skills и MCP-серверов в переносимые плагины с общим манифестом и структурой каталогов" + }, + "tr": { + "name": "Agent Plugins", + "description": "Agent Skills ve MCP sunucularını ortak bir bildirim ve dizin düzenine sahip taşınabilir eklentiler olarak paketlemek için açık, sağlayıcıdan bağımsız v1.0.0 çalışma taslağı" + }, + "zh-Hans": { + "name": "Agent Plugins", + "description": "开放、厂商中立的 v1.0.0 工作草案，通过统一清单和目录结构将 Agent Skills 与 MCP 服务器打包为可移植插件" + }, + "zh-Hant": { + "name": "Agent Plugins", + "description": "開放、廠商中立的 v1.0.0 工作草案，透過統一清單與目錄結構將 Agent Skills 和 MCP 伺服器封裝為可攜式外掛程式" + } + }, + "id": "agent-plugins", + "name": "Agent Plugins", + "url": "https://agent-plugins.org", + "description": "Open, vendor-neutral v1.0.0 working-draft standard for packaging Agent Skills and MCP servers as portable plugins with a shared manifest and directory layout", + "lastVerifiedAt": "2026-08-07" } ] }, diff --git a/manifests/models/deepseek-v4-pro-preview.json b/manifests/models/deepseek-v4-pro-preview.json new file mode 100644 index 00000000..9d64b7ad --- /dev/null +++ b/manifests/models/deepseek-v4-pro-preview.json @@ -0,0 +1,109 @@ +{ + "$schema": "../$schemas/model.schema.json", + "id": "deepseek-v4-pro-preview", + "name": "DeepSeek-V4-Pro-Preview", + "description": "DeepSeek's flagship open-weight V4 model with 1.6T total and 49B active parameters, 1M context, dual thinking modes, JSON output, and tool calls.", + "translations": { + "de": { + "description": "DeepSeeks Open-Weight-V4-Flaggschiff mit 1,6 Bio. Gesamt- und 49 Mrd. aktiven Parametern, 1 Mio. Kontext, zwei Denkmodi, JSON und Werkzeugaufrufen." + }, + "es": { + "description": "El modelo V4 insignia y de pesos abiertos de DeepSeek, con 1,6T parámetros totales y 49B activos, contexto de 1M, dos modos de pensamiento, JSON y herramientas." + }, + "fr": { + "description": "Le modèle V4 phare à poids ouverts de DeepSeek, avec 1,6 Bn de paramètres dont 49 Md actifs, contexte 1 M, deux modes de réflexion, JSON et outils." + }, + "id": { + "description": "Model V4 open-weight unggulan DeepSeek dengan 1,6T parameter total dan 49B aktif, konteks 1M, dua mode berpikir, keluaran JSON, dan pemanggilan alat." + }, + "ja": { + "description": "総1.6兆・アクティブ490億パラメータ、100万コンテキスト、2種類の思考モード、JSON出力、ツール呼び出しを備えたDeepSeekの旗艦オープンウェイトV4モデル。" + }, + "ko": { + "description": "총 1.6T 및 활성 49B 매개변수, 100만 컨텍스트, 이중 사고 모드, JSON 출력과 도구 호출을 지원하는 DeepSeek의 플래그십 오픈 웨이트 V4 모델입니다." + }, + "pt": { + "description": "O modelo V4 principal e de pesos abertos da DeepSeek, com 1,6T parâmetros totais e 49B ativos, contexto de 1M, dois modos de pensamento, JSON e ferramentas." + }, + "ru": { + "description": "Флагманская открытая модель DeepSeek V4: 1,6 трлн параметров, 49 млрд активных, контекст 1 млн, два режима рассуждений, JSON и инструменты." + }, + "tr": { + "description": "DeepSeek'in 1,6T toplam ve 49B aktif parametreli, 1M bağlamlı, çift düşünme modlu, JSON çıktılı ve araç çağrılı amiral açık ağırlıklı V4 modeli." + }, + "zh-Hans": { + "description": "DeepSeek 的旗舰开放权重 V4 模型，1.6 万亿总参数、490 亿激活参数，支持 100 万上下文、双思考模式、JSON 输出和工具调用。" + }, + "zh-Hant": { + "description": "DeepSeek 的旗艦開放權重 V4 模型，1.6 兆總參數、490 億啟用參數，支援 100 萬上下文、雙思考模式、JSON 輸出和工具呼叫。" + } + }, + "verified": true, + "sources": [ + { + "url": "https://api-docs.deepseek.com/news/news260424/", + "title": "DeepSeek V4 Preview release", + "fields": [ + "name", + "description", + "size", + "activeParameters", + "releaseDate", + "lifecycle", + "contextWindow", + "capabilities" + ] + }, + { + "url": "https://api-docs.deepseek.com/quick_start/pricing", + "title": "DeepSeek current models and pricing", + "fields": ["lifecycle"] + }, + { + "url": "https://api-docs.deepseek.com/updates/", + "title": "DeepSeek V4 Flash July 31 API update", + "fields": ["lifecycle"] + }, + { + "url": "https://artificialanalysis.ai/models/deepseek-v4-pro-0424", + "title": "Artificial Analysis DeepSeek V4 Pro preview model page", + "fields": ["platformUrls.artificialAnalysis"] + } + ], + "lastVerifiedAt": "2026-08-13", + "verifiedBy": "codex-agent", + "confidence": "high", + "websiteUrl": "https://www.deepseek.com", + "docsUrl": "https://api-docs.deepseek.com/quick_start/pricing", + "vendor": "DeepSeek", + "size": "1.6T", + "activeParameters": "49B", + "contextWindow": 1000000, + "maxOutput": 384000, + "tokenPricing": { + "status": "unavailable", + "reason": "historical-price-unverified", + "primaryOffer": null, + "offers": [] + }, + "releaseDate": "2026-04-24", + "lifecycle": "maintained", + "knowledgeCutoff": null, + "inputModalities": ["text"], + "outputModalities": ["text"], + "capabilities": ["function-calling", "tool-choice", "structured-outputs", "reasoning"], + "benchmarks": { + "sweBench": null, + "terminalBench": null, + "mmmu": null, + "mmmuPro": null, + "webDevArena": null, + "sciCode": null, + "liveCodeBench": null + }, + "platformUrls": { + "huggingface": "https://huggingface.co/deepseek-ai/DeepSeek-V4-Pro", + "artificialAnalysis": "https://artificialanalysis.ai/models/deepseek-v4-pro-0424", + "openrouter": null + } +} diff --git a/manifests/models/deepseek-v4-pro.json b/manifests/models/deepseek-v4-pro.json index f5819b75..27ab4951 100644 --- a/manifests/models/deepseek-v4-pro.json +++ b/manifests/models/deepseek-v4-pro.json @@ -1,86 +1,75 @@ { "$schema": "../$schemas/model.schema.json", "id": "deepseek-v4-pro", - "name": "DeepSeek-V4-Pro-Preview", - "description": "DeepSeek's flagship open-weight V4 model with 1.6T total and 49B active parameters, 1M context, dual thinking modes, JSON output, and tool calls.", + "name": "DeepSeek-V4-Pro", + "description": "DeepSeek's current V4 Pro model with a 1M-token context window, thinking and non-thinking modes, JSON output, tool calls, and Responses API support.", "translations": { "de": { - "description": "DeepSeeks Open-Weight-V4-Flaggschiff mit 1,6 Bio. Gesamt- und 49 Mrd. aktiven Parametern, 1 Mio. Kontext, zwei Denkmodi, JSON und Werkzeugaufrufen." + "description": "DeepSeeks aktuelles V4-Pro-Modell mit 1 Million Token Kontext, Denk- und Nicht-Denkmodus, JSON-Ausgabe, Werkzeugaufrufen und Unterstützung der Responses API." }, "es": { - "description": "El modelo V4 insignia y de pesos abiertos de DeepSeek, con 1,6T parámetros totales y 49B activos, contexto de 1M, dos modos de pensamiento, JSON y herramientas." + "description": "El modelo V4 Pro actual de DeepSeek, con contexto de un millón de tokens, modos con y sin razonamiento, salida JSON, herramientas y compatibilidad con Responses API." }, "fr": { - "description": "Le modèle V4 phare à poids ouverts de DeepSeek, avec 1,6 Bn de paramètres dont 49 Md actifs, contexte 1 M, deux modes de réflexion, JSON et outils." + "description": "Le modèle V4 Pro actuel de DeepSeek avec un contexte d’un million de jetons, des modes avec et sans raisonnement, une sortie JSON, des appels d’outils et la prise en charge de Responses API." }, "id": { - "description": "Model V4 open-weight unggulan DeepSeek dengan 1,6T parameter total dan 49B aktif, konteks 1M, dua mode berpikir, keluaran JSON, dan pemanggilan alat." + "description": "Model V4 Pro terbaru DeepSeek dengan konteks 1 juta token, mode berpikir dan non-berpikir, keluaran JSON, pemanggilan alat, serta dukungan Responses API." }, "ja": { - "description": "総1.6兆・アクティブ490億パラメータ、100万コンテキスト、2種類の思考モード、JSON出力、ツール呼び出しを備えたDeepSeekの旗艦オープンウェイトV4モデル。" + "description": "100万トークンのコンテキスト、思考・非思考モード、JSON出力、ツール呼び出し、Responses API対応を備えたDeepSeekの現行V4 Proモデル。" }, "ko": { - "description": "총 1.6T 및 활성 49B 매개변수, 100만 컨텍스트, 이중 사고 모드, JSON 출력과 도구 호출을 지원하는 DeepSeek의 플래그십 오픈 웨이트 V4 모델입니다." + "description": "100만 토큰 컨텍스트, 사고 및 비사고 모드, JSON 출력, 도구 호출과 Responses API를 지원하는 DeepSeek의 최신 V4 Pro 모델입니다." }, "pt": { - "description": "O modelo V4 principal e de pesos abertos da DeepSeek, com 1,6T parâmetros totais e 49B ativos, contexto de 1M, dois modos de pensamento, JSON e ferramentas." + "description": "O modelo V4 Pro atual da DeepSeek, com contexto de 1 milhão de tokens, modos com e sem raciocínio, saída JSON, chamadas de ferramentas e suporte à Responses API." }, "ru": { - "description": "Флагманская открытая модель DeepSeek V4: 1,6 трлн параметров, 49 млрд активных, контекст 1 млн, два режима рассуждений, JSON и инструменты." + "description": "Текущая модель DeepSeek V4 Pro с контекстом 1 млн токенов, режимами с рассуждением и без него, выводом JSON, вызовами инструментов и поддержкой Responses API." }, "tr": { - "description": "DeepSeek'in 1,6T toplam ve 49B aktif parametreli, 1M bağlamlı, çift düşünme modlu, JSON çıktılı ve araç çağrılı amiral açık ağırlıklı V4 modeli." + "description": "DeepSeek'in 1 milyon token bağlam, düşünme ve düşünmeme modları, JSON çıktısı, araç çağrıları ve Responses API desteği sunan güncel V4 Pro modeli." }, "zh-Hans": { - "description": "DeepSeek 的旗舰开放权重 V4 模型，1.6 万亿总参数、490 亿激活参数，支持 100 万上下文、双思考模式、JSON 输出和工具调用。" + "description": "DeepSeek 当前的 V4 Pro 模型，提供 100 万 token 上下文、思考与非思考模式、JSON 输出、工具调用和 Responses API 支持。" }, "zh-Hant": { - "description": "DeepSeek 的旗艦開放權重 V4 模型，1.6 兆總參數、490 億啟用參數，支援 100 萬上下文、雙思考模式、JSON 輸出和工具呼叫。" + "description": "DeepSeek 目前的 V4 Pro 模型，提供 100 萬 token 上下文、思考與非思考模式、JSON 輸出、工具呼叫和 Responses API 支援。" } }, "verified": true, "sources": [ { - "url": "https://api-docs.deepseek.com/news/news260424/", - "title": "DeepSeek V4 Preview release", + "url": "https://api-docs.deepseek.com/quick_start/pricing/", + "title": "DeepSeek V4 Pro 0813 models and pricing", "fields": [ "name", "description", - "size", - "activeParameters", - "releaseDate", - "lifecycle", - "contextWindow", - "capabilities" - ] - }, - { - "url": "https://api-docs.deepseek.com/quick_start/pricing", - "title": "DeepSeek models and pricing", - "fields": [ "docsUrl", "contextWindow", "maxOutput", "tokenPricing", + "lifecycle", "inputModalities", "outputModalities", "capabilities" ] }, { - "url": "https://api-docs.deepseek.com/updates/", - "title": "DeepSeek V4 Flash July 31 API update", - "fields": ["name", "lifecycle"] + "url": "https://artificialanalysis.ai/models/deepseek-v4-pro", + "title": "Artificial Analysis DeepSeek V4 Pro 0813 model page", + "fields": ["releaseDate", "platformUrls.artificialAnalysis"] } ], - "lastVerifiedAt": "2026-07-31", + "lastVerifiedAt": "2026-08-13", "verifiedBy": "codex-agent", "confidence": "high", "websiteUrl": "https://www.deepseek.com", - "docsUrl": "https://api-docs.deepseek.com/quick_start/pricing", + "docsUrl": "https://api-docs.deepseek.com/quick_start/pricing/", "vendor": "DeepSeek", - "size": "1.6T", - "activeParameters": "49B", + "size": null, + "activeParameters": null, "contextWindow": 1000000, "maxOutput": 384000, "tokenPricing": { @@ -108,7 +97,7 @@ } ] }, - "releaseDate": "2026-04-24", + "releaseDate": "2026-08-13", "lifecycle": "latest", "knowledgeCutoff": null, "inputModalities": ["text"], @@ -124,8 +113,8 @@ "liveCodeBench": null }, "platformUrls": { - "huggingface": "https://huggingface.co/deepseek-ai/DeepSeek-V4-Pro", - "artificialAnalysis": null, + "huggingface": null, + "artificialAnalysis": "https://artificialanalysis.ai/models/deepseek-v4-pro", "openrouter": null } } diff --git a/manifests/models/grok-4-5.json b/manifests/models/grok-4-5.json index 03693c08..5ac81b49 100644 --- a/manifests/models/grok-4-5.json +++ b/manifests/models/grok-4-5.json @@ -62,9 +62,14 @@ "url": "https://docs.x.ai/developers/release-notes", "title": "SpaceXAI API release notes", "fields": ["releaseDate"] + }, + { + "url": "https://artificialanalysis.ai/models/grok-4-5", + "title": "Artificial Analysis Grok 4.5 model page", + "fields": ["platformUrls.artificialAnalysis"] } ], - "lastVerifiedAt": "2026-07-21", + "lastVerifiedAt": "2026-08-11", "verifiedBy": "codex-agent", "confidence": "high", "websiteUrl": "https://x.ai", @@ -91,7 +96,7 @@ "rates": { "input": 2, "output": 6, - "cacheRead": 0.5, + "cacheRead": 0.3, "cacheWrite": null } } @@ -114,5 +119,9 @@ "sciCode": null, "liveCodeBench": null }, - "platformUrls": { "huggingface": null, "artificialAnalysis": null, "openrouter": null } + "platformUrls": { + "huggingface": null, + "artificialAnalysis": "https://artificialanalysis.ai/models/grok-4-5", + "openrouter": null + } } diff --git a/manifests/models/grok-4-6.json b/manifests/models/grok-4-6.json new file mode 100644 index 00000000..3ffedb86 --- /dev/null +++ b/manifests/models/grok-4-6.json @@ -0,0 +1,123 @@ +{ + "$schema": "../$schemas/model.schema.json", + "id": "grok-4-6", + "name": "Grok 4.6", + "description": "SpaceXAI's frontier model for coding, long-running agents, knowledge work, and ambitious interactive and visual projects.", + "translations": { + "de": { + "description": "SpaceXAIs Frontier-Modell für Programmierung, lang laufende Agenten, Wissensarbeit sowie anspruchsvolle interaktive und visuelle Projekte." + }, + "es": { + "description": "El modelo de frontera de SpaceXAI para programación, agentes de larga duración, trabajo del conocimiento y proyectos interactivos y visuales ambiciosos." + }, + "fr": { + "description": "Le modèle de pointe de SpaceXAI pour le codage, les agents de longue durée, le travail intellectuel et les projets interactifs et visuels ambitieux." + }, + "id": { + "description": "Model frontier SpaceXAI untuk coding, agen jangka panjang, pekerjaan pengetahuan, serta proyek interaktif dan visual yang ambisius." + }, + "ja": { + "description": "コーディング、長時間稼働するエージェント、知識労働、野心的なインタラクティブ・ビジュアルプロジェクト向けのSpaceXAIフロンティアモデル。" + }, + "ko": { + "description": "코딩, 장기 실행 에이전트, 지식 업무와 야심 찬 인터랙티브 및 시각 프로젝트를 위한 SpaceXAI의 프런티어 모델입니다." + }, + "pt": { + "description": "O modelo de fronteira da SpaceXAI para programação, agentes de longa duração, trabalho de conhecimento e projetos interativos e visuais ambiciosos." + }, + "ru": { + "description": "Передовая модель SpaceXAI для программирования, длительно работающих агентов, интеллектуальной работы и сложных интерактивных и визуальных проектов." + }, + "tr": { + "description": "SpaceXAI'ın kodlama, uzun süre çalışan ajanlar, bilgi çalışmaları ve iddialı etkileşimli ve görsel projeler için öncü modeli." + }, + "zh-Hans": { + "description": "SpaceXAI 面向编码、长时间运行的智能体、知识工作以及复杂交互与视觉项目的前沿模型。" + }, + "zh-Hant": { + "description": "SpaceXAI 面向程式設計、長時間運行的智慧體、知識工作以及複雜互動與視覺專案的前沿模型。" + } + }, + "verified": true, + "sources": [ + { + "url": "https://docs.x.ai/developers/models/grok-4.6", + "title": "Grok 4.6 model documentation", + "fields": [ + "name", + "docsUrl", + "contextWindow", + "maxOutput", + "tokenPricing", + "knowledgeCutoff", + "inputModalities", + "outputModalities", + "capabilities" + ] + }, + { + "url": "https://x.ai/news/grok-4-6", + "title": "Introducing Grok 4.6", + "fields": ["description", "releaseDate", "lifecycle", "websiteUrl"] + }, + { + "url": "https://artificialanalysis.ai/models/grok-4-6", + "title": "Artificial Analysis Grok 4.6 model page", + "fields": ["platformUrls.artificialAnalysis"] + } + ], + "lastVerifiedAt": "2026-08-13", + "verifiedBy": "codex-agent", + "confidence": "high", + "websiteUrl": "https://x.ai", + "docsUrl": "https://docs.x.ai/developers/models/grok-4.6", + "vendor": "xAI", + "size": null, + "activeParameters": null, + "contextWindow": 500000, + "maxOutput": null, + "tokenPricing": { + "status": "available", + "primaryOffer": "global-standard", + "offers": [ + { + "id": "global-standard", + "currency": "USD", + "region": "global", + "serviceTier": "standard", + "effectiveFrom": null, + "effectiveTo": null, + "tiers": [ + { + "condition": { "metric": "inputTokens", "min": 1, "max": 200000 }, + "rates": { "input": 2, "output": 6, "cacheRead": 0.5, "cacheWrite": null } + }, + { + "condition": { "metric": "inputTokens", "min": 200001, "max": 500000 }, + "rates": { "input": 4, "output": 12, "cacheRead": 1, "cacheWrite": null } + } + ] + } + ] + }, + "releaseDate": "2026-08-12", + "lifecycle": "latest", + "knowledgeCutoff": "2026-02-01", + "inputModalities": ["text", "image"], + "outputModalities": ["text"], + "capabilities": ["function-calling", "tool-choice", "structured-outputs", "reasoning"], + "benchmarks": { + "sweBench": null, + "terminalBench": null, + "mmmu": null, + "mmmuPro": null, + "webDevArena": null, + "sciCode": null, + "liveCodeBench": null + }, + "platformUrls": { + "huggingface": null, + "artificialAnalysis": "https://artificialanalysis.ai/models/grok-4-6", + "openrouter": null + } +} diff --git a/manifests/models/muse-spark-1-1.json b/manifests/models/muse-spark-1-1.json index 968470d9..40e4c6e8 100644 --- a/manifests/models/muse-spark-1-1.json +++ b/manifests/models/muse-spark-1-1.json @@ -63,10 +63,10 @@ { "url": "https://models.dev/api.json", "title": "Models.dev Meta pricing record", - "fields": ["referenceTokenPricing"] + "fields": ["lifecycle", "referenceTokenPricing"] } ], - "lastVerifiedAt": "2026-07-31", + "lastVerifiedAt": "2026-08-11", "verifiedBy": "codex-agent", "confidence": "high", "websiteUrl": "https://ai.meta.com", @@ -89,11 +89,11 @@ "source": { "name": "Models.dev — Meta", "url": "https://models.dev/api.json", - "observedAt": "2026-07-31" + "observedAt": "2026-08-11" } }, "releaseDate": "2026-07-09", - "lifecycle": "latest", + "lifecycle": "maintained", "knowledgeCutoff": null, "inputModalities": ["text", "image", "pdf", "audio", "video"], "outputModalities": ["text"], diff --git a/manifests/models/muse-spark-1-2.json b/manifests/models/muse-spark-1-2.json new file mode 100644 index 00000000..0066af54 --- /dev/null +++ b/manifests/models/muse-spark-1-2.json @@ -0,0 +1,107 @@ +{ + "$schema": "../$schemas/model.schema.json", + "id": "muse-spark-1-2", + "name": "Muse Spark 1.2", + "description": "Meta's coding-focused reasoning model for code generation, complex debugging, codebase understanding, long-horizon tasks, and end-to-end developer workflows.", + "translations": { + "de": { + "description": "Metas auf Programmierung fokussiertes Reasoning-Modell für Codegenerierung, komplexes Debugging, Codeverständnis, langfristige Aufgaben und durchgängige Entwicklerabläufe." + }, + "es": { + "description": "El modelo de razonamiento de Meta centrado en programación para generar código, depurar problemas complejos, comprender repositorios y completar flujos de desarrollo prolongados." + }, + "fr": { + "description": "Le modèle de raisonnement de Meta axé sur le code pour la génération, le débogage complexe, la compréhension des dépôts et les workflows de développement de longue durée." + }, + "id": { + "description": "Model penalaran Meta yang berfokus pada coding untuk pembuatan kode, debugging kompleks, pemahaman basis kode, tugas jangka panjang, dan alur pengembangan menyeluruh." + }, + "ja": { + "description": "コード生成、複雑なデバッグ、コードベース理解、長期タスク、エンドツーエンドの開発ワークフローに特化したMetaの推論モデル。" + }, + "ko": { + "description": "코드 생성, 복잡한 디버깅, 코드베이스 이해, 장기 작업 및 엔드투엔드 개발 워크플로에 초점을 맞춘 Meta의 추론 모델입니다." + }, + "pt": { + "description": "Modelo de raciocínio da Meta focado em programação para gerar código, depurar problemas complexos, compreender bases de código e concluir fluxos de desenvolvimento prolongados." + }, + "ru": { + "description": "Модель рассуждений Meta для генерации кода, сложной отладки, понимания кодовых баз, длительных задач и сквозных процессов разработки." + }, + "tr": { + "description": "Meta'nın kod üretimi, karmaşık hata ayıklama, kod tabanı anlama, uzun süreli görevler ve uçtan uca geliştirme iş akışlarına odaklanan akıl yürütme modeli." + }, + "zh-Hans": { + "description": "Meta 面向代码生成、复杂调试、代码库理解、长周期任务和端到端开发工作流的编码推理模型。" + }, + "zh-Hant": { + "description": "Meta 面向程式碼生成、複雜除錯、程式碼庫理解、長週期任務和端對端開發工作流程的程式設計推理模型。" + } + }, + "verified": true, + "sources": [ + { + "url": "https://research.meta.ai/blog/introducing-muse-code-and-muse-spark-1-2", + "title": "Introducing Muse Code and Muse Spark 1.2", + "fields": ["name", "description", "releaseDate", "lifecycle", "websiteUrl"] + }, + { + "url": "https://developer.meta.com/ai/", + "title": "Meta Model API", + "fields": ["docsUrl"] + }, + { + "url": "https://models.dev/api.json", + "title": "Models.dev Meta model record", + "fields": [ + "contextWindow", + "referenceTokenPricing", + "inputModalities", + "outputModalities", + "capabilities" + ] + } + ], + "lastVerifiedAt": "2026-08-11", + "verifiedBy": "codex-agent", + "confidence": "high", + "websiteUrl": "https://ai.meta.com", + "docsUrl": "https://developer.meta.com/ai/", + "vendor": "Meta", + "size": null, + "activeParameters": null, + "contextWindow": 1048576, + "maxOutput": null, + "tokenPricing": { + "status": "unavailable", + "reason": "official-price-not-published", + "primaryOffer": null, + "offers": [] + }, + "referenceTokenPricing": { + "currency": "USD", + "basis": "first-party-api", + "rates": { "input": 1.25, "output": 4.25, "cacheRead": 0.15, "cacheWrite": null }, + "source": { + "name": "Models.dev — Meta", + "url": "https://models.dev/api.json", + "observedAt": "2026-08-11" + } + }, + "releaseDate": "2026-08-05", + "lifecycle": "latest", + "knowledgeCutoff": null, + "inputModalities": ["text", "image", "pdf", "audio", "video"], + "outputModalities": ["text"], + "capabilities": ["function-calling", "tool-choice", "structured-outputs", "reasoning"], + "benchmarks": { + "sweBench": null, + "terminalBench": null, + "mmmu": null, + "mmmuPro": null, + "webDevArena": null, + "sciCode": null, + "liveCodeBench": null + }, + "platformUrls": { "huggingface": null, "artificialAnalysis": null, "openrouter": null } +} diff --git a/manifests/models/qwen3-8-max.json b/manifests/models/qwen3-8-max.json new file mode 100644 index 00000000..b9b56963 --- /dev/null +++ b/manifests/models/qwen3-8-max.json @@ -0,0 +1,104 @@ +{ + "$schema": "../$schemas/model.schema.json", + "id": "qwen3-8-max", + "name": "Qwen3.8-Max", + "description": "Alibaba's 2.4T-parameter flagship model with 95B active parameters for coding, real-world work, long-horizon tasks, and multimodal agents.", + "translations": { + "de": { + "description": "Alibabas Flaggschiffmodell mit 2,4 Billionen Parametern und 95 Milliarden aktiven Parametern für Programmierung, reale Arbeitsaufgaben, langfristige Aufgaben und multimodale Agenten." + }, + "es": { + "description": "El modelo insignia de Alibaba con 2,4 billones de parámetros y 95 mil millones activos para programación, trabajo real, tareas prolongadas y agentes multimodales." + }, + "fr": { + "description": "Le modèle phare d’Alibaba doté de 2,4 billions de paramètres, dont 95 milliards actifs, pour le codage, le travail réel, les tâches longues et les agents multimodaux." + }, + "id": { + "description": "Model unggulan Alibaba dengan 2,4 triliun parameter dan 95 miliar parameter aktif untuk coding, pekerjaan dunia nyata, tugas jangka panjang, dan agen multimodal." + }, + "ja": { + "description": "総2.4兆・アクティブ950億パラメータを備え、コーディング、実務、長期タスク、マルチモーダルエージェントに対応するAlibabaのフラッグシップモデル。" + }, + "ko": { + "description": "총 2.4조 및 활성 950억 매개변수로 코딩, 실제 업무, 장기 작업과 멀티모달 에이전트를 지원하는 Alibaba의 플래그십 모델입니다." + }, + "pt": { + "description": "O modelo principal da Alibaba com 2,4 trilhões de parâmetros e 95 bilhões ativos para programação, trabalho real, tarefas prolongadas e agentes multimodais." + }, + "ru": { + "description": "Флагманская модель Alibaba с 2,4 трлн параметров, из которых 95 млрд активны, для программирования, реальных рабочих задач, длительной автономной работы и мультимодальных агентов." + }, + "tr": { + "description": "Alibaba'nın kodlama, gerçek iş görevleri, uzun süreli çalışmalar ve çok modlu ajanlar için 2,4 trilyon toplam ve 95 milyar aktif parametreli amiral modeli." + }, + "zh-Hans": { + "description": "阿里巴巴拥有 2.4 万亿总参数、950 亿激活参数的旗舰模型，面向编码、真实工作、长周期任务和多模态智能体。" + }, + "zh-Hant": { + "description": "阿里巴巴擁有 2.4 兆總參數、950 億啟用參數的旗艦模型，面向程式設計、真實工作、長週期任務和多模態智慧體。" + } + }, + "verified": true, + "sources": [ + { + "url": "https://qwen.ai/blog?id=qwen3.8", + "title": "Qwen3.8-Max: A New Bar for Coding and Cowork", + "fields": [ + "name", + "description", + "websiteUrl", + "docsUrl", + "size", + "activeParameters", + "contextWindow", + "maxOutput", + "releaseDate", + "lifecycle", + "inputModalities", + "outputModalities", + "capabilities" + ] + }, + { + "url": "https://artificialanalysis.ai/models/qwen3-8-max", + "title": "Artificial Analysis Qwen3.8 Max model page", + "fields": ["platformUrls.artificialAnalysis"] + } + ], + "lastVerifiedAt": "2026-08-13", + "verifiedBy": "codex-agent", + "confidence": "high", + "websiteUrl": "https://qwen.ai", + "docsUrl": "https://qwen.ai/blog?id=qwen3.8", + "vendor": "Alibaba", + "size": "2.4T", + "activeParameters": "95B", + "contextWindow": 1000000, + "maxOutput": 65536, + "tokenPricing": { + "status": "unavailable", + "reason": "official-price-not-published", + "primaryOffer": null, + "offers": [] + }, + "releaseDate": "2026-08-03", + "lifecycle": "latest", + "knowledgeCutoff": null, + "inputModalities": ["text", "image", "pdf", "video"], + "outputModalities": ["text"], + "capabilities": ["function-calling", "tool-choice", "reasoning"], + "benchmarks": { + "sweBench": null, + "terminalBench": null, + "mmmu": null, + "mmmuPro": null, + "webDevArena": null, + "sciCode": null, + "liveCodeBench": null + }, + "platformUrls": { + "huggingface": null, + "artificialAnalysis": "https://artificialanalysis.ai/models/qwen3-8-max", + "openrouter": null + } +} diff --git a/manifests/vendors/alibaba.json b/manifests/vendors/alibaba.json index 861dcb98..fa9adaa8 100644 --- a/manifests/vendors/alibaba.json +++ b/manifests/vendors/alibaba.json @@ -10,7 +10,7 @@ { "id": "qwen-max", "name": "Qwen Max", - "modelIds": ["qwen3-6-max-preview", "qwen3-7-max"] + "modelIds": ["qwen3-6-max-preview", "qwen3-7-max", "qwen3-8-max"] }, { "id": "qwen-plus", @@ -91,9 +91,14 @@ "url": "https://x.com/AlibabaGroup", "title": "Alibaba twitter community", "fields": ["communityUrls.twitter"] + }, + { + "url": "https://qwen.ai/blog?id=qwen3.8", + "title": "Qwen3.8-Max: A New Bar for Coding and Cowork", + "fields": ["modelSeries"] } ], - "lastVerifiedAt": "2026-07-28", + "lastVerifiedAt": "2026-08-13", "verifiedBy": "codex-agent", "confidence": "high", "websiteUrl": "https://www.alibabagroup.com", diff --git a/manifests/vendors/deepseek.json b/manifests/vendors/deepseek.json index 4c6a015a..7299ea92 100644 --- a/manifests/vendors/deepseek.json +++ b/manifests/vendors/deepseek.json @@ -18,6 +18,7 @@ "deepseek-v3-terminus", "deepseek-v3-2-exp", "deepseek-3-2", + "deepseek-v4-pro-preview", "deepseek-v4-pro" ] }, @@ -89,9 +90,14 @@ "url": "https://www.reddit.com/r/DeepSeekAI", "title": "DeepSeek reddit community", "fields": ["communityUrls.reddit"] + }, + { + "url": "https://api-docs.deepseek.com/quick_start/pricing/", + "title": "DeepSeek V4 Pro 0813 models and pricing", + "fields": ["modelSeries"] } ], - "lastVerifiedAt": "2026-07-28", + "lastVerifiedAt": "2026-08-13", "verifiedBy": "codex-agent", "confidence": "high", "websiteUrl": "https://www.deepseek.com", diff --git a/manifests/vendors/meta.json b/manifests/vendors/meta.json index 23529dc2..bd04145c 100644 --- a/manifests/vendors/meta.json +++ b/manifests/vendors/meta.json @@ -15,7 +15,7 @@ { "id": "muse", "name": "Muse", - "modelIds": ["muse-spark-1-1"] + "modelIds": ["muse-spark-1-1", "muse-spark-1-2"] } ], "description": "Technology company building AI tools and open-source models like Llama, advancing research through open science to connect people and enable innovation worldwide.", @@ -85,9 +85,14 @@ "url": "https://ai.meta.com/blog", "title": "Meta blog community", "fields": ["communityUrls.blog"] + }, + { + "url": "https://research.meta.ai/blog/introducing-muse-code-and-muse-spark-1-2", + "title": "Introducing Muse Code and Muse Spark 1.2", + "fields": ["modelSeries"] } ], - "lastVerifiedAt": "2026-07-28", + "lastVerifiedAt": "2026-08-11", "verifiedBy": "codex-agent", "confidence": "high", "websiteUrl": "https://ai.meta.com", diff --git a/manifests/vendors/xai.json b/manifests/vendors/xai.json index 5a8142ab..5685bff0 100644 --- a/manifests/vendors/xai.json +++ b/manifests/vendors/xai.json @@ -10,7 +10,7 @@ { "id": "grok", "name": "Grok", - "modelIds": ["grok-4", "grok-4-20", "grok-4-3", "grok-4-5"] + "modelIds": ["grok-4", "grok-4-20", "grok-4-3", "grok-4-5", "grok-4-6"] }, { "id": "grok-code", @@ -90,9 +90,14 @@ "url": "https://x.ai/news", "title": "xAI blog community", "fields": ["communityUrls.blog"] + }, + { + "url": "https://x.ai/news/grok-4-6", + "title": "Introducing Grok 4.6", + "fields": ["modelSeries"] } ], - "lastVerifiedAt": "2026-07-28", + "lastVerifiedAt": "2026-08-13", "verifiedBy": "codex-agent", "confidence": "high", "websiteUrl": "https://x.ai", diff --git a/scripts/validate/i18n-exact-english-baseline.json b/scripts/validate/i18n-exact-english-baseline.json index 5e8b16ed..7a257c09 100644 --- a/scripts/validate/i18n-exact-english-baseline.json +++ b/scripts/validate/i18n-exact-english-baseline.json @@ -56,6 +56,7 @@ "manifest:de:manifests/collections.json#features.sections[0].items[0].translations.de.name": "Agent Skills", "manifest:de:manifests/collections.json#features.sections[0].items[1].translations.de.name": "Model Context Protocol (MCP)", "manifest:de:manifests/collections.json#features.sections[0].items[2].translations.de.name": "AGENTS.md", + "manifest:de:manifests/collections.json#features.sections[0].items[3].translations.de.name": "Agent Plugins", "manifest:de:manifests/collections.json#features.sections[1].items[0].translations.de.name": "Slash Commands", "manifest:de:manifests/collections.json#specifications.sections[0].items[0].translations.de.name": "Conventional Commits", "manifest:de:manifests/collections.json#specifications.sections[0].items[1].translations.de.name": "Semantic Versioning", @@ -78,6 +79,7 @@ "manifest:de:manifests/collections.json#tools.sections[1].items[5].translations.de.name": "Google Code Wiki", "manifest:es:manifests/collections.json#features.sections[0].items[1].translations.es.name": "Model Context Protocol (MCP)", "manifest:es:manifests/collections.json#features.sections[0].items[2].translations.es.name": "AGENTS.md", + "manifest:es:manifests/collections.json#features.sections[0].items[3].translations.es.name": "Agent Plugins", "manifest:es:manifests/collections.json#specifications.sections[0].items[0].translations.es.name": "Conventional Commits", "manifest:es:manifests/collections.json#specifications.sections[0].items[1].translations.es.name": "Semantic Versioning", "manifest:es:manifests/collections.json#specifications.sections[1].items[0].translations.es.name": "Agent2Agent (A2A)", @@ -99,6 +101,7 @@ "manifest:es:manifests/collections.json#tools.sections[1].items[5].translations.es.name": "Google Code Wiki", "manifest:fr:manifests/collections.json#features.sections[0].items[1].translations.fr.name": "Model Context Protocol (MCP)", "manifest:fr:manifests/collections.json#features.sections[0].items[2].translations.fr.name": "AGENTS.md", + "manifest:fr:manifests/collections.json#features.sections[0].items[3].translations.fr.name": "Agent Plugins", "manifest:fr:manifests/collections.json#specifications.sections[0].items[0].translations.fr.name": "Conventional Commits", "manifest:fr:manifests/collections.json#specifications.sections[0].items[1].translations.fr.name": "Semantic Versioning", "manifest:fr:manifests/collections.json#specifications.sections[1].items[0].translations.fr.name": "Agent2Agent (A2A)", @@ -121,6 +124,7 @@ "manifest:id:manifests/collections.json#articles.sections[3].translations.id.title": "Lovable Engineering", "manifest:id:manifests/collections.json#features.sections[0].items[1].translations.id.name": "Model Context Protocol (MCP)", "manifest:id:manifests/collections.json#features.sections[0].items[2].translations.id.name": "AGENTS.md", + "manifest:id:manifests/collections.json#features.sections[0].items[3].translations.id.name": "Agent Plugins", "manifest:id:manifests/collections.json#features.sections[1].items[0].translations.id.name": "Slash Commands", "manifest:id:manifests/collections.json#specifications.sections[0].items[0].translations.id.name": "Conventional Commits", "manifest:id:manifests/collections.json#specifications.sections[0].items[1].translations.id.name": "Semantic Versioning", @@ -144,6 +148,7 @@ "manifest:ja:manifests/collections.json#articles.sections[3].translations.ja.title": "Lovable Engineering", "manifest:ja:manifests/collections.json#features.sections[0].items[1].translations.ja.name": "Model Context Protocol (MCP)", "manifest:ja:manifests/collections.json#features.sections[0].items[2].translations.ja.name": "AGENTS.md", + "manifest:ja:manifests/collections.json#features.sections[0].items[3].translations.ja.name": "Agent Plugins", "manifest:ja:manifests/collections.json#specifications.sections[0].items[0].translations.ja.name": "Conventional Commits", "manifest:ja:manifests/collections.json#specifications.sections[0].items[1].translations.ja.name": "Semantic Versioning", "manifest:ja:manifests/collections.json#specifications.sections[1].items[0].translations.ja.name": "Agent2Agent (A2A)", @@ -166,6 +171,7 @@ "manifest:ko:manifests/collections.json#articles.sections[3].translations.ko.title": "Lovable Engineering", "manifest:ko:manifests/collections.json#features.sections[0].items[1].translations.ko.name": "Model Context Protocol (MCP)", "manifest:ko:manifests/collections.json#features.sections[0].items[2].translations.ko.name": "AGENTS.md", + "manifest:ko:manifests/collections.json#features.sections[0].items[3].translations.ko.name": "Agent Plugins", "manifest:ko:manifests/collections.json#specifications.sections[0].items[0].translations.ko.name": "Conventional Commits", "manifest:ko:manifests/collections.json#specifications.sections[0].items[1].translations.ko.name": "Semantic Versioning", "manifest:ko:manifests/collections.json#specifications.sections[1].items[0].translations.ko.name": "Agent2Agent (A2A)", @@ -188,6 +194,7 @@ "manifest:pt:manifests/collections.json#articles.sections[3].translations.pt.title": "Lovable Engineering", "manifest:pt:manifests/collections.json#features.sections[0].items[1].translations.pt.name": "Model Context Protocol (MCP)", "manifest:pt:manifests/collections.json#features.sections[0].items[2].translations.pt.name": "AGENTS.md", + "manifest:pt:manifests/collections.json#features.sections[0].items[3].translations.pt.name": "Agent Plugins", "manifest:pt:manifests/collections.json#specifications.sections[0].items[0].translations.pt.name": "Conventional Commits", "manifest:pt:manifests/collections.json#specifications.sections[0].items[1].translations.pt.name": "Semantic Versioning", "manifest:pt:manifests/collections.json#specifications.sections[1].items[0].translations.pt.name": "Agent2Agent (A2A)", @@ -210,6 +217,7 @@ "manifest:ru:manifests/collections.json#articles.sections[3].translations.ru.title": "Lovable Engineering", "manifest:ru:manifests/collections.json#features.sections[0].items[1].translations.ru.name": "Model Context Protocol (MCP)", "manifest:ru:manifests/collections.json#features.sections[0].items[2].translations.ru.name": "AGENTS.md", + "manifest:ru:manifests/collections.json#features.sections[0].items[3].translations.ru.name": "Agent Plugins", "manifest:ru:manifests/collections.json#specifications.sections[0].items[0].translations.ru.name": "Conventional Commits", "manifest:ru:manifests/collections.json#specifications.sections[0].items[1].translations.ru.name": "Semantic Versioning", "manifest:ru:manifests/collections.json#specifications.sections[1].items[0].translations.ru.name": "Agent2Agent (A2A)", @@ -232,6 +240,7 @@ "manifest:tr:manifests/collections.json#articles.sections[3].translations.tr.title": "Lovable Engineering", "manifest:tr:manifests/collections.json#features.sections[0].items[1].translations.tr.name": "Model Context Protocol (MCP)", "manifest:tr:manifests/collections.json#features.sections[0].items[2].translations.tr.name": "AGENTS.md", + "manifest:tr:manifests/collections.json#features.sections[0].items[3].translations.tr.name": "Agent Plugins", "manifest:tr:manifests/collections.json#specifications.sections[0].items[0].translations.tr.name": "Conventional Commits", "manifest:tr:manifests/collections.json#specifications.sections[0].items[1].translations.tr.name": "Semantic Versioning", "manifest:tr:manifests/collections.json#specifications.sections[1].items[0].translations.tr.name": "Agent2Agent (A2A)", @@ -252,6 +261,7 @@ "manifest:tr:manifests/collections.json#tools.sections[1].items[4].translations.tr.name": "CLIProxyAPI", "manifest:tr:manifests/collections.json#tools.sections[1].items[5].translations.tr.name": "Google Code Wiki", "manifest:zh-Hans:manifests/collections.json#features.sections[0].items[2].translations.zh-Hans.name": "AGENTS.md", + "manifest:zh-Hans:manifests/collections.json#features.sections[0].items[3].translations.zh-Hans.name": "Agent Plugins", "manifest:zh-Hans:manifests/collections.json#specifications.sections[2].items[1].translations.zh-Hans.name": "Open Responses", "manifest:zh-Hans:manifests/collections.json#tools.sections[0].items[0].translations.zh-Hans.name": "OpenSpec", "manifest:zh-Hans:manifests/collections.json#tools.sections[0].items[1].translations.zh-Hans.name": "Spec Kit", @@ -265,6 +275,7 @@ "manifest:zh-Hans:manifests/collections.json#tools.sections[1].items[4].translations.zh-Hans.name": "CLIProxyAPI", "manifest:zh-Hans:manifests/collections.json#tools.sections[1].items[5].translations.zh-Hans.name": "Google Code Wiki", "manifest:zh-Hant:manifests/collections.json#features.sections[0].items[2].translations.zh-Hant.name": "AGENTS.md", + "manifest:zh-Hant:manifests/collections.json#features.sections[0].items[3].translations.zh-Hant.name": "Agent Plugins", "manifest:zh-Hant:manifests/collections.json#specifications.sections[2].items[1].translations.zh-Hant.name": "Open Responses", "manifest:zh-Hant:manifests/collections.json#tools.sections[0].items[0].translations.zh-Hant.name": "OpenSpec", "manifest:zh-Hant:manifests/collections.json#tools.sections[0].items[1].translations.zh-Hant.name": "Spec Kit", diff --git a/src/lib/generated/clis.ts b/src/lib/generated/clis.ts index 1812d05b..b9447453 100644 --- a/src/lib/generated/clis.ts +++ b/src/lib/generated/clis.ts @@ -25,6 +25,7 @@ import KiloCodeCli from '../../../manifests/clis/kilo-code-cli.json' import KimiCli from '../../../manifests/clis/kimi-cli.json' import KiroCli from '../../../manifests/clis/kiro-cli.json' import Kode from '../../../manifests/clis/kode.json' +import MuseCode from '../../../manifests/clis/muse-code.json' import Omp from '../../../manifests/clis/omp.json' import Opencode from '../../../manifests/clis/opencode.json' import Pi from '../../../manifests/clis/pi.json' @@ -55,6 +56,7 @@ export const clisData = [ KimiCli, KiroCli, Kode, + MuseCode, Omp, Opencode, Pi, diff --git a/src/lib/generated/metadata.ts b/src/lib/generated/metadata.ts index 3b7b3af1..1d033182 100644 --- a/src/lib/generated/metadata.ts +++ b/src/lib/generated/metadata.ts @@ -3044,6 +3044,76 @@ export const collectionsMetadata: Record = { description: 'Open-format markdown file providing coding agents with project-specific instructions and context', }, + { + translations: { + en: { + name: 'Agent Plugins', + description: + 'Open, vendor-neutral v1.0.0 working-draft standard for packaging Agent Skills and MCP servers as portable plugins with a shared manifest and directory layout', + }, + de: { + name: 'Agent Plugins', + description: + 'Offener, herstellerneutraler Arbeitsentwurf v1.0.0 zum Verpacken von Agent Skills und MCP-Servern als portable Plugins mit gemeinsamem Manifest und Verzeichnislayout', + }, + es: { + name: 'Agent Plugins', + description: + 'Borrador abierto y neutral respecto a proveedores v1.0.0 para empaquetar Agent Skills y servidores MCP como plugins portátiles con un manifiesto y una estructura de directorios comunes', + }, + fr: { + name: 'Agent Plugins', + description: + 'Projet de norme ouvert et neutre vis-à-vis des fournisseurs v1.0.0 pour regrouper Agent Skills et serveurs MCP dans des plugins portables avec un manifeste et une arborescence communs', + }, + id: { + name: 'Agent Plugins', + description: + 'Draf kerja v1.0.0 yang terbuka dan netral vendor untuk mengemas Agent Skills dan server MCP sebagai plugin portabel dengan manifes dan tata letak direktori bersama', + }, + ja: { + name: 'Agent Plugins', + description: + 'Agent SkillsとMCPサーバーを共通のマニフェストとディレクトリ構成を持つポータブルなプラグインとしてパッケージ化する、オープンでベンダー中立なv1.0.0作業草案', + }, + ko: { + name: 'Agent Plugins', + description: + 'Agent Skills와 MCP 서버를 공통 매니페스트 및 디렉터리 구조를 갖춘 이식 가능한 플러그인으로 패키징하기 위한 개방형 벤더 중립 v1.0.0 작업 초안', + }, + pt: { + name: 'Agent Plugins', + description: + 'Rascunho aberto e neutro em relação a fornecedores v1.0.0 para empacotar Agent Skills e servidores MCP como plugins portáteis com manifesto e estrutura de diretórios compartilhados', + }, + ru: { + name: 'Agent Plugins', + description: + 'Открытый независимый от поставщиков рабочий проект v1.0.0 для упаковки Agent Skills и MCP-серверов в переносимые плагины с общим манифестом и структурой каталогов', + }, + tr: { + name: 'Agent Plugins', + description: + 'Agent Skills ve MCP sunucularını ortak bir bildirim ve dizin düzenine sahip taşınabilir eklentiler olarak paketlemek için açık, sağlayıcıdan bağımsız v1.0.0 çalışma taslağı', + }, + 'zh-Hans': { + name: 'Agent Plugins', + description: + '开放、厂商中立的 v1.0.0 工作草案，通过统一清单和目录结构将 Agent Skills 与 MCP 服务器打包为可移植插件', + }, + 'zh-Hant': { + name: 'Agent Plugins', + description: + '開放、廠商中立的 v1.0.0 工作草案，透過統一清單與目錄結構將 Agent Skills 和 MCP 伺服器封裝為可攜式外掛程式', + }, + }, + id: 'agent-plugins', + name: 'Agent Plugins', + url: 'https://agent-plugins.org', + description: + 'Open, vendor-neutral v1.0.0 working-draft standard for packaging Agent Skills and MCP servers as portable plugins with a shared manifest and directory layout', + lastVerifiedAt: '2026-08-07', + }, ], }, { @@ -3566,10 +3636,10 @@ export const faqMetadata: Record = { export const stackCounts: Record = { ides: 9, - clis: 27, + clis: 28, desktops: 12, extensions: 18, - models: 123, + models: 127, 'model-providers': 17, vendors: 47, } diff --git a/src/lib/generated/models.ts b/src/lib/generated/models.ts index f62467a8..aceed5e7 100644 --- a/src/lib/generated/models.ts +++ b/src/lib/generated/models.ts @@ -36,6 +36,7 @@ import DeepseekV3Terminus from '../../../manifests/models/deepseek-v3-terminus.j import DeepseekV4Flash from '../../../manifests/models/deepseek-v4-flash.json' import DeepseekV4FlashPreview from '../../../manifests/models/deepseek-v4-flash-preview.json' import DeepseekV4Pro from '../../../manifests/models/deepseek-v4-pro.json' +import DeepseekV4ProPreview from '../../../manifests/models/deepseek-v4-pro-preview.json' import Devstral2 from '../../../manifests/models/devstral-2.json' import DevstralSmall2 from '../../../manifests/models/devstral-small-2.json' import Gemini20Flash from '../../../manifests/models/gemini-2-0-flash.json' @@ -88,6 +89,7 @@ import Grok4 from '../../../manifests/models/grok-4.json' import Grok41Fast from '../../../manifests/models/grok-4-1-fast.json' import Grok43 from '../../../manifests/models/grok-4-3.json' import Grok45 from '../../../manifests/models/grok-4-5.json' +import Grok46 from '../../../manifests/models/grok-4-6.json' import Grok420 from '../../../manifests/models/grok-4-20.json' import Grok4Fast from '../../../manifests/models/grok-4-fast.json' import GrokCodeFast1 from '../../../manifests/models/grok-code-fast-1.json' @@ -112,6 +114,7 @@ import MinimaxM3 from '../../../manifests/models/minimax-m3.json' import MistralMedium35 from '../../../manifests/models/mistral-medium-3-5.json' import MistralSmall4 from '../../../manifests/models/mistral-small-4.json' import MuseSpark11 from '../../../manifests/models/muse-spark-1-1.json' +import MuseSpark12 from '../../../manifests/models/muse-spark-1-2.json' import O3 from '../../../manifests/models/o3.json' import O3Mini from '../../../manifests/models/o3-mini.json' import O4Mini from '../../../manifests/models/o4-mini.json' @@ -124,6 +127,7 @@ import Qwen36MaxPreview from '../../../manifests/models/qwen3-6-max-preview.json import Qwen36Plus from '../../../manifests/models/qwen3-6-plus.json' import Qwen37Max from '../../../manifests/models/qwen3-7-max.json' import Qwen37Plus from '../../../manifests/models/qwen3-7-plus.json' +import Qwen38Max from '../../../manifests/models/qwen3-8-max.json' import Qwen3Coder30bA3b from '../../../manifests/models/qwen3-coder-30b-a3b.json' import Qwen3Coder480bA35b from '../../../manifests/models/qwen3-coder-480b-a35b.json' import Qwen3CoderNext from '../../../manifests/models/qwen3-coder-next.json' @@ -161,6 +165,7 @@ export const modelsData = [ DeepseekV3, DeepseekV4FlashPreview, DeepseekV4Flash, + DeepseekV4ProPreview, DeepseekV4Pro, Devstral2, DevstralSmall2, @@ -213,6 +218,7 @@ export const modelsData = [ Grok41Fast, Grok43, Grok45, + Grok46, Grok420, Grok4Fast, Grok4, @@ -238,6 +244,7 @@ export const modelsData = [ MistralMedium35, MistralSmall4, MuseSpark11, + MuseSpark12, O3Mini, O3, O4Mini, @@ -250,6 +257,7 @@ export const modelsData = [ Qwen36Plus, Qwen37Max, Qwen37Plus, + Qwen38Max, Qwen3Coder30bA3b, Qwen3Coder480bA35b, Qwen3CoderNext, diff --git a/tests/model-intelligence-index.test.ts b/tests/model-intelligence-index.test.ts index 57e57ad7..2ccdefe5 100644 --- a/tests/model-intelligence-index.test.ts +++ b/tests/model-intelligence-index.test.ts @@ -189,6 +189,31 @@ describe('model intelligence index', () => { .filter(point => point.modelId === 'qwen3-6-27b' || point.modelId === 'qwen3-6-35b-a3b') .map(point => point.series) ).toEqual(['Qwen Open', 'Qwen Open']) + expect( + qwenSeries[0]?.points.map(point => [ + point.modelId, + point.score, + point.estimated, + point.configuration, + ]) + ).toEqual([ + ['qwen3-6-max-preview', 41, true, 'Qwen3.6 Max Preview'], + ['qwen3-7-max', 47, false, 'Qwen3.7 Max'], + ['qwen3-8-max', 58, false, 'Qwen3.8 Max'], + ]) + }) + + it('connects Grok 4.6 to the flagship Grok line', () => { + const grokSeries = modelIntelligenceSeries.find( + series => series.vendor === 'xAI' && series.name === 'Grok' + ) + + expect(grokSeries?.points.at(-1)).toMatchObject({ + modelId: 'grok-4-6', + score: 61, + estimated: false, + configuration: 'Grok 4.6 (high)', + }) }) it('orders Claude product lines from Opus through Fable', () => { @@ -239,8 +264,13 @@ describe('model intelligence index', () => { 'deepseek-v3-terminus', 'deepseek-v3-2-exp', 'deepseek-3-2', + 'deepseek-v4-pro-preview', 'deepseek-v4-pro', ]) + expect(deepSeekSeries[0]?.points.slice(-2).map(point => [point.modelId, point.score])).toEqual([ + ['deepseek-v4-pro-preview', 45], + ['deepseek-v4-pro', 53], + ]) expect( deepSeekSeries[1]?.points.map(point => [ point.modelId, @@ -249,8 +279,8 @@ describe('model intelligence index', () => { point.configuration, ]) ).toEqual([ - ['deepseek-v4-flash-preview', 40, false, 'DeepSeek V4 Flash (max)'], - ['deepseek-v4-flash', 50, false, 'DeepSeek V4 Flash 0731 (max)'], + ['deepseek-v4-flash-preview', 42, false, 'DeepSeek V4 Flash (max)'], + ['deepseek-v4-flash', 52, false, 'DeepSeek V4 Flash 0731 (max)'], ]) }) @@ -467,7 +497,7 @@ describe('model intelligence index', () => { expect(modelIntelligenceMeta.methodologyUrl).toBe( 'https://artificialanalysis.ai/methodology/intelligence-benchmarking' ) - expect(modelIntelligenceMeta.indexVersion).toBe('4.1') + expect(modelIntelligenceMeta.indexVersion).toBe('4.1.1') expect(modelIntelligenceMeta.observedAt).toMatch(/^\d{4}-\d{2}-\d{2}$/) }) })