[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"portal-settings:stajic:fr":3,"public-menus:all":38,"post:why-does-ai-retrieve-some-sources-and-ignore-others-the-retrieval-authority-layer:fr":221,"related:post:why-does-ai-retrieve-some-sources-and-ignore-others-the-retrieval-authority-layer:fr:1":1329},{"statusCode":4,"data":5,"message":37},200,{"tenantId":6,"lang":7,"defaultLang":8,"siteUrl":9,"contactEmail":10,"brandName":11,"logoUrl":12,"siteName":11,"siteDescription":13,"ogImage":10,"robotsIndex":14,"socialLinks":10,"reservedSlugs":10,"seoPolicy":15},"stajic","fr","de","https:\u002F\u002Fstajic.de",null,"Stajic Platform","\u002FLogo_Planet.svg","Stajic Portal",true,{"branding":16,"relatedContent":17,"crossDomainLinks":18},{"logoUrl":12},{"enabled":14},[19,22,25,28,31,34],{"url":20,"label":21,"isActive":14,"showInFooter":14,"includeInSameAs":14},"https:\u002F\u002Ffigure.rocks","figure.rocks",{"url":23,"label":24,"isActive":14,"showInFooter":14,"includeInSameAs":14},"https:\u002F\u002Floving.rocks","loving.rocks",{"url":26,"label":27,"isActive":14,"showInFooter":14,"includeInSameAs":14},"https:\u002F\u002Fbazify.com","bazify.com",{"url":29,"label":30,"isActive":14,"showInFooter":14,"includeInSameAs":14},"https:\u002F\u002Fbazify.de","bazify.de",{"url":32,"label":33,"isActive":14,"showInFooter":14,"includeInSameAs":14},"https:\u002F\u002Fbazify.at","bazify.at",{"url":35,"label":36,"isActive":14,"showInFooter":14,"includeInSameAs":14},"https:\u002F\u002Fbazify.ba","bazify.ba","Portal settings resolved",[39,45],{"id":40,"name":41,"location":42,"isActive":14,"isDefault":43,"items":44},1,"main-navigation","header",false,[],{"id":46,"name":47,"location":48,"isActive":14,"isDefault":14,"items":49},4,"main-menu","sidebar",[50,66,79,93,103,118,133],{"id":51,"title":52,"url":60,"target":61,"icon":62,"isActive":14,"type":63,"productId":10,"categoryId":10,"shopCategoryId":10,"articleId":10,"pageId":64,"portfolioId":10,"children":65},"item-18",{"de":53,"en":54,"es":55,"fr":56,"it":54,"ru":57,"sr":58,"zh":59},"Startseite","Home","Inicio","Accueil","Главная","Почетна","首页","\u002Ffull-stack-web-developer-munich-performance-seo-and-maintainable-builds","_self","i-lucide-home","page",111,[],{"id":67,"title":68,"url":75,"target":61,"icon":76,"isActive":14,"type":63,"productId":10,"categoryId":10,"shopCategoryId":10,"articleId":10,"pageId":77,"portfolioId":10,"children":78},"item-22",{"de":69,"en":69,"es":70,"fr":69,"it":71,"ru":72,"sr":73,"zh":74},"Vision","Visión","Visione","Видение","Визија","想象","\u002Fueber-uns-webdesign-muenchen-webaplikation","i-lucide-eye",113,[],{"id":80,"title":81,"url":89,"target":61,"icon":90,"isActive":14,"type":63,"productId":10,"categoryId":10,"shopCategoryId":10,"articleId":10,"pageId":91,"portfolioId":10,"children":92},"item-19",{"de":82,"en":83,"es":84,"fr":83,"it":85,"ru":86,"sr":87,"zh":88},"Leistungen","Services","Servicios","Servizi","Услуги","Услуге","服务","\u002Fservices-dienstleistungen-muenchen","i-lucide-wrench",116,[],{"id":94,"title":95,"url":99,"target":61,"icon":100,"isActive":14,"type":63,"productId":10,"categoryId":10,"shopCategoryId":10,"articleId":10,"pageId":101,"portfolioId":10,"children":102},"item-23",{"de":96,"en":96,"es":96,"fr":96,"it":96,"ru":97,"sr":97,"zh":98},"Blog","Блог","博客","\u002Fblog","i-lucide-book-open",112,[],{"id":104,"title":105,"url":114,"target":61,"icon":115,"isActive":14,"type":63,"productId":10,"categoryId":10,"shopCategoryId":10,"articleId":10,"pageId":116,"portfolioId":10,"children":117},"item-32",{"de":106,"en":107,"es":108,"fr":109,"it":110,"ru":111,"sr":112,"zh":113},"Neue Technologien","New Technologies","Nuevas tecnologías","Nouvelles technologies","Nuove tecnologie","Новые технологии","Нове технологије","新技术！","\u002Fneue-webtechnologien","i-lucide-sparkles",122,[],{"id":119,"title":120,"url":129,"target":61,"icon":130,"isActive":14,"type":63,"productId":10,"categoryId":10,"shopCategoryId":10,"articleId":10,"pageId":131,"portfolioId":10,"children":132},"item-20",{"de":121,"en":122,"es":123,"fr":124,"it":125,"ru":126,"sr":127,"zh":128},"Kontakt","Contact us!","Contacto","Contact","Contatto","Контакт","Контактирајте нас","联系我们！","\u002Fcontact","i-lucide-mail",115,[],{"id":134,"title":135,"url":144,"target":61,"icon":145,"isActive":14,"type":63,"productId":10,"categoryId":10,"shopCategoryId":10,"articleId":10,"pageId":146,"portfolioId":10,"children":147},"item-21",{"de":136,"en":137,"es":138,"fr":139,"it":140,"ru":141,"sr":142,"zh":143},"Unsere Arbeit","Our Work","Nuestro trabajo","Nos réalisations","I nostri lavori","Наши работы","Наши радови","文件夹","\u002Fportfolio","i-lucide-briefcase",114,[148,161,175,191,206],{"id":149,"title":150,"url":144,"target":61,"icon":159,"isActive":14,"type":63,"productId":10,"categoryId":10,"shopCategoryId":10,"articleId":10,"pageId":146,"portfolioId":10,"children":160},"item-24",{"de":151,"en":152,"es":153,"fr":154,"it":155,"ru":156,"sr":157,"zh":158},"Alle Projekte","All Projects","Todos los proyectos","Tous les projets","Tutti i progetti","Все проекты","Сви пројекти","所有项目","i-lucide-grid-3x3",[],{"id":162,"title":163,"url":171,"target":61,"icon":172,"isActive":14,"type":173,"productId":10,"categoryId":10,"shopCategoryId":10,"articleId":10,"pageId":10,"portfolioId":10,"children":174},"item-29",{"de":164,"en":165,"es":166,"fr":167,"it":168,"ru":169,"sr":170,"zh":143},"Local Roots, Global Reach","Local Roots - Global Reach","Empresa local ","Entreprise locale","Azienda locale","Местная компания","Локално предузеће глобално тржиште","\u002Fportfolio\u002Flocal-roots-global-reach-communication-media-systems-for-modern-business","i-lucide-globe-2","custom",[],{"id":176,"title":177,"url":186,"target":61,"icon":187,"isActive":14,"type":188,"productId":10,"categoryId":10,"shopCategoryId":10,"articleId":10,"pageId":10,"portfolioId":189,"children":190},"menu-2916d060-8681-482a-af9a-c107a8bc30d8",{"de":178,"en":179,"es":180,"fr":181,"it":182,"ru":183,"sr":184,"zh":185},"Von Rohdaten zu verifizierbarem KI-Kontext: Eine praktische Architektur für agentenbereite Informationen","From Raw Data to Verifiable AI Context:","De datos sin procesar a contexto de IA verificable: una arquitectura práctica para información lista para agentes","Des données brutes au contexte d'IA vérifiable : une architecture pratique pour des informations prêtes pour les agents","Dai dati grezzi a un contesto IA verificabile: Un'architettura pratica per informazioni pronte per gli agenti","От необработанных данных к верифицируемому контексту ИИ: практическая архитектура информации, готовой для агентов","Od sirovih podataka do proverljivog AI konteksta: Praktična arhitektura za informacije spremne za agente","从原始数据到可验证的 AI 上下文：面向智能体就绪信息的实用架构","\u002Fportfolio\u002Ffrom-raw-data-to-verifiable-ai-context-a-practical-architecture-for-agent-ready-information","i-lucide-database","portfolio",22,[],{"id":192,"title":193,"url":202,"target":61,"icon":203,"isActive":14,"type":188,"productId":10,"categoryId":10,"shopCategoryId":10,"articleId":10,"pageId":10,"portfolioId":204,"children":205},"menu-e2061cd7-ff53-49d5-a0e8-3d49090577c0",{"de":194,"en":195,"es":196,"fr":197,"it":198,"ru":199,"sr":200,"zh":201},"Präsenzplattformen — Unternehmenskommunikation","Presence Platforms — Enterprise Communication","Plataformas de presencia — Comunicación empresarial","Plateformes de présence — Communication d'entreprise","Presence Platforms — Comunicazione aziendale","Платформы присутствия — Корпоративные коммуникации","Presence platforme — Poslovna komunikacija","Presence 平台 — 企业通信","\u002Fportfolio\u002Fpresence-platforms-enterprise-communication-that-wins-without-competing","i-lucide-badge-percent",21,[],{"id":207,"title":208,"url":217,"target":61,"icon":218,"isActive":14,"type":188,"productId":10,"categoryId":10,"shopCategoryId":10,"articleId":10,"pageId":10,"portfolioId":219,"children":220},"menu-a781582e-e90d-418e-b8c9-b070bc37c37e",{"de":209,"en":210,"es":211,"fr":212,"it":213,"ru":214,"sr":215,"zh":216},"Maximierung des POS-ROI durch Konsistenz","Maximizing POS ROI Through Consistent","Maximizar el ROI del POS a través de constante","Maximiser le ROI du PDV grâce à la cohérence","Massimizzare il ROI del POS attraverso costante","Максимизация ROI POS за счет последовательного","Maksimizacija POS ROI-ja kroz dosledno","通过持续最大化 POS 投资回报率","\u002Fportfolio\u002Ffrom-vision-to-value-maximizing-pos-roi-through-consistent-ci-creation-and-strategic-branding","i-lucide-briefcase-business",11,[],{"statusCode":4,"data":222,"message":1328},{"id":223,"title":224,"slug":225,"content":226,"contentJson":227,"excerpt":684,"featuredImage":685,"featuredImageAlt":686,"featuredImageCaption":10,"featuredImageTitle":10,"featuredImageCopyright":10,"featuredImageAuthor":10,"featuredImageSourceUrl":10,"featuredImageLicense":10,"featuredImageIsAiGenerated":43,"status":687,"publishedAt":688,"createdAt":689,"updatedAt":690,"seoLocalePaths":691,"categories":700,"author":709,"translations":713},"496","Pourquoi l'IA récupère-t-elle certaines sources et en ignore-t-elle d'autres ? — La couche d'autorité de récupération","why-does-ai-retrieve-some-sources-and-ignore-others-the-retrieval-authority-layer","\u003Ch2 id=\"section-1\">Question\u003C\u002Fh2>\n\u003Cp>Pourquoi un système de recherche par IA choisit-il une source tout en ignorant une autre source qui semble traiter exactement du même sujet ?\u003C\u002Fp>\n\u003Cp>La pertinence sémantique ne constitue qu'une partie de la réponse. Deux documents peuvent être également similaires à une requête tout en différant radicalement en termes d'autorité, de fraîcheur, de provenance, de portée, d'autorisations et d'applicabilité.\u003C\u002Fp>\n\u003Cp>Cet article utilise le terme Couche d'Autorité de Récupération pour désigner la frontière de sélection des sources qui se situe entre la recherche d'informations potentiellement pertinentes et le traitement de ces informations comme des preuves appropriées.\u003C\u002Fp>\n\u003Cblockquote class=\"border-l-4 border-gray-300 pl-4 italic\">Une Couche d'Autorité de Récupération est l'ensemble des règles de sélection des sources et des propriétés des sources qui déterminent quelles informations récupérables méritent de devenir des preuves pour une question particulière.\u003Ccite class=\"block mt-2 text-sm\">— Définition de travail — un concept systémique utilisé dans cet article, et non un terme de recherche standardisé.\u003C\u002Fcite>\u003C\u002Fblockquote>\n\u003Cnav class=\"editorjs-toc\" data-editorjs-toc=\"true\" aria-label=\"Sommaire\">\u003Cstrong class=\"editorjs-toc__title\">Sommaire\u003C\u002Fstrong>\u003Col class=\"editorjs-toc__list editorjs-toc__list--depth-0\">\u003Cli class=\"editorjs-toc__item\">\u003Ca href=\"#section-1\" class=\"editorjs-toc__link\">Question\u003C\u002Fa>\u003C\u002Fli>\u003Cli class=\"editorjs-toc__item\">\u003Ca href=\"#section-7\" class=\"editorjs-toc__link\">Ce que cela signifie vraiment\u003C\u002Fa>\u003C\u002Fli>\u003Cli class=\"editorjs-toc__item\">\u003Ca href=\"#section-13\" class=\"editorjs-toc__link\">Exemple le plus simple\u003C\u002Fa>\u003C\u002Fli>\u003Cli class=\"editorjs-toc__item\">\u003Ca href=\"#section-19\" class=\"editorjs-toc__link\">Là où l'exemple cesse de fonctionner\u003C\u002Fa>\u003C\u002Fli>\u003Cli class=\"editorjs-toc__item\">\u003Ca href=\"#section-23\" class=\"editorjs-toc__link\">Réponse directe\u003C\u002Fa>\u003C\u002Fli>\u003Cli class=\"editorjs-toc__item\">\u003Ca href=\"#section-27\" class=\"editorjs-toc__link\">Pourquoi la similarité seule ne suffit pas\u003C\u002Fa>\u003C\u002Fli>\u003Cli class=\"editorjs-toc__item\">\u003Ca href=\"#section-31\" class=\"editorjs-toc__link\">Les six signaux d'autorité\u003C\u002Fa>\u003Col class=\"editorjs-toc__list editorjs-toc__list--depth-1\">\u003Cli class=\"editorjs-toc__item\">\u003Ca href=\"#section-32\" class=\"editorjs-toc__link\">1. Identité de la source\u003C\u002Fa>\u003C\u002Fli>\u003Cli class=\"editorjs-toc__item\">\u003Ca href=\"#section-34\" class=\"editorjs-toc__link\">2. Provenance\u003C\u002Fa>\u003C\u002Fli>\u003Cli class=\"editorjs-toc__item\">\u003Ca href=\"#section-36\" class=\"editorjs-toc__link\">3. Fraîcheur\u003C\u002Fa>\u003C\u002Fli>\u003Cli class=\"editorjs-toc__item\">\u003Ca href=\"#section-38\" class=\"editorjs-toc__link\">4. Applicabilité\u003C\u002Fa>\u003C\u002Fli>\u003Cli class=\"editorjs-toc__item\">\u003Ca href=\"#section-40\" class=\"editorjs-toc__link\">5. Spécificité des preuves\u003C\u002Fa>\u003C\u002Fli>\u003Cli class=\"editorjs-toc__item\">\u003Ca href=\"#section-42\" class=\"editorjs-toc__link\">6. Légitimité de l'accès\u003C\u002Fa>\u003C\u002Fli>\u003C\u002Fol>\u003C\u002Fli>\u003Cli class=\"editorjs-toc__item\">\u003Ca href=\"#section-45\" class=\"editorjs-toc__link\">Un modèle d'autorité minimal\u003C\u002Fa>\u003C\u002Fli>\u003Cli class=\"editorjs-toc__item\">\u003Ca href=\"#section-49\" class=\"editorjs-toc__link\">Connexion à l'information prête pour les agents\u003C\u002Fa>\u003C\u002Fli>\u003Cli class=\"editorjs-toc__item\">\u003Ca href=\"#section-53\" class=\"editorjs-toc__link\">Méthode de diagnostic \u002F décision\u003C\u002Fa>\u003C\u002Fli>\u003Cli class=\"editorjs-toc__item\">\u003Ca href=\"#section-57\" class=\"editorjs-toc__link\">Modes de défaillance courants\u003C\u002Fa>\u003C\u002Fli>\u003Cli class=\"editorjs-toc__item\">\u003Ca href=\"#section-62\" class=\"editorjs-toc__link\">Cas limite : sources faisant autorité en conflit\u003C\u002Fa>\u003C\u002Fli>\u003Cli class=\"editorjs-toc__item\">\u003Ca href=\"#section-66\" class=\"editorjs-toc__link\">Limites\u003C\u002Fa>\u003C\u002Fli>\u003Cli class=\"editorjs-toc__item\">\u003Ca href=\"#section-70\" class=\"editorjs-toc__link\">Qu'est-ce qui changerait cette réponse ?\u003C\u002Fa>\u003C\u002Fli>\u003Cli class=\"editorjs-toc__item\">\u003Ca href=\"#section-73\" class=\"editorjs-toc__link\">Conclusion\u003C\u002Fa>\u003C\u002Fli>\u003Cli class=\"editorjs-toc__item\">\u003Ca href=\"#section-78\" class=\"editorjs-toc__link\">Sources Primaires\u003C\u002Fa>\u003C\u002Fli>\u003C\u002Fol>\u003C\u002Fnav>\n\u003Ch2 id=\"section-7\">Ce que cela signifie vraiment\u003C\u002Fh2>\n\u003Cp>Un moteur de récupération répond à une question relativement étroite : Quels éléments semblent pertinents pour cette requête ?\u003C\u002Fp>\n\u003Cp>Un système d'IA en production doit généralement répondre à une question plus difficile : Lesquels de ces éléments pertinents devraient être autorisés à influencer la réponse ?\u003C\u002Fp>\n\u003Cp>Cette seconde question introduit l'identité de la source, la version, l'autorité, la période de validité, la provenance et la portée. Une ancienne politique pertinente, un article de blog copié ou un résumé non officiel peut obtenir un très bon classement tout en restant la mauvaise preuve pour la décision.\u003C\u002Fp>\n\u003Cpre class=\"code-block\">\u003Ccode>Question\n   ↓\nRetrieval Trigger\n   ↓\nCandidate Sources\n   ↓\nRETRIEVAL AUTHORITY LAYER\n   ↓\nQualified Evidence\n   ↓\nContext \u002F Reasoning\n   ↓\nAnswer Validity Boundary\n   ↓\nAnswer\u003C\u002Fcode>\u003C\u002Fpre>\n\u003Ca href=\"https:\u002F\u002Fstajic.de\u002Ffr\u002Fblog\u002Fwhen-should-an-ai-stop-trusting-its-own-knowledge-the-retrieval-trigger\" class=\"editorjs-link-tool block border border-gray-200 dark:border-gray-700 rounded-lg p-4 transition text-gray-900 dark:text-gray-100 hover:border-primary-500 hover:bg-primary-50 dark:hover:bg-gray-900 hover:text-gray-900 dark:hover:text-gray-100\">\u003Cstrong class=\"block font-semibold\">Quand une IA devrait-elle cesser de faire confiance à ses propres connaissances ? — Le Déclencheur de Récupération\u003C\u002Fstrong>\u003Cp class=\"text-sm text-gray-600 dark:text-gray-400\">Le Déclencheur de Récupération décide quand des preuves externes sont nécessaires. La Couche d'Autorité de Récupération détermine quelles sources doivent les fournir.\u003C\u002Fp>\u003C\u002Fa>\n\u003Ch2 id=\"section-13\">Exemple le plus simple\u003C\u002Fh2>\n\u003Cp>Imaginez qu'un employé demande à un assistant d'entreprise : Combien de jours de vacances les employés reçoivent-ils ?\u003C\u002Fp>\n\u003Cdiv class=\"overflow-x-auto\">\u003Ctable class=\"w-full border-collapse\">\u003Cthead>\u003Ctr>\u003Cth class=\"border border-gray-300 px-4 py-2 text-left font-semibold\">Candidat\u003C\u002Fth>\u003Cth class=\"border border-gray-300 px-4 py-2 text-left font-semibold\">Contenu\u003C\u002Fth>\u003Cth class=\"border border-gray-300 px-4 py-2 text-left font-semibold\">Pertinence sémantique\u003C\u002Fth>\u003C\u002Ftr>\u003C\u002Fthead>\u003Ctbody>\u003Ctr>\u003Ctd class=\"border border-gray-300 px-4 py-2\">A\u003C\u002Ftd>\u003Ctd class=\"border border-gray-300 px-4 py-2\">Politique RH actuelle, version approuvée 6.2\u003C\u002Ftd>\u003Ctd class=\"border border-gray-300 px-4 py-2\">Élevée\u003C\u002Ftd>\u003C\u002Ftr>\u003Ctr>\u003Ctd class=\"border border-gray-300 px-4 py-2\">B\u003C\u002Ftd>\u003Ctd class=\"border border-gray-300 px-4 py-2\">Ancien guide d'intégration copié de la version 4.8\u003C\u002Ftd>\u003Ctd class=\"border border-gray-300 px-4 py-2\">Élevée\u003C\u002Ftd>\u003C\u002Ftr>\u003C\u002Ftbody>\u003C\u002Ftable>\u003C\u002Fdiv>\n\u003Cp>Les deux passages peuvent contenir une formulation presque identique. Un score de similarité peut donc les placer très proches l'un de l'autre.\u003C\u002Fp>\n\u003Cpre class=\"code-block\">\u003Ccode>Semantic relevance\n        ≠\nSource authority\n        ≠\nApplicability\u003C\u002Fcode>\u003C\u002Fpre>\n\u003Cp>La politique RH actuelle approuvée devrait normalement dominer car elle constitue la source de référence pour l'affirmation. L'ancien guide reste un texte pertinent, mais il ne devrait pas avoir le même poids probatoire.\u003C\u002Fp>\n\u003Ch2 id=\"section-19\">Là où l'exemple cesse de fonctionner\u003C\u002Fh2>\n\u003Cp>L'autorité n'est pas toujours une hiérarchie universelle dans laquelle une source l'emporte à chaque fois.\u003C\u002Fp>\n\u003Cp>Une base de données transactionnelle peut faire autorité pour le statut actuel d'une commande. Un manuel de produit peut faire autorité pour le fonctionnement prévu. Un contrat signé peut régir une obligation commerciale. Un article scientifique primaire peut être la meilleure source pour son expérience, tandis qu'une revue systématique ultérieure peut être plus solide pour une conclusion plus large.\u003C\u002Fp>\n\u003Cp>L'autorité est donc relative à l'affirmation. La classe de source correcte dépend de ce qui est demandé.\u003C\u002Fp>\n\u003Ch2 id=\"section-23\">Réponse directe\u003C\u002Fh2>\n\u003Cp>Une source devrait devenir une preuve de récupération préférée lorsqu'elle n'est pas seulement pertinente, mais aussi suffisamment faisant autorité, actuelle, applicable, traçable et accessible pour l'affirmation spécifique à laquelle elle répond.\u003C\u002Fp>\n\u003Cpre class=\"code-block\">\u003Ccode>Useful retrieval evidence\n=\nrelevance\n+ authority\n+ provenance\n+ freshness\n+ applicability\n+ accessibility\u003C\u002Fcode>\u003C\u002Fpre>\n\u003Cp>Ces propriétés ne garantissent pas la vérité. Elles créent une base plus solide pour décider quelle information récupérée mérite d'entrer dans le processus de raisonnement.\u003C\u002Fp>\n\u003Ch2 id=\"section-27\">Pourquoi la similarité seule ne suffit pas\u003C\u002Fh2>\n\u003Cp>La récupération vectorielle est extrêmement utile car elle peut trouver des informations sémantiquement liées même lorsque la formulation diffère. Mais la similarité sémantique ne sait pas par elle-même si un document est obsolète, remplacé, non officiel, non autorisé ou valide pour une autre juridiction.\u003C\u002Fp>\n\u003Cp>L'architecture RAG originale a montré comment une mémoire externe non paramétrique peut augmenter les connaissances du modèle. Les systèmes de production doivent ajouter des contrôles spécifiques à l'application autour de ce processus de récupération.\u003C\u002Fp>\n\u003Ca href=\"https:\u002F\u002Fstajic.de\u002Ffr\u002Fblog\u002Fwhere-does-an-llm-get-its-data-rag-data-sources-in-python\" class=\"editorjs-link-tool block border border-gray-200 dark:border-gray-700 rounded-lg p-4 transition text-gray-900 dark:text-gray-100 hover:border-primary-500 hover:bg-primary-50 dark:hover:bg-gray-900 hover:text-gray-900 dark:hover:text-gray-100\">\u003Cstrong class=\"block font-semibold\">Où un LLM obtient-il ses données ? Sources de données RAG en Python\u003C\u002Fstrong>\u003Cp class=\"text-sm text-gray-600 dark:text-gray-400\">Les fichiers, bases de données, API, recherches et index vectoriels peuvent tous fournir des connaissances à l'exécution. La méthode de récupération doit correspondre à la source et à la question.\u003C\u002Fp>\u003C\u002Fa>\n\u003Ch2 id=\"section-31\">Les six signaux d'autorité\u003C\u002Fh2>\n\u003Ch3 id=\"section-32\">1. Identité de la source\u003C\u002Fh3>\n\u003Cp>Le système doit savoir ce qu'est réellement la source : politique, contrat, API, enregistrement de base de données, article de recherche, page de documentation, fichier utilisateur ou commentaire tiers.\u003C\u002Fp>\n\u003Ch3 id=\"section-34\">2. Provenance\u003C\u002Fh3>\n\u003Cp>Les preuves doivent préserver leur origine, la manière dont elles sont entrées dans le système et quel objet ou document original peut être inspecté si l'affirmation est contestée.\u003C\u002Fp>\n\u003Ch3 id=\"section-36\">3. Fraîcheur\u003C\u002Fh3>\n\u003Cp>La fraîcheur importe lorsque le fait sous-jacent change. Une réponse API actuelle et un article vieux de cinq ans peuvent être également pertinents pour les mots d'une requête, mais pas pour son exigence temporelle.\u003C\u002Fp>\n\u003Ch3 id=\"section-38\">4. Applicabilité\u003C\u002Fh3>\n\u003Cp>Un fait peut être correct mais applicable à la mauvaise version de produit, pays, locataire, groupe de clients, contrat, version logicielle ou période.\u003C\u002Fp>\n\u003Ch3 id=\"section-40\">5. Spécificité des preuves\u003C\u002Fh3>\n\u003Cp>Une preuve primaire qui soutient directement l'affirmation est généralement plus forte qu'une chaîne de résumés qui ne font que la répéter.\u003C\u002Fp>\n\u003Ch3 id=\"section-42\">6. Légitimité de l'accès\u003C\u002Fh3>\n\u003Cp>Un document techniquement récupérable n'est pas nécessairement un contexte autorisé. L'autorité de récupération doit opérer dans les limites du locataire, des permissions, de la confidentialité et de la minimisation des données.\u003C\u002Fp>\n\u003Ca href=\"https:\u002F\u002Fstajic.de\u002Ffr\u002Fblog\u002Fwhat-is-context-engineering-what-the-model-receives-before-it-answers\" class=\"editorjs-link-tool block border border-gray-200 dark:border-gray-700 rounded-lg p-4 transition text-gray-900 dark:text-gray-100 hover:border-primary-500 hover:bg-primary-50 dark:hover:bg-gray-900 hover:text-gray-900 dark:hover:text-gray-100\">\u003Cstrong class=\"block font-semibold\">Qu'est-ce que l'ingénierie de contexte ? Ce que le modèle reçoit avant de répondre\u003C\u002Fstrong>\u003Cp class=\"text-sm text-gray-600 dark:text-gray-400\">L'assemblage du contexte est aussi une frontière de validité et de sécurité. Les informations récupérées doivent préserver la portée, l'autorité et les contraintes d'accès.\u003C\u002Fp>\u003C\u002Fa>\n\u003Ch2 id=\"section-45\">Un modèle d'autorité minimal\u003C\u002Fh2>\n\u003Cp>Un système pratique peut représenter ces propriétés explicitement plutôt que d'espérer que le classement par similarité les capture.\u003C\u002Fp>\n\u003Cpre class=\"code-block\">\u003Ccode>{\n  "source": "hr-policy-6.2",\n  "sourceType": "approved_policy",\n  "effectiveFrom": "2026-01-01",\n  "supersedes": "hr-policy-5.9",\n  "jurisdiction": "DE",\n  "authority": "primary",\n  "accessScope": "employee",\n  "provenance": "hr-policy-system"\n}\u003C\u002Fcode>\u003C\u002Fpre>\n\u003Cp>Le récupérateur peut toujours utiliser la pertinence sémantique ou lexicale. La couche d'autorité ajoute des contraintes et des signaux de reclassement que la pertinence seule ne peut fournir.\u003C\u002Fp>\n\u003Ch2 id=\"section-49\">Connexion à l'information prête pour les agents\u003C\u002Fh2>\n\u003Cp>Cela devient particulièrement important lorsque le consommateur n'est pas un humain lisant des documents mais un agent IA prenant des décisions.\u003C\u002Fp>\n\u003Cp>Un objet d'information prêt pour les agents devrait préserver l'identité, la provenance, les conflits, les inconnues et les conditions de validité avant l'étape de récupération. Sinon, le système tente de reconstruire l'autorité après que l'information a déjà été aplatie en texte anonyme.\u003C\u002Fp>\n\u003Ca href=\"https:\u002F\u002Fstajic.de\u002Ffr\u002Fportfolio\u002Ffrom-raw-data-to-verifiable-ai-context-a-practical-architecture-for-agent-ready-information\" class=\"editorjs-link-tool block border border-gray-200 dark:border-gray-700 rounded-lg p-4 transition text-gray-900 dark:text-gray-100 hover:border-primary-500 hover:bg-primary-50 dark:hover:bg-gray-900 hover:text-gray-900 dark:hover:text-gray-100\">\u003Cstrong class=\"block font-semibold\">Des données brutes au contexte IA vérifiable : une architecture pratique pour l'information prête pour les agents\u003C\u002Fstrong>\u003Cp class=\"text-sm text-gray-600 dark:text-gray-400\">Une architecture de portefeuille pour transformer des données hétérogènes en objets d'information contrôlés avec identité, provenance, conflits et validité.\u003C\u002Fp>\u003C\u002Fa>\n\u003Ch2 id=\"section-53\">Méthode de diagnostic \u002F décision\u003C\u002Fh2>\n\u003Cp>Pour chaque affirmation récupérée de grande valeur, posez six questions :\u003C\u002Fp>\n\u003Col>\u003Cli>Quelle source exacte a produit cette preuve ?\u003C\u002Fli>\u003Cli>Cette source fait-elle autorité pour ce type d'affirmation ?\u003C\u002Fli>\u003Cli>L'information est-elle toujours d'actualité ?\u003C\u002Fli>\u003Cli>S'applique-t-elle à cet utilisateur, cette version, cette juridiction et cette période ?\u003C\u002Fli>\u003Cli>L'affirmation peut-elle être retracée jusqu'à la preuve originale ?\u003C\u002Fli>\u003Cli>L'utilisateur ou l'agent était-il autorisé à la récupérer ?\u003C\u002Fli>\u003C\u002Fol>\n\u003Cp>Si le système ne peut pas répondre à ces questions, le problème n'est plus seulement la qualité de la récupération. C'est un problème de gouvernance des preuves.\u003C\u002Fp>\n\u003Ch2 id=\"section-57\">Modes de défaillance courants\u003C\u002Fh2>\n\u003Cp>Le score de similarité le plus élevé l'emporte. Cela confond la proximité sémantique avec l'autorité probante.\u003C\u002Fp>\n\u003Cp>La source la plus récente l'emporte toujours. Un commentaire plus récent peut être moins faisant autorité qu'un contrat ou une norme plus ancienne qui fait foi.\u003C\u002Fp>\n\u003Cp>Officiel signifie universellement applicable. Une source officielle peut toujours appartenir à une autre juridiction, version ou portée.\u003C\u002Fp>\n\u003Cp>Public signifie sûr à récupérer. Une information peut être publique mais inappropriée pour la tâche, tandis qu'une information privée peut être légitime uniquement dans un contexte autorisé.\u003C\u002Fp>\n\u003Ch2 id=\"section-62\">Cas limite : sources faisant autorité en conflit\u003C\u002Fh2>\n\u003Cp>Le cas difficile n'est pas une source faisant autorité contre une source manifestement faible. C'est deux sources apparemment faisant autorité qui sont en désaccord.\u003C\u002Fp>\n\u003Cp>À ce stade, la couche d'autorité devrait préserver le conflit au lieu de l'effondrer silencieusement en une seule réponse.\u003C\u002Fp>\n\u003Cpre class=\"code-block\">\u003Ccode>Retrieve\n   ↓\nQualify source authority\n   ↓\nDetect conflict\n   ↓\nPreserve disagreement\n   ↓\nReason about scope \u002F version \u002F provenance\n   ↓\nAnswer only inside the supported boundary\u003C\u002Fcode>\u003C\u002Fpre>\n\u003Ch2 id=\"section-66\">Limites\u003C\u002Fh2>\n\u003Cp>La couche d'autorité de récupération est une architecture conceptuelle, pas un algorithme de classement universel.\u003C\u002Fp>\n\u003Cp>Différents domaines nécessitent différents modèles d'autorité. Les données juridiques, scientifiques, commerciales, opérationnelles et personnelles ne peuvent pas toutes utiliser la même hiérarchie de sources.\u003C\u002Fp>\n\u003Cp>Elle ne peut pas non plus garantir qu'un moteur de recherche externe ou un système d'IA classera la page d'un éditeur. Les systèmes de classement externes restent hors du contrôle de l'éditeur.\u003C\u002Fp>\n\u003Ch2 id=\"section-70\">Qu'est-ce qui changerait cette réponse ?\u003C\u002Fh2>\n\u003Cp>Les systèmes de récupération peuvent de plus en plus apprendre automatiquement la fiabilité des sources, combiner la provenance structurée avec le classement neuronal et négocier l'accès via des protocoles d'agents.\u003C\u002Fp>\n\u003Cp>Cela peut changer la façon dont les signaux d'autorité sont mis en œuvre. Cela ne supprime pas la distinction entre trouver une information pertinente et décider si cette information mérite un poids probant.\u003C\u002Fp>\n\u003Ch2 id=\"section-73\">Conclusion\u003C\u002Fh2>\n\u003Cp>Le Déclencheur de Récupération répond : Ai-je besoin de preuves externes ?\u003C\u002Fp>\n\u003Cp>La Couche d'Autorité de Récupération répond à la question suivante : Quelle source externe mérite de devenir une preuve ?\u003C\u002Fp>\n\u003Cpre class=\"code-block\">\u003Ccode>Model Knowledge\n   ↓\nRetrieval Trigger\n   ↓\nCandidate Sources\n   ↓\nRetrieval Authority Layer\n   ↓\nQualified Evidence\n   ↓\nReasoning\n   ↓\nAnswer\u003C\u002Fcode>\u003C\u002Fpre>\n\u003Cp>Cette distinction fait passer le RAG de la récupération de documents vers la récupération contrôlée de preuves.\u003C\u002Fp>\n\u003Ch2 id=\"section-78\">Sources Primaires\u003C\u002Fh2>\n\u003Cp>Patrick Lewis et al., \u003Ca href=\"https:\u002F\u002Farxiv.org\u002Fabs\u002F2005.11401\" target=\"_blank\">Retrieval-Augmented Generation for Knowledge-Intensive NLP Tasks\u003C\u002Fa> (2020).\u003C\u002Fp>\n\u003Cp>Akari Asai et al., \u003Ca href=\"https:\u002F\u002Farxiv.org\u002Fabs\u002F2310.11511\" target=\"_blank\">Self-RAG: Learning to Retrieve, Generate, and Critique through Self-Reflection\u003C\u002Fa> (2023).\u003C\u002Fp>\n\u003Cp>Soyeong Jeong et al., \u003Ca href=\"https:\u002F\u002Farxiv.org\u002Fabs\u002F2403.14403\" target=\"_blank\">Adaptive-RAG: Learning to Adapt Retrieval-Augmented Large Language Models through Question Complexity\u003C\u002Fa> (2024).\u003C\u002Fp>",{"time":228,"blocks":229,"version":683},1792072583886,[230,236,242,247,252,260,267,272,277,282,287,293,304,309,314,331,336,341,346,351,356,361,366,371,376,381,386,391,396,401,410,415,420,425,430,435,440,445,450,455,460,465,470,475,484,489,494,499,504,509,514,519,528,533,538,553,558,563,568,573,578,583,588,593,598,603,608,613,618,623,628,633,638,643,648,653,658,663,668,673,678],{"id":231,"data":232,"type":42,"tunes":235},"XpcNQW15-B",{"text":233,"level":234},"Question",2,{},{"id":237,"data":238,"type":240,"tunes":241},"ren2ISkSca",{"text":239},"Pourquoi un système de recherche par IA choisit-il une source tout en ignorant une autre source qui semble traiter exactement du même sujet ?","paragraph",{},{"id":243,"data":244,"type":240,"tunes":246},"X_bzzHXG4I",{"text":245},"La pertinence sémantique ne constitue qu'une partie de la réponse. Deux documents peuvent être également similaires à une requête tout en différant radicalement en termes d'autorité, de fraîcheur, de provenance, de portée, d'autorisations et d'applicabilité.",{},{"id":248,"data":249,"type":240,"tunes":251},"elm6ePGXbF",{"text":250},"Cet article utilise le terme Couche d'Autorité de Récupération pour désigner la frontière de sélection des sources qui se situe entre la recherche d'informations potentiellement pertinentes et le traitement de ces informations comme des preuves appropriées.",{},{"id":253,"data":254,"type":258,"tunes":259},"TeJ97szQOn",{"text":255,"caption":256,"alignment":257},"Une Couche d'Autorité de Récupération est l'ensemble des règles de sélection des sources et des propriétés des sources qui déterminent quelles informations récupérables méritent de devenir des preuves pour une question particulière.","Définition de travail — un concept systémique utilisé dans cet article, et non un terme de recherche standardisé.","left","quote",{},{"id":261,"data":262,"type":265,"tunes":266},"JfN6zo2EAW",{"title":263,"maxLevel":264,"minLevel":234},"Sommaire",3,"tableOfContents",{},{"id":268,"data":269,"type":42,"tunes":271},"RoTT8hQShH",{"text":270,"level":234},"Ce que cela signifie vraiment",{},{"id":273,"data":274,"type":240,"tunes":276},"Cns53IeLiV",{"text":275},"Un moteur de récupération répond à une question relativement étroite : Quels éléments semblent pertinents pour cette requête ?",{},{"id":278,"data":279,"type":240,"tunes":281},"2qvTGNQESy",{"text":280},"Un système d'IA en production doit généralement répondre à une question plus difficile : Lesquels de ces éléments pertinents devraient être autorisés à influencer la réponse ?",{},{"id":283,"data":284,"type":240,"tunes":286},"KneZrq8Dgr",{"text":285},"Cette seconde question introduit l'identité de la source, la version, l'autorité, la période de validité, la provenance et la portée. Une ancienne politique pertinente, un article de blog copié ou un résumé non officiel peut obtenir un très bon classement tout en restant la mauvaise preuve pour la décision.",{},{"id":288,"data":289,"type":291,"tunes":292},"DYVDznB9AW",{"code":290},"Question\n   ↓\nRetrieval Trigger\n   ↓\nCandidate Sources\n   ↓\nRETRIEVAL AUTHORITY LAYER\n   ↓\nQualified Evidence\n   ↓\nContext \u002F Reasoning\n   ↓\nAnswer Validity Boundary\n   ↓\nAnswer","code",{},{"id":294,"data":295,"type":302,"tunes":303},"mBPGlmAz5D",{"link":296,"meta":297},"https:\u002F\u002Fstajic.de\u002Ffr\u002Fblog\u002Fwhen-should-an-ai-stop-trusting-its-own-knowledge-the-retrieval-trigger",{"image":298,"title":300,"description":301},{"url":299},"","Quand une IA devrait-elle cesser de faire confiance à ses propres connaissances ? — Le Déclencheur de Récupération","Le Déclencheur de Récupération décide quand des preuves externes sont nécessaires. La Couche d'Autorité de Récupération détermine quelles sources doivent les fournir.","linkTool",{},{"id":305,"data":306,"type":42,"tunes":308},"zuuDXruzRL",{"text":307,"level":234},"Exemple le plus simple",{},{"id":310,"data":311,"type":240,"tunes":313},"QP0qUL8Ciw",{"text":312},"Imaginez qu'un employé demande à un assistant d'entreprise : Combien de jours de vacances les employés reçoivent-ils ?",{},{"id":315,"data":316,"type":329,"tunes":330},"2Ev2t8opWr",{"content":317,"stretched":43,"withHeadings":14},[318,322,326],[319,320,321],"Candidat","Contenu","Pertinence sémantique",[323,324,325],"A","Politique RH actuelle, version approuvée 6.2","Élevée",[327,328,325],"B","Ancien guide d'intégration copié de la version 4.8","table",{},{"id":332,"data":333,"type":240,"tunes":335},"wtPrU1Dnqd",{"text":334},"Les deux passages peuvent contenir une formulation presque identique. Un score de similarité peut donc les placer très proches l'un de l'autre.",{},{"id":337,"data":338,"type":291,"tunes":340},"v97zTkFBzt",{"code":339},"Semantic relevance\n        ≠\nSource authority\n        ≠\nApplicability",{},{"id":342,"data":343,"type":240,"tunes":345},"kAKa32Hh_B",{"text":344},"La politique RH actuelle approuvée devrait normalement dominer car elle constitue la source de référence pour l'affirmation. L'ancien guide reste un texte pertinent, mais il ne devrait pas avoir le même poids probatoire.",{},{"id":347,"data":348,"type":42,"tunes":350},"gUE1_pfNG9",{"text":349,"level":234},"Là où l'exemple cesse de fonctionner",{},{"id":352,"data":353,"type":240,"tunes":355},"3c0r-JJC4s",{"text":354},"L'autorité n'est pas toujours une hiérarchie universelle dans laquelle une source l'emporte à chaque fois.",{},{"id":357,"data":358,"type":240,"tunes":360},"q1E1rsEJq0",{"text":359},"Une base de données transactionnelle peut faire autorité pour le statut actuel d'une commande. Un manuel de produit peut faire autorité pour le fonctionnement prévu. Un contrat signé peut régir une obligation commerciale. Un article scientifique primaire peut être la meilleure source pour son expérience, tandis qu'une revue systématique ultérieure peut être plus solide pour une conclusion plus large.",{},{"id":362,"data":363,"type":240,"tunes":365},"oCUaHjt58Z",{"text":364},"L'autorité est donc relative à l'affirmation. La classe de source correcte dépend de ce qui est demandé.",{},{"id":367,"data":368,"type":42,"tunes":370},"4CxlnDa9Cc",{"text":369,"level":234},"Réponse directe",{},{"id":372,"data":373,"type":240,"tunes":375},"QW5_w33xWI",{"text":374},"Une source devrait devenir une preuve de récupération préférée lorsqu'elle n'est pas seulement pertinente, mais aussi suffisamment faisant autorité, actuelle, applicable, traçable et accessible pour l'affirmation spécifique à laquelle elle répond.",{},{"id":377,"data":378,"type":291,"tunes":380},"owyjAlbwY4",{"code":379},"Useful retrieval evidence\n=\nrelevance\n+ authority\n+ provenance\n+ freshness\n+ applicability\n+ accessibility",{},{"id":382,"data":383,"type":240,"tunes":385},"IBCawe0zx3",{"text":384},"Ces propriétés ne garantissent pas la vérité. Elles créent une base plus solide pour décider quelle information récupérée mérite d'entrer dans le processus de raisonnement.",{},{"id":387,"data":388,"type":42,"tunes":390},"4vZIEUjrdq",{"text":389,"level":234},"Pourquoi la similarité seule ne suffit pas",{},{"id":392,"data":393,"type":240,"tunes":395},"1PFBw4V0ni",{"text":394},"La récupération vectorielle est extrêmement utile car elle peut trouver des informations sémantiquement liées même lorsque la formulation diffère. Mais la similarité sémantique ne sait pas par elle-même si un document est obsolète, remplacé, non officiel, non autorisé ou valide pour une autre juridiction.",{},{"id":397,"data":398,"type":240,"tunes":400},"vfaNbSsKCP",{"text":399},"L'architecture RAG originale a montré comment une mémoire externe non paramétrique peut augmenter les connaissances du modèle. Les systèmes de production doivent ajouter des contrôles spécifiques à l'application autour de ce processus de récupération.",{},{"id":402,"data":403,"type":302,"tunes":409},"RKi9NxoiRb",{"link":404,"meta":405},"https:\u002F\u002Fstajic.de\u002Ffr\u002Fblog\u002Fwhere-does-an-llm-get-its-data-rag-data-sources-in-python",{"image":406,"title":407,"description":408},{"url":299},"Où un LLM obtient-il ses données ? Sources de données RAG en Python","Les fichiers, bases de données, API, recherches et index vectoriels peuvent tous fournir des connaissances à l'exécution. La méthode de récupération doit correspondre à la source et à la question.",{},{"id":411,"data":412,"type":42,"tunes":414},"hULFRV_NCz",{"text":413,"level":234},"Les six signaux d'autorité",{},{"id":416,"data":417,"type":42,"tunes":419},"7lJ3RudSld",{"text":418,"level":264},"1. Identité de la source",{},{"id":421,"data":422,"type":240,"tunes":424},"J2JcDKRmnb",{"text":423},"Le système doit savoir ce qu'est réellement la source : politique, contrat, API, enregistrement de base de données, article de recherche, page de documentation, fichier utilisateur ou commentaire tiers.",{},{"id":426,"data":427,"type":42,"tunes":429},"bKliUrdBTI",{"text":428,"level":264},"2. Provenance",{},{"id":431,"data":432,"type":240,"tunes":434},"LRJssGYfAY",{"text":433},"Les preuves doivent préserver leur origine, la manière dont elles sont entrées dans le système et quel objet ou document original peut être inspecté si l'affirmation est contestée.",{},{"id":436,"data":437,"type":42,"tunes":439},"N-KXk6Cp2H",{"text":438,"level":264},"3. Fraîcheur",{},{"id":441,"data":442,"type":240,"tunes":444},"FJKVbQVgOr",{"text":443},"La fraîcheur importe lorsque le fait sous-jacent change. Une réponse API actuelle et un article vieux de cinq ans peuvent être également pertinents pour les mots d'une requête, mais pas pour son exigence temporelle.",{},{"id":446,"data":447,"type":42,"tunes":449},"8Wxcbhl365",{"text":448,"level":264},"4. Applicabilité",{},{"id":451,"data":452,"type":240,"tunes":454},"8w01u-EYq6",{"text":453},"Un fait peut être correct mais applicable à la mauvaise version de produit, pays, locataire, groupe de clients, contrat, version logicielle ou période.",{},{"id":456,"data":457,"type":42,"tunes":459},"NYQm_3y_PX",{"text":458,"level":264},"5. Spécificité des preuves",{},{"id":461,"data":462,"type":240,"tunes":464},"uqqpEuQdxC",{"text":463},"Une preuve primaire qui soutient directement l'affirmation est généralement plus forte qu'une chaîne de résumés qui ne font que la répéter.",{},{"id":466,"data":467,"type":42,"tunes":469},"GggJkQwr1C",{"text":468,"level":264},"6. Légitimité de l'accès",{},{"id":471,"data":472,"type":240,"tunes":474},"spnPYg1KRU",{"text":473},"Un document techniquement récupérable n'est pas nécessairement un contexte autorisé. L'autorité de récupération doit opérer dans les limites du locataire, des permissions, de la confidentialité et de la minimisation des données.",{},{"id":476,"data":477,"type":302,"tunes":483},"yvkNCwWFiC",{"link":478,"meta":479},"https:\u002F\u002Fstajic.de\u002Ffr\u002Fblog\u002Fwhat-is-context-engineering-what-the-model-receives-before-it-answers",{"image":480,"title":481,"description":482},{"url":299},"Qu'est-ce que l'ingénierie de contexte ? Ce que le modèle reçoit avant de répondre","L'assemblage du contexte est aussi une frontière de validité et de sécurité. Les informations récupérées doivent préserver la portée, l'autorité et les contraintes d'accès.",{},{"id":485,"data":486,"type":42,"tunes":488},"b_9T6lmCVJ",{"text":487,"level":234},"Un modèle d'autorité minimal",{},{"id":490,"data":491,"type":240,"tunes":493},"VpHVLljSPR",{"text":492},"Un système pratique peut représenter ces propriétés explicitement plutôt que d'espérer que le classement par similarité les capture.",{},{"id":495,"data":496,"type":291,"tunes":498},"L4W7ZQSTYs",{"code":497},"{\n  \"source\": \"hr-policy-6.2\",\n  \"sourceType\": \"approved_policy\",\n  \"effectiveFrom\": \"2026-01-01\",\n  \"supersedes\": \"hr-policy-5.9\",\n  \"jurisdiction\": \"DE\",\n  \"authority\": \"primary\",\n  \"accessScope\": \"employee\",\n  \"provenance\": \"hr-policy-system\"\n}",{},{"id":500,"data":501,"type":240,"tunes":503},"VISBdGcdtA",{"text":502},"Le récupérateur peut toujours utiliser la pertinence sémantique ou lexicale. La couche d'autorité ajoute des contraintes et des signaux de reclassement que la pertinence seule ne peut fournir.",{},{"id":505,"data":506,"type":42,"tunes":508},"GcbcUHNia6",{"text":507,"level":234},"Connexion à l'information prête pour les agents",{},{"id":510,"data":511,"type":240,"tunes":513},"CVqLmnn5CD",{"text":512},"Cela devient particulièrement important lorsque le consommateur n'est pas un humain lisant des documents mais un agent IA prenant des décisions.",{},{"id":515,"data":516,"type":240,"tunes":518},"MkvgpsEqJc",{"text":517},"Un objet d'information prêt pour les agents devrait préserver l'identité, la provenance, les conflits, les inconnues et les conditions de validité avant l'étape de récupération. Sinon, le système tente de reconstruire l'autorité après que l'information a déjà été aplatie en texte anonyme.",{},{"id":520,"data":521,"type":302,"tunes":527},"5KX3sugIpt",{"link":522,"meta":523},"https:\u002F\u002Fstajic.de\u002Ffr\u002Fportfolio\u002Ffrom-raw-data-to-verifiable-ai-context-a-practical-architecture-for-agent-ready-information",{"image":524,"title":525,"description":526},{"url":299},"Des données brutes au contexte IA vérifiable : une architecture pratique pour l'information prête pour les agents","Une architecture de portefeuille pour transformer des données hétérogènes en objets d'information contrôlés avec identité, provenance, conflits et validité.",{},{"id":529,"data":530,"type":42,"tunes":532},"O6bYy5LenW",{"text":531,"level":234},"Méthode de diagnostic \u002F décision",{},{"id":534,"data":535,"type":240,"tunes":537},"cZeNmj3Z_3",{"text":536},"Pour chaque affirmation récupérée de grande valeur, posez six questions :",{},{"id":539,"data":540,"type":551,"tunes":552},"lWP8qNNrxw",{"meta":541,"items":543,"style":550},{"counterType":542},"numeric",[544,545,546,547,548,549],"Quelle source exacte a produit cette preuve ?","Cette source fait-elle autorité pour ce type d'affirmation ?","L'information est-elle toujours d'actualité ?","S'applique-t-elle à cet utilisateur, cette version, cette juridiction et cette période ?","L'affirmation peut-elle être retracée jusqu'à la preuve originale ?","L'utilisateur ou l'agent était-il autorisé à la récupérer ?","ordered","list",{},{"id":554,"data":555,"type":240,"tunes":557},"zzgllTPRBV",{"text":556},"Si le système ne peut pas répondre à ces questions, le problème n'est plus seulement la qualité de la récupération. C'est un problème de gouvernance des preuves.",{},{"id":559,"data":560,"type":42,"tunes":562},"gmdM4uPGYl",{"text":561,"level":234},"Modes de défaillance courants",{},{"id":564,"data":565,"type":240,"tunes":567},"GjbXINllzx",{"text":566},"Le score de similarité le plus élevé l'emporte. Cela confond la proximité sémantique avec l'autorité probante.",{},{"id":569,"data":570,"type":240,"tunes":572},"DapWd1cmb5",{"text":571},"La source la plus récente l'emporte toujours. Un commentaire plus récent peut être moins faisant autorité qu'un contrat ou une norme plus ancienne qui fait foi.",{},{"id":574,"data":575,"type":240,"tunes":577},"VVGLrdBjxS",{"text":576},"Officiel signifie universellement applicable. Une source officielle peut toujours appartenir à une autre juridiction, version ou portée.",{},{"id":579,"data":580,"type":240,"tunes":582},"vBrSvczrCq",{"text":581},"Public signifie sûr à récupérer. Une information peut être publique mais inappropriée pour la tâche, tandis qu'une information privée peut être légitime uniquement dans un contexte autorisé.",{},{"id":584,"data":585,"type":42,"tunes":587},"gN24OPQDE2",{"text":586,"level":234},"Cas limite : sources faisant autorité en conflit",{},{"id":589,"data":590,"type":240,"tunes":592},"qcST_7WVLw",{"text":591},"Le cas difficile n'est pas une source faisant autorité contre une source manifestement faible. C'est deux sources apparemment faisant autorité qui sont en désaccord.",{},{"id":594,"data":595,"type":240,"tunes":597},"E9gsL-vSes",{"text":596},"À ce stade, la couche d'autorité devrait préserver le conflit au lieu de l'effondrer silencieusement en une seule réponse.",{},{"id":599,"data":600,"type":291,"tunes":602},"CBZ5ctc6_X",{"code":601},"Retrieve\n   ↓\nQualify source authority\n   ↓\nDetect conflict\n   ↓\nPreserve disagreement\n   ↓\nReason about scope \u002F version \u002F provenance\n   ↓\nAnswer only inside the supported boundary",{},{"id":604,"data":605,"type":42,"tunes":607},"xeQqLNvjCb",{"text":606,"level":234},"Limites",{},{"id":609,"data":610,"type":240,"tunes":612},"raTFpBhsAR",{"text":611},"La couche d'autorité de récupération est une architecture conceptuelle, pas un algorithme de classement universel.",{},{"id":614,"data":615,"type":240,"tunes":617},"G6tXMIP-u1",{"text":616},"Différents domaines nécessitent différents modèles d'autorité. Les données juridiques, scientifiques, commerciales, opérationnelles et personnelles ne peuvent pas toutes utiliser la même hiérarchie de sources.",{},{"id":619,"data":620,"type":240,"tunes":622},"SWCg4MmSIb",{"text":621},"Elle ne peut pas non plus garantir qu'un moteur de recherche externe ou un système d'IA classera la page d'un éditeur. Les systèmes de classement externes restent hors du contrôle de l'éditeur.",{},{"id":624,"data":625,"type":42,"tunes":627},"j_xI-ClxxP",{"text":626,"level":234},"Qu'est-ce qui changerait cette réponse ?",{},{"id":629,"data":630,"type":240,"tunes":632},"xrctSdgP3p",{"text":631},"Les systèmes de récupération peuvent de plus en plus apprendre automatiquement la fiabilité des sources, combiner la provenance structurée avec le classement neuronal et négocier l'accès via des protocoles d'agents.",{},{"id":634,"data":635,"type":240,"tunes":637},"Ka8dOdJHXO",{"text":636},"Cela peut changer la façon dont les signaux d'autorité sont mis en œuvre. Cela ne supprime pas la distinction entre trouver une information pertinente et décider si cette information mérite un poids probant.",{},{"id":639,"data":640,"type":42,"tunes":642},"E6SLB3m5CI",{"text":641,"level":234},"Conclusion",{},{"id":644,"data":645,"type":240,"tunes":647},"0_oDIxt-xd",{"text":646},"Le Déclencheur de Récupération répond : Ai-je besoin de preuves externes ?",{},{"id":649,"data":650,"type":240,"tunes":652},"mCMupYycYk",{"text":651},"La Couche d'Autorité de Récupération répond à la question suivante : Quelle source externe mérite de devenir une preuve ?",{},{"id":654,"data":655,"type":291,"tunes":657},"rSohO6bowX",{"code":656},"Model Knowledge\n   ↓\nRetrieval Trigger\n   ↓\nCandidate Sources\n   ↓\nRetrieval Authority Layer\n   ↓\nQualified Evidence\n   ↓\nReasoning\n   ↓\nAnswer",{},{"id":659,"data":660,"type":240,"tunes":662},"f8GMr_eelV",{"text":661},"Cette distinction fait passer le RAG de la récupération de documents vers la récupération contrôlée de preuves.",{},{"id":664,"data":665,"type":42,"tunes":667},"hOWp43J5Dw",{"text":666,"level":234},"Sources Primaires",{},{"id":669,"data":670,"type":240,"tunes":672},"evqsbllTgC",{"text":671},"Patrick Lewis et al., \u003Ca href=\"https:\u002F\u002Farxiv.org\u002Fabs\u002F2005.11401\" target=\"_blank\">Retrieval-Augmented Generation for Knowledge-Intensive NLP Tasks\u003C\u002Fa> (2020).",{},{"id":674,"data":675,"type":240,"tunes":677},"gz0XrNndq4",{"text":676},"Akari Asai et al., \u003Ca href=\"https:\u002F\u002Farxiv.org\u002Fabs\u002F2310.11511\" target=\"_blank\">Self-RAG: Learning to Retrieve, Generate, and Critique through Self-Reflection\u003C\u002Fa> (2023).",{},{"id":679,"data":680,"type":240,"tunes":682},"NgkPqz5Poz",{"text":681},"Soyeong Jeong et al., \u003Ca href=\"https:\u002F\u002Farxiv.org\u002Fabs\u002F2403.14403\" target=\"_blank\">Adaptive-RAG: Learning to Adapt Retrieval-Augmented Large Language Models through Question Complexity\u003C\u002Fa> (2024).",{},"2.31","Trouver un document pertinent ne suffit pas. Les systèmes d’IA doivent aussi distinguer les sources actuelles, faisant autorité, applicables et traçables des sources qui semblent simplement similaires. La couche d’autorité de récupération définit cette frontière de sélection des sources.","\u002Fuploads\u002F2026\u002F10\u002Fwhy-does-ai-retrieve-some-sources-and-ignore-others-the-retrieval-authority-layer-1792072410616-0evp1z.webp","why-does-ai-retrieve-some-sources-and-ignore-others-the-retrieval-authority-layer-1792072410616-0evp1z","PUBLISHED","2026-10-15T07:51:00.000Z","2026-10-15T13:51:04.579Z","2026-10-15T20:37:39.839Z",{"en":692,"de":693,"sr":694,"es":695,"fr":696,"it":697,"ru":698,"zh":699},"\u002Fblog\u002Fwhy-does-ai-retrieve-some-sources-and-ignore-others-the-retrieval-authority-layer","\u002Fde\u002Fblog\u002Fwhy-does-ai-retrieve-some-sources-and-ignore-others-the-retrieval-authority-layer","\u002Fsr\u002Fblog\u002Fwhy-does-ai-retrieve-some-sources-and-ignore-others-the-retrieval-authority-layer","\u002Fes\u002Fblog\u002Fwhy-does-ai-retrieve-some-sources-and-ignore-others-the-retrieval-authority-layer","\u002Ffr\u002Fblog\u002Fwhy-does-ai-retrieve-some-sources-and-ignore-others-the-retrieval-authority-layer","\u002Fit\u002Fblog\u002Fwhy-does-ai-retrieve-some-sources-and-ignore-others-the-retrieval-authority-layer","\u002Fru\u002Fblog\u002Fwhy-does-ai-retrieve-some-sources-and-ignore-others-the-retrieval-authority-layer","\u002Fzh\u002Fblog\u002Fwhy-does-ai-retrieve-some-sources-and-ignore-others-the-retrieval-authority-layer",[701,705],{"id":702,"name":703,"slug":704},65,"Système d’autorité thématique","topical-authority",{"id":706,"name":707,"slug":708},106,"Runbook : Rollback IA","ai-rollback",{"id":710,"login":711,"email":299,"displayName":712},"20","rooth8233","Aleksandar Stajić",[714,1068],{"lang":715,"title":716,"content":717,"contentJson":718,"excerpt":1067},"en","Why Does AI Retrieve Some Sources and Ignore Others? — The Retrieval Authority Layer","{\"time\":1792072370632,\"blocks\":[{\"id\":\"XpcNQW15-B\",\"data\":{\"text\":\"Question\",\"level\":2},\"type\":\"header\",\"tunes\":{}},{\"id\":\"ren2ISkSca\",\"data\":{\"text\":\"Why does an AI retrieval system choose one source while ignoring another source that appears to discuss exactly the same subject?\"},\"type\":\"paragraph\",\"tunes\":{}},{\"id\":\"X_bzzHXG4I\",\"data\":{\"text\":\"Semantic relevance is only part of the answer. Two documents can be equally similar to a query while differing radically in authority, freshness, provenance, scope, permissions and applicability.\"},\"type\":\"paragraph\",\"tunes\":{}},{\"id\":\"elm6ePGXbF\",\"data\":{\"text\":\"This article uses the term Retrieval Authority Layer for the source-selection boundary that sits between finding potentially relevant information and treating that information as suitable evidence.\"},\"type\":\"paragraph\",\"tunes\":{}},{\"id\":\"TeJ97szQOn\",\"data\":{\"text\":\"A Retrieval Authority Layer is the set of source-selection rules and source properties that determine which retrievable information deserves to become evidence for a particular question.\",\"caption\":\"Working definition — a systems concept used in this article, not a standardized research term.\",\"alignment\":\"left\"},\"type\":\"quote\",\"tunes\":{}},{\"id\":\"JfN6zo2EAW\",\"data\":{\"title\":\"Contents\",\"maxLevel\":3,\"minLevel\":2},\"type\":\"tableOfContents\",\"tunes\":{}},{\"id\":\"RoTT8hQShH\",\"data\":{\"text\":\"What This Really Means\",\"level\":2},\"type\":\"header\",\"tunes\":{}},{\"id\":\"Cns53IeLiV\",\"data\":{\"text\":\"A retriever answers a relatively narrow question: Which items appear relevant to this query?\"},\"type\":\"paragraph\",\"tunes\":{}},{\"id\":\"2qvTGNQESy\",\"data\":{\"text\":\"A production AI system usually needs to answer a harder one: Which of those relevant items should be allowed to influence the answer?\"},\"type\":\"paragraph\",\"tunes\":{}},{\"id\":\"KneZrq8Dgr\",\"data\":{\"text\":\"That second question introduces source identity, version, authority, validity period, provenance and scope. A relevant old policy, copied blog post or unofficial summary can rank extremely well while still being the wrong evidence for the decision.\"},\"type\":\"paragraph\",\"tunes\":{}},{\"id\":\"DYVDznB9AW\",\"data\":{\"code\":\"Question\\n   ↓\\nRetrieval Trigger\\n   ↓\\nCandidate Sources\\n   ↓\\nRETRIEVAL AUTHORITY LAYER\\n   ↓\\nQualified Evidence\\n   ↓\\nContext \u002F Reasoning\\n   ↓\\nAnswer Validity Boundary\\n   ↓\\nAnswer\"},\"type\":\"code\",\"tunes\":{}},{\"id\":\"mBPGlmAz5D\",\"data\":{\"link\":\"https:\u002F\u002Fstajic.de\u002Fblog\u002Fwhen-should-an-ai-stop-trusting-its-own-knowledge-the-retrieval-trigger\",\"meta\":{\"image\":{\"url\":\"\"},\"title\":\"When Should an AI Stop Trusting Its Own Knowledge? — The Retrieval Trigger\",\"description\":\"The Retrieval Trigger decides when external evidence is required. The Retrieval Authority Layer asks which sources should provide it.\"}},\"type\":\"linkTool\",\"tunes\":{}},{\"id\":\"zuuDXruzRL\",\"data\":{\"text\":\"Simplest Example\",\"level\":2},\"type\":\"header\",\"tunes\":{}},{\"id\":\"QP0qUL8Ciw\",\"data\":{\"text\":\"Imagine an employee asks an enterprise assistant: How many vacation days do employees receive?\"},\"type\":\"paragraph\",\"tunes\":{}},{\"id\":\"2Ev2t8opWr\",\"data\":{\"content\":[[\"Candidate\",\"Content\",\"Semantic relevance\"],[\"A\",\"Current HR policy, approved version 6.2\",\"High\"],[\"B\",\"Old onboarding guide copied from version 4.8\",\"High\"]],\"stretched\":false,\"withHeadings\":true},\"type\":\"table\",\"tunes\":{}},{\"id\":\"wtPrU1Dnqd\",\"data\":{\"text\":\"Both passages can contain nearly identical wording. A similarity score may therefore place them very close together.\"},\"type\":\"paragraph\",\"tunes\":{}},{\"id\":\"v97zTkFBzt\",\"data\":{\"code\":\"Semantic relevance\\n        ≠\\nSource authority\\n        ≠\\nApplicability\"},\"type\":\"code\",\"tunes\":{}},{\"id\":\"kAKa32Hh_B\",\"data\":{\"text\":\"The current approved HR policy should normally dominate because it is the governing source for the claim. The older guide remains relevant text, but it should not carry the same evidential weight.\"},\"type\":\"paragraph\",\"tunes\":{}},{\"id\":\"gUE1_pfNG9\",\"data\":{\"text\":\"Where the Example Stops Working\",\"level\":2},\"type\":\"header\",\"tunes\":{}},{\"id\":\"3c0r-JJC4s\",\"data\":{\"text\":\"Authority is not always a universal hierarchy in which one source wins every time.\"},\"type\":\"paragraph\",\"tunes\":{}},{\"id\":\"q1E1rsEJq0\",\"data\":{\"text\":\"A transactional database may be authoritative for the current status of an order. A product manual may be authoritative for intended operation. A signed contract may control a commercial obligation. A primary scientific paper may be the best source for its experiment, while a later systematic review may be stronger for a broader conclusion.\"},\"type\":\"paragraph\",\"tunes\":{}},{\"id\":\"oCUaHjt58Z\",\"data\":{\"text\":\"Authority is therefore claim-relative. The correct source class depends on what is being asked.\"},\"type\":\"paragraph\",\"tunes\":{}},{\"id\":\"4CxlnDa9Cc\",\"data\":{\"text\":\"Direct Answer\",\"level\":2},\"type\":\"header\",\"tunes\":{}},{\"id\":\"QW5_w33xWI\",\"data\":{\"text\":\"A source should become preferred retrieval evidence when it is not merely relevant, but also sufficiently authoritative, current, applicable, traceable and accessible for the specific claim being answered.\"},\"type\":\"paragraph\",\"tunes\":{}},{\"id\":\"owyjAlbwY4\",\"data\":{\"code\":\"Useful retrieval evidence\\n=\\nrelevance\\n+ authority\\n+ provenance\\n+ freshness\\n+ applicability\\n+ accessibility\"},\"type\":\"code\",\"tunes\":{}},{\"id\":\"IBCawe0zx3\",\"data\":{\"text\":\"These properties do not guarantee truth. They create a stronger basis for deciding which retrieved information deserves to enter the reasoning process.\"},\"type\":\"paragraph\",\"tunes\":{}},{\"id\":\"4vZIEUjrdq\",\"data\":{\"text\":\"Why Similarity Alone Is Not Enough\",\"level\":2},\"type\":\"header\",\"tunes\":{}},{\"id\":\"1PFBw4V0ni\",\"data\":{\"text\":\"Vector retrieval is extremely useful because it can find semantically related information even when the wording differs. But semantic similarity does not know by itself whether a document is obsolete, superseded, unofficial, unauthorized or valid for another jurisdiction.\"},\"type\":\"paragraph\",\"tunes\":{}},{\"id\":\"vfaNbSsKCP\",\"data\":{\"text\":\"The original RAG architecture showed how external non-parametric memory can augment model knowledge. Production systems must add application-specific controls around that retrieval process.\"},\"type\":\"paragraph\",\"tunes\":{}},{\"id\":\"RKi9NxoiRb\",\"data\":{\"link\":\"https:\u002F\u002Fstajic.de\u002Fblog\u002Fwhere-does-an-llm-get-its-data-rag-data-sources-in-python\",\"meta\":{\"image\":{\"url\":\"\"},\"title\":\"Where Does an LLM Get Its Data? RAG Data Sources in Python\",\"description\":\"Files, databases, APIs, search and vector indexes can all provide runtime knowledge. The retrieval method should match the source and the question.\"}},\"type\":\"linkTool\",\"tunes\":{}},{\"id\":\"hULFRV_NCz\",\"data\":{\"text\":\"The Six Authority Signals\",\"level\":2},\"type\":\"header\",\"tunes\":{}},{\"id\":\"7lJ3RudSld\",\"data\":{\"text\":\"1. Source Identity\",\"level\":3},\"type\":\"header\",\"tunes\":{}},{\"id\":\"J2JcDKRmnb\",\"data\":{\"text\":\"The system should know what the source actually is: policy, contract, API, database record, research paper, documentation page, user file or third-party commentary.\"},\"type\":\"paragraph\",\"tunes\":{}},{\"id\":\"bKliUrdBTI\",\"data\":{\"text\":\"2. Provenance\",\"level\":3},\"type\":\"header\",\"tunes\":{}},{\"id\":\"LRJssGYfAY\",\"data\":{\"text\":\"Evidence should preserve where it came from, how it entered the system and which original object or document can be inspected if the claim is challenged.\"},\"type\":\"paragraph\",\"tunes\":{}},{\"id\":\"N-KXk6Cp2H\",\"data\":{\"text\":\"3. Freshness\",\"level\":3},\"type\":\"header\",\"tunes\":{}},{\"id\":\"FJKVbQVgOr\",\"data\":{\"text\":\"Freshness matters when the underlying fact changes. A current API response and a five-year-old article may be equally relevant to the words of a query but not to its temporal requirement.\"},\"type\":\"paragraph\",\"tunes\":{}},{\"id\":\"8Wxcbhl365\",\"data\":{\"text\":\"4. Applicability\",\"level\":3},\"type\":\"header\",\"tunes\":{}},{\"id\":\"8w01u-EYq6\",\"data\":{\"text\":\"A fact may be correct but applicable to the wrong product version, country, tenant, customer group, contract, software release or time period.\"},\"type\":\"paragraph\",\"tunes\":{}},{\"id\":\"NYQm_3y_PX\",\"data\":{\"text\":\"5. Evidence Specificity\",\"level\":3},\"type\":\"header\",\"tunes\":{}},{\"id\":\"uqqpEuQdxC\",\"data\":{\"text\":\"Primary evidence that directly supports the claim is generally stronger than a chain of summaries that merely repeat it.\"},\"type\":\"paragraph\",\"tunes\":{}},{\"id\":\"GggJkQwr1C\",\"data\":{\"text\":\"6. Access Legitimacy\",\"level\":3},\"type\":\"header\",\"tunes\":{}},{\"id\":\"spnPYg1KRU\",\"data\":{\"text\":\"A technically retrievable document is not necessarily authorized context. Retrieval authority must operate inside tenant, permission, confidentiality and data-minimization boundaries.\"},\"type\":\"paragraph\",\"tunes\":{}},{\"id\":\"yvkNCwWFiC\",\"data\":{\"link\":\"https:\u002F\u002Fstajic.de\u002Fblog\u002Fwhat-is-context-engineering-what-the-model-receives-before-it-answers\",\"meta\":{\"image\":{\"url\":\"\"},\"title\":\"What Is Context Engineering? What the Model Receives Before It Answers\",\"description\":\"Context assembly is also a validity and security boundary. Retrieved information must preserve scope, authority and access constraints.\"}},\"type\":\"linkTool\",\"tunes\":{}},{\"id\":\"b_9T6lmCVJ\",\"data\":{\"text\":\"A Minimal Authority Model\",\"level\":2},\"type\":\"header\",\"tunes\":{}},{\"id\":\"VpHVLljSPR\",\"data\":{\"text\":\"A practical system can represent these properties explicitly rather than hoping that similarity ranking captures them.\"},\"type\":\"paragraph\",\"tunes\":{}},{\"id\":\"L4W7ZQSTYs\",\"data\":{\"code\":\"{\\n  \\\"source\\\": \\\"hr-policy-6.2\\\",\\n  \\\"sourceType\\\": \\\"approved_policy\\\",\\n  \\\"effectiveFrom\\\": \\\"2026-01-01\\\",\\n  \\\"supersedes\\\": \\\"hr-policy-5.9\\\",\\n  \\\"jurisdiction\\\": \\\"DE\\\",\\n  \\\"authority\\\": \\\"primary\\\",\\n  \\\"accessScope\\\": \\\"employee\\\",\\n  \\\"provenance\\\": \\\"hr-policy-system\\\"\\n}\"},\"type\":\"code\",\"tunes\":{}},{\"id\":\"VISBdGcdtA\",\"data\":{\"text\":\"The retriever can still use semantic or lexical relevance. The authority layer adds constraints and reranking signals that relevance alone cannot provide.\"},\"type\":\"paragraph\",\"tunes\":{}},{\"id\":\"GcbcUHNia6\",\"data\":{\"text\":\"Connection to Agent-Ready Information\",\"level\":2},\"type\":\"header\",\"tunes\":{}},{\"id\":\"CVqLmnn5CD\",\"data\":{\"text\":\"This becomes especially important when the consumer is not a human reading documents but an AI agent making decisions.\"},\"type\":\"paragraph\",\"tunes\":{}},{\"id\":\"MkvgpsEqJc\",\"data\":{\"text\":\"An agent-ready information object should preserve identity, provenance, conflicts, unknowns and validity conditions before the retrieval stage. Otherwise the system tries to reconstruct authority after information has already been flattened into anonymous text.\"},\"type\":\"paragraph\",\"tunes\":{}},{\"id\":\"5KX3sugIpt\",\"data\":{\"link\":\"https:\u002F\u002Fstajic.de\u002Fportfolio\u002Ffrom-raw-data-to-verifiable-ai-context-a-practical-architecture-for-agent-ready-information\",\"meta\":{\"image\":{\"url\":\"\"},\"title\":\"From Raw Data to Verifiable AI Context: A Practical Architecture for Agent-Ready Information\",\"description\":\"A portfolio architecture for turning heterogeneous data into controlled information objects with identity, provenance, conflicts and validity.\"}},\"type\":\"linkTool\",\"tunes\":{}},{\"id\":\"O6bYy5LenW\",\"data\":{\"text\":\"Diagnostic \u002F Decision Method\",\"level\":2},\"type\":\"header\",\"tunes\":{}},{\"id\":\"cZeNmj3Z_3\",\"data\":{\"text\":\"For every high-value retrieved claim, ask six questions:\"},\"type\":\"paragraph\",\"tunes\":{}},{\"id\":\"lWP8qNNrxw\",\"data\":{\"meta\":{\"counterType\":\"numeric\"},\"items\":[\"What exact source produced this evidence?\",\"Is this source authoritative for this type of claim?\",\"Is the information still current?\",\"Does it apply to this user, version, jurisdiction and time period?\",\"Can the claim be traced back to the original evidence?\",\"Was the user or agent authorized to retrieve it?\"],\"style\":\"ordered\"},\"type\":\"list\",\"tunes\":{}},{\"id\":\"zzgllTPRBV\",\"data\":{\"text\":\"If the system cannot answer these questions, the problem is no longer only retrieval quality. It is an evidence-governance problem.\"},\"type\":\"paragraph\",\"tunes\":{}},{\"id\":\"gmdM4uPGYl\",\"data\":{\"text\":\"Common Failure Modes\",\"level\":2},\"type\":\"header\",\"tunes\":{}},{\"id\":\"GjbXINllzx\",\"data\":{\"text\":\"The highest similarity score wins. This confuses semantic closeness with evidential authority.\"},\"type\":\"paragraph\",\"tunes\":{}},{\"id\":\"DapWd1cmb5\",\"data\":{\"text\":\"The newest source always wins. A newer commentary may be less authoritative than an older governing contract or standard.\"},\"type\":\"paragraph\",\"tunes\":{}},{\"id\":\"VVGLrdBjxS\",\"data\":{\"text\":\"Official means universally applicable. An official source may still belong to another jurisdiction, version or scope.\"},\"type\":\"paragraph\",\"tunes\":{}},{\"id\":\"vBrSvczrCq\",\"data\":{\"text\":\"Public means safe to retrieve. Information can be public yet inappropriate for the task, while private information can be legitimate only inside an authorized context.\"},\"type\":\"paragraph\",\"tunes\":{}},{\"id\":\"gN24OPQDE2\",\"data\":{\"text\":\"Edge Case: Conflicting Authoritative Sources\",\"level\":2},\"type\":\"header\",\"tunes\":{}},{\"id\":\"qcST_7WVLw\",\"data\":{\"text\":\"The difficult case is not an authoritative source versus an obviously weak source. It is two apparently authoritative sources that disagree.\"},\"type\":\"paragraph\",\"tunes\":{}},{\"id\":\"E9gsL-vSes\",\"data\":{\"text\":\"At that point, the authority layer should preserve the conflict instead of silently collapsing it into one answer.\"},\"type\":\"paragraph\",\"tunes\":{}},{\"id\":\"CBZ5ctc6_X\",\"data\":{\"code\":\"Retrieve\\n   ↓\\nQualify source authority\\n   ↓\\nDetect conflict\\n   ↓\\nPreserve disagreement\\n   ↓\\nReason about scope \u002F version \u002F provenance\\n   ↓\\nAnswer only inside the supported boundary\"},\"type\":\"code\",\"tunes\":{}},{\"id\":\"xeQqLNvjCb\",\"data\":{\"text\":\"Limitations\",\"level\":2},\"type\":\"header\",\"tunes\":{}},{\"id\":\"raTFpBhsAR\",\"data\":{\"text\":\"The Retrieval Authority Layer is a conceptual architecture, not a universal ranking algorithm.\"},\"type\":\"paragraph\",\"tunes\":{}},{\"id\":\"G6tXMIP-u1\",\"data\":{\"text\":\"Different domains require different authority models. Legal, scientific, commercial, operational and personal data cannot all use the same source hierarchy.\"},\"type\":\"paragraph\",\"tunes\":{}},{\"id\":\"SWCg4MmSIb\",\"data\":{\"text\":\"It also cannot guarantee that an external search or AI system will rank a publisher's page. External ranking systems remain outside the publisher's control.\"},\"type\":\"paragraph\",\"tunes\":{}},{\"id\":\"j_xI-ClxxP\",\"data\":{\"text\":\"What Would Change This Answer?\",\"level\":2},\"type\":\"header\",\"tunes\":{}},{\"id\":\"xrctSdgP3p\",\"data\":{\"text\":\"Retrieval systems may increasingly learn source reliability automatically, combine structured provenance with neural ranking and negotiate access through agent protocols.\"},\"type\":\"paragraph\",\"tunes\":{}},{\"id\":\"Ka8dOdJHXO\",\"data\":{\"text\":\"That may change how authority signals are implemented. It does not remove the distinction between finding relevant information and deciding whether that information deserves evidential weight.\"},\"type\":\"paragraph\",\"tunes\":{}},{\"id\":\"E6SLB3m5CI\",\"data\":{\"text\":\"Conclusion\",\"level\":2},\"type\":\"header\",\"tunes\":{}},{\"id\":\"0_oDIxt-xd\",\"data\":{\"text\":\"The Retrieval Trigger answers: Do I need external evidence?\"},\"type\":\"paragraph\",\"tunes\":{}},{\"id\":\"mCMupYycYk\",\"data\":{\"text\":\"The Retrieval Authority Layer answers the next question: Which external source deserves to become evidence?\"},\"type\":\"paragraph\",\"tunes\":{}},{\"id\":\"rSohO6bowX\",\"data\":{\"code\":\"Model Knowledge\\n   ↓\\nRetrieval Trigger\\n   ↓\\nCandidate Sources\\n   ↓\\nRetrieval Authority Layer\\n   ↓\\nQualified Evidence\\n   ↓\\nReasoning\\n   ↓\\nAnswer\"},\"type\":\"code\",\"tunes\":{}},{\"id\":\"f8GMr_eelV\",\"data\":{\"text\":\"That distinction moves RAG from document retrieval toward controlled evidence retrieval.\"},\"type\":\"paragraph\",\"tunes\":{}},{\"id\":\"hOWp43J5Dw\",\"data\":{\"text\":\"Primary Sources\",\"level\":2},\"type\":\"header\",\"tunes\":{}},{\"id\":\"evqsbllTgC\",\"data\":{\"text\":\"Patrick Lewis et al., \u003Ca href=\\\"https:\u002F\u002Farxiv.org\u002Fabs\u002F2005.11401\\\" target=\\\"_blank\\\">Retrieval-Augmented Generation for Knowledge-Intensive NLP Tasks\u003C\u002Fa> (2020).\"},\"type\":\"paragraph\",\"tunes\":{}},{\"id\":\"gz0XrNndq4\",\"data\":{\"text\":\"Akari Asai et al., \u003Ca href=\\\"https:\u002F\u002Farxiv.org\u002Fabs\u002F2310.11511\\\" target=\\\"_blank\\\">Self-RAG: Learning to Retrieve, Generate, and Critique through Self-Reflection\u003C\u002Fa> (2023).\"},\"type\":\"paragraph\",\"tunes\":{}},{\"id\":\"NgkPqz5Poz\",\"data\":{\"text\":\"Soyeong Jeong et al., \u003Ca href=\\\"https:\u002F\u002Farxiv.org\u002Fabs\u002F2403.14403\\\" target=\\\"_blank\\\">Adaptive-RAG: Learning to Adapt Retrieval-Augmented Large Language Models through Question Complexity\u003C\u002Fa> (2024).\"},\"type\":\"paragraph\",\"tunes\":{}}],\"version\":\"2.31.6\"}",{"time":719,"blocks":720,"version":1066},1792072370632,[721,724,728,732,736,741,745,749,753,757,761,764,772,776,780,793,797,800,804,808,812,816,820,824,828,831,835,839,843,847,855,859,863,867,870,874,878,882,886,890,894,898,902,906,914,918,922,925,929,933,937,941,949,953,957,968,972,976,980,984,988,992,996,1000,1004,1007,1011,1015,1019,1023,1027,1031,1035,1038,1042,1046,1049,1053,1057,1060,1063],{"id":231,"data":722,"type":42,"tunes":723},{"text":233,"level":234},{},{"id":237,"data":725,"type":240,"tunes":727},{"text":726},"Why does an AI retrieval system choose one source while ignoring another source that appears to discuss exactly the same subject?",{},{"id":243,"data":729,"type":240,"tunes":731},{"text":730},"Semantic relevance is only part of the answer. Two documents can be equally similar to a query while differing radically in authority, freshness, provenance, scope, permissions and applicability.",{},{"id":248,"data":733,"type":240,"tunes":735},{"text":734},"This article uses the term Retrieval Authority Layer for the source-selection boundary that sits between finding potentially relevant information and treating that information as suitable evidence.",{},{"id":253,"data":737,"type":258,"tunes":740},{"text":738,"caption":739,"alignment":257},"A Retrieval Authority Layer is the set of source-selection rules and source properties that determine which retrievable information deserves to become evidence for a particular question.","Working definition — a systems concept used in this article, not a standardized research term.",{},{"id":261,"data":742,"type":265,"tunes":744},{"title":743,"maxLevel":264,"minLevel":234},"Contents",{},{"id":268,"data":746,"type":42,"tunes":748},{"text":747,"level":234},"What This Really Means",{},{"id":273,"data":750,"type":240,"tunes":752},{"text":751},"A retriever answers a relatively narrow question: Which items appear relevant to this query?",{},{"id":278,"data":754,"type":240,"tunes":756},{"text":755},"A production AI system usually needs to answer a harder one: Which of those relevant items should be allowed to influence the answer?",{},{"id":283,"data":758,"type":240,"tunes":760},{"text":759},"That second question introduces source identity, version, authority, validity period, provenance and scope. A relevant old policy, copied blog post or unofficial summary can rank extremely well while still being the wrong evidence for the decision.",{},{"id":288,"data":762,"type":291,"tunes":763},{"code":290},{},{"id":294,"data":765,"type":302,"tunes":771},{"link":766,"meta":767},"https:\u002F\u002Fstajic.de\u002Fblog\u002Fwhen-should-an-ai-stop-trusting-its-own-knowledge-the-retrieval-trigger",{"image":768,"title":769,"description":770},{"url":299},"When Should an AI Stop Trusting Its Own Knowledge? — The Retrieval Trigger","The Retrieval Trigger decides when external evidence is required. The Retrieval Authority Layer asks which sources should provide it.",{},{"id":305,"data":773,"type":42,"tunes":775},{"text":774,"level":234},"Simplest Example",{},{"id":310,"data":777,"type":240,"tunes":779},{"text":778},"Imagine an employee asks an enterprise assistant: How many vacation days do employees receive?",{},{"id":315,"data":781,"type":329,"tunes":792},{"content":782,"stretched":43,"withHeadings":14},[783,787,790],[784,785,786],"Candidate","Content","Semantic relevance",[323,788,789],"Current HR policy, approved version 6.2","High",[327,791,789],"Old onboarding guide copied from version 4.8",{},{"id":332,"data":794,"type":240,"tunes":796},{"text":795},"Both passages can contain nearly identical wording. A similarity score may therefore place them very close together.",{},{"id":337,"data":798,"type":291,"tunes":799},{"code":339},{},{"id":342,"data":801,"type":240,"tunes":803},{"text":802},"The current approved HR policy should normally dominate because it is the governing source for the claim. The older guide remains relevant text, but it should not carry the same evidential weight.",{},{"id":347,"data":805,"type":42,"tunes":807},{"text":806,"level":234},"Where the Example Stops Working",{},{"id":352,"data":809,"type":240,"tunes":811},{"text":810},"Authority is not always a universal hierarchy in which one source wins every time.",{},{"id":357,"data":813,"type":240,"tunes":815},{"text":814},"A transactional database may be authoritative for the current status of an order. A product manual may be authoritative for intended operation. A signed contract may control a commercial obligation. A primary scientific paper may be the best source for its experiment, while a later systematic review may be stronger for a broader conclusion.",{},{"id":362,"data":817,"type":240,"tunes":819},{"text":818},"Authority is therefore claim-relative. The correct source class depends on what is being asked.",{},{"id":367,"data":821,"type":42,"tunes":823},{"text":822,"level":234},"Direct Answer",{},{"id":372,"data":825,"type":240,"tunes":827},{"text":826},"A source should become preferred retrieval evidence when it is not merely relevant, but also sufficiently authoritative, current, applicable, traceable and accessible for the specific claim being answered.",{},{"id":377,"data":829,"type":291,"tunes":830},{"code":379},{},{"id":382,"data":832,"type":240,"tunes":834},{"text":833},"These properties do not guarantee truth. They create a stronger basis for deciding which retrieved information deserves to enter the reasoning process.",{},{"id":387,"data":836,"type":42,"tunes":838},{"text":837,"level":234},"Why Similarity Alone Is Not Enough",{},{"id":392,"data":840,"type":240,"tunes":842},{"text":841},"Vector retrieval is extremely useful because it can find semantically related information even when the wording differs. But semantic similarity does not know by itself whether a document is obsolete, superseded, unofficial, unauthorized or valid for another jurisdiction.",{},{"id":397,"data":844,"type":240,"tunes":846},{"text":845},"The original RAG architecture showed how external non-parametric memory can augment model knowledge. Production systems must add application-specific controls around that retrieval process.",{},{"id":402,"data":848,"type":302,"tunes":854},{"link":849,"meta":850},"https:\u002F\u002Fstajic.de\u002Fblog\u002Fwhere-does-an-llm-get-its-data-rag-data-sources-in-python",{"image":851,"title":852,"description":853},{"url":299},"Where Does an LLM Get Its Data? RAG Data Sources in Python","Files, databases, APIs, search and vector indexes can all provide runtime knowledge. The retrieval method should match the source and the question.",{},{"id":411,"data":856,"type":42,"tunes":858},{"text":857,"level":234},"The Six Authority Signals",{},{"id":416,"data":860,"type":42,"tunes":862},{"text":861,"level":264},"1. Source Identity",{},{"id":421,"data":864,"type":240,"tunes":866},{"text":865},"The system should know what the source actually is: policy, contract, API, database record, research paper, documentation page, user file or third-party commentary.",{},{"id":426,"data":868,"type":42,"tunes":869},{"text":428,"level":264},{},{"id":431,"data":871,"type":240,"tunes":873},{"text":872},"Evidence should preserve where it came from, how it entered the system and which original object or document can be inspected if the claim is challenged.",{},{"id":436,"data":875,"type":42,"tunes":877},{"text":876,"level":264},"3. Freshness",{},{"id":441,"data":879,"type":240,"tunes":881},{"text":880},"Freshness matters when the underlying fact changes. A current API response and a five-year-old article may be equally relevant to the words of a query but not to its temporal requirement.",{},{"id":446,"data":883,"type":42,"tunes":885},{"text":884,"level":264},"4. Applicability",{},{"id":451,"data":887,"type":240,"tunes":889},{"text":888},"A fact may be correct but applicable to the wrong product version, country, tenant, customer group, contract, software release or time period.",{},{"id":456,"data":891,"type":42,"tunes":893},{"text":892,"level":264},"5. Evidence Specificity",{},{"id":461,"data":895,"type":240,"tunes":897},{"text":896},"Primary evidence that directly supports the claim is generally stronger than a chain of summaries that merely repeat it.",{},{"id":466,"data":899,"type":42,"tunes":901},{"text":900,"level":264},"6. Access Legitimacy",{},{"id":471,"data":903,"type":240,"tunes":905},{"text":904},"A technically retrievable document is not necessarily authorized context. Retrieval authority must operate inside tenant, permission, confidentiality and data-minimization boundaries.",{},{"id":476,"data":907,"type":302,"tunes":913},{"link":908,"meta":909},"https:\u002F\u002Fstajic.de\u002Fblog\u002Fwhat-is-context-engineering-what-the-model-receives-before-it-answers",{"image":910,"title":911,"description":912},{"url":299},"What Is Context Engineering? What the Model Receives Before It Answers","Context assembly is also a validity and security boundary. Retrieved information must preserve scope, authority and access constraints.",{},{"id":485,"data":915,"type":42,"tunes":917},{"text":916,"level":234},"A Minimal Authority Model",{},{"id":490,"data":919,"type":240,"tunes":921},{"text":920},"A practical system can represent these properties explicitly rather than hoping that similarity ranking captures them.",{},{"id":495,"data":923,"type":291,"tunes":924},{"code":497},{},{"id":500,"data":926,"type":240,"tunes":928},{"text":927},"The retriever can still use semantic or lexical relevance. The authority layer adds constraints and reranking signals that relevance alone cannot provide.",{},{"id":505,"data":930,"type":42,"tunes":932},{"text":931,"level":234},"Connection to Agent-Ready Information",{},{"id":510,"data":934,"type":240,"tunes":936},{"text":935},"This becomes especially important when the consumer is not a human reading documents but an AI agent making decisions.",{},{"id":515,"data":938,"type":240,"tunes":940},{"text":939},"An agent-ready information object should preserve identity, provenance, conflicts, unknowns and validity conditions before the retrieval stage. Otherwise the system tries to reconstruct authority after information has already been flattened into anonymous text.",{},{"id":520,"data":942,"type":302,"tunes":948},{"link":943,"meta":944},"https:\u002F\u002Fstajic.de\u002Fportfolio\u002Ffrom-raw-data-to-verifiable-ai-context-a-practical-architecture-for-agent-ready-information",{"image":945,"title":946,"description":947},{"url":299},"From Raw Data to Verifiable AI Context: A Practical Architecture for Agent-Ready Information","A portfolio architecture for turning heterogeneous data into controlled information objects with identity, provenance, conflicts and validity.",{},{"id":529,"data":950,"type":42,"tunes":952},{"text":951,"level":234},"Diagnostic \u002F Decision Method",{},{"id":534,"data":954,"type":240,"tunes":956},{"text":955},"For every high-value retrieved claim, ask six questions:",{},{"id":539,"data":958,"type":551,"tunes":967},{"meta":959,"items":960,"style":550},{"counterType":542},[961,962,963,964,965,966],"What exact source produced this evidence?","Is this source authoritative for this type of claim?","Is the information still current?","Does it apply to this user, version, jurisdiction and time period?","Can the claim be traced back to the original evidence?","Was the user or agent authorized to retrieve it?",{},{"id":554,"data":969,"type":240,"tunes":971},{"text":970},"If the system cannot answer these questions, the problem is no longer only retrieval quality. It is an evidence-governance problem.",{},{"id":559,"data":973,"type":42,"tunes":975},{"text":974,"level":234},"Common Failure Modes",{},{"id":564,"data":977,"type":240,"tunes":979},{"text":978},"The highest similarity score wins. This confuses semantic closeness with evidential authority.",{},{"id":569,"data":981,"type":240,"tunes":983},{"text":982},"The newest source always wins. A newer commentary may be less authoritative than an older governing contract or standard.",{},{"id":574,"data":985,"type":240,"tunes":987},{"text":986},"Official means universally applicable. An official source may still belong to another jurisdiction, version or scope.",{},{"id":579,"data":989,"type":240,"tunes":991},{"text":990},"Public means safe to retrieve. Information can be public yet inappropriate for the task, while private information can be legitimate only inside an authorized context.",{},{"id":584,"data":993,"type":42,"tunes":995},{"text":994,"level":234},"Edge Case: Conflicting Authoritative Sources",{},{"id":589,"data":997,"type":240,"tunes":999},{"text":998},"The difficult case is not an authoritative source versus an obviously weak source. It is two apparently authoritative sources that disagree.",{},{"id":594,"data":1001,"type":240,"tunes":1003},{"text":1002},"At that point, the authority layer should preserve the conflict instead of silently collapsing it into one answer.",{},{"id":599,"data":1005,"type":291,"tunes":1006},{"code":601},{},{"id":604,"data":1008,"type":42,"tunes":1010},{"text":1009,"level":234},"Limitations",{},{"id":609,"data":1012,"type":240,"tunes":1014},{"text":1013},"The Retrieval Authority Layer is a conceptual architecture, not a universal ranking algorithm.",{},{"id":614,"data":1016,"type":240,"tunes":1018},{"text":1017},"Different domains require different authority models. Legal, scientific, commercial, operational and personal data cannot all use the same source hierarchy.",{},{"id":619,"data":1020,"type":240,"tunes":1022},{"text":1021},"It also cannot guarantee that an external search or AI system will rank a publisher's page. External ranking systems remain outside the publisher's control.",{},{"id":624,"data":1024,"type":42,"tunes":1026},{"text":1025,"level":234},"What Would Change This Answer?",{},{"id":629,"data":1028,"type":240,"tunes":1030},{"text":1029},"Retrieval systems may increasingly learn source reliability automatically, combine structured provenance with neural ranking and negotiate access through agent protocols.",{},{"id":634,"data":1032,"type":240,"tunes":1034},{"text":1033},"That may change how authority signals are implemented. It does not remove the distinction between finding relevant information and deciding whether that information deserves evidential weight.",{},{"id":639,"data":1036,"type":42,"tunes":1037},{"text":641,"level":234},{},{"id":644,"data":1039,"type":240,"tunes":1041},{"text":1040},"The Retrieval Trigger answers: Do I need external evidence?",{},{"id":649,"data":1043,"type":240,"tunes":1045},{"text":1044},"The Retrieval Authority Layer answers the next question: Which external source deserves to become evidence?",{},{"id":654,"data":1047,"type":291,"tunes":1048},{"code":656},{},{"id":659,"data":1050,"type":240,"tunes":1052},{"text":1051},"That distinction moves RAG from document retrieval toward controlled evidence retrieval.",{},{"id":664,"data":1054,"type":42,"tunes":1056},{"text":1055,"level":234},"Primary Sources",{},{"id":669,"data":1058,"type":240,"tunes":1059},{"text":671},{},{"id":674,"data":1061,"type":240,"tunes":1062},{"text":676},{},{"id":679,"data":1064,"type":240,"tunes":1065},{"text":681},{},"2.31.6","Finding a relevant document is not enough. AI systems also need to distinguish sources that are current, authoritative, applicable and traceable from sources that merely look similar. The Retrieval Authority Layer defines that source-selection boundary.",{"lang":7,"title":224,"content":226,"contentJson":1069,"excerpt":684},{"time":228,"blocks":1070,"version":683},[1071,1074,1077,1080,1083,1086,1089,1092,1095,1098,1101,1104,1109,1112,1115,1122,1125,1128,1131,1134,1137,1140,1143,1146,1149,1152,1155,1158,1161,1164,1169,1172,1175,1178,1181,1184,1187,1190,1193,1196,1199,1202,1205,1208,1213,1216,1219,1222,1225,1228,1231,1234,1239,1242,1245,1250,1253,1256,1259,1262,1265,1268,1271,1274,1277,1280,1283,1286,1289,1292,1295,1298,1301,1304,1307,1310,1313,1316,1319,1322,1325],{"id":231,"data":1072,"type":42,"tunes":1073},{"text":233,"level":234},{},{"id":237,"data":1075,"type":240,"tunes":1076},{"text":239},{},{"id":243,"data":1078,"type":240,"tunes":1079},{"text":245},{},{"id":248,"data":1081,"type":240,"tunes":1082},{"text":250},{},{"id":253,"data":1084,"type":258,"tunes":1085},{"text":255,"caption":256,"alignment":257},{},{"id":261,"data":1087,"type":265,"tunes":1088},{"title":263,"maxLevel":264,"minLevel":234},{},{"id":268,"data":1090,"type":42,"tunes":1091},{"text":270,"level":234},{},{"id":273,"data":1093,"type":240,"tunes":1094},{"text":275},{},{"id":278,"data":1096,"type":240,"tunes":1097},{"text":280},{},{"id":283,"data":1099,"type":240,"tunes":1100},{"text":285},{},{"id":288,"data":1102,"type":291,"tunes":1103},{"code":290},{},{"id":294,"data":1105,"type":302,"tunes":1108},{"link":296,"meta":1106},{"image":1107,"title":300,"description":301},{"url":299},{},{"id":305,"data":1110,"type":42,"tunes":1111},{"text":307,"level":234},{},{"id":310,"data":1113,"type":240,"tunes":1114},{"text":312},{},{"id":315,"data":1116,"type":329,"tunes":1121},{"content":1117,"stretched":43,"withHeadings":14},[1118,1119,1120],[319,320,321],[323,324,325],[327,328,325],{},{"id":332,"data":1123,"type":240,"tunes":1124},{"text":334},{},{"id":337,"data":1126,"type":291,"tunes":1127},{"code":339},{},{"id":342,"data":1129,"type":240,"tunes":1130},{"text":344},{},{"id":347,"data":1132,"type":42,"tunes":1133},{"text":349,"level":234},{},{"id":352,"data":1135,"type":240,"tunes":1136},{"text":354},{},{"id":357,"data":1138,"type":240,"tunes":1139},{"text":359},{},{"id":362,"data":1141,"type":240,"tunes":1142},{"text":364},{},{"id":367,"data":1144,"type":42,"tunes":1145},{"text":369,"level":234},{},{"id":372,"data":1147,"type":240,"tunes":1148},{"text":374},{},{"id":377,"data":1150,"type":291,"tunes":1151},{"code":379},{},{"id":382,"data":1153,"type":240,"tunes":1154},{"text":384},{},{"id":387,"data":1156,"type":42,"tunes":1157},{"text":389,"level":234},{},{"id":392,"data":1159,"type":240,"tunes":1160},{"text":394},{},{"id":397,"data":1162,"type":240,"tunes":1163},{"text":399},{},{"id":402,"data":1165,"type":302,"tunes":1168},{"link":404,"meta":1166},{"image":1167,"title":407,"description":408},{"url":299},{},{"id":411,"data":1170,"type":42,"tunes":1171},{"text":413,"level":234},{},{"id":416,"data":1173,"type":42,"tunes":1174},{"text":418,"level":264},{},{"id":421,"data":1176,"type":240,"tunes":1177},{"text":423},{},{"id":426,"data":1179,"type":42,"tunes":1180},{"text":428,"level":264},{},{"id":431,"data":1182,"type":240,"tunes":1183},{"text":433},{},{"id":436,"data":1185,"type":42,"tunes":1186},{"text":438,"level":264},{},{"id":441,"data":1188,"type":240,"tunes":1189},{"text":443},{},{"id":446,"data":1191,"type":42,"tunes":1192},{"text":448,"level":264},{},{"id":451,"data":1194,"type":240,"tunes":1195},{"text":453},{},{"id":456,"data":1197,"type":42,"tunes":1198},{"text":458,"level":264},{},{"id":461,"data":1200,"type":240,"tunes":1201},{"text":463},{},{"id":466,"data":1203,"type":42,"tunes":1204},{"text":468,"level":264},{},{"id":471,"data":1206,"type":240,"tunes":1207},{"text":473},{},{"id":476,"data":1209,"type":302,"tunes":1212},{"link":478,"meta":1210},{"image":1211,"title":481,"description":482},{"url":299},{},{"id":485,"data":1214,"type":42,"tunes":1215},{"text":487,"level":234},{},{"id":490,"data":1217,"type":240,"tunes":1218},{"text":492},{},{"id":495,"data":1220,"type":291,"tunes":1221},{"code":497},{},{"id":500,"data":1223,"type":240,"tunes":1224},{"text":502},{},{"id":505,"data":1226,"type":42,"tunes":1227},{"text":507,"level":234},{},{"id":510,"data":1229,"type":240,"tunes":1230},{"text":512},{},{"id":515,"data":1232,"type":240,"tunes":1233},{"text":517},{},{"id":520,"data":1235,"type":302,"tunes":1238},{"link":522,"meta":1236},{"image":1237,"title":525,"description":526},{"url":299},{},{"id":529,"data":1240,"type":42,"tunes":1241},{"text":531,"level":234},{},{"id":534,"data":1243,"type":240,"tunes":1244},{"text":536},{},{"id":539,"data":1246,"type":551,"tunes":1249},{"meta":1247,"items":1248,"style":550},{"counterType":542},[544,545,546,547,548,549],{},{"id":554,"data":1251,"type":240,"tunes":1252},{"text":556},{},{"id":559,"data":1254,"type":42,"tunes":1255},{"text":561,"level":234},{},{"id":564,"data":1257,"type":240,"tunes":1258},{"text":566},{},{"id":569,"data":1260,"type":240,"tunes":1261},{"text":571},{},{"id":574,"data":1263,"type":240,"tunes":1264},{"text":576},{},{"id":579,"data":1266,"type":240,"tunes":1267},{"text":581},{},{"id":584,"data":1269,"type":42,"tunes":1270},{"text":586,"level":234},{},{"id":589,"data":1272,"type":240,"tunes":1273},{"text":591},{},{"id":594,"data":1275,"type":240,"tunes":1276},{"text":596},{},{"id":599,"data":1278,"type":291,"tunes":1279},{"code":601},{},{"id":604,"data":1281,"type":42,"tunes":1282},{"text":606,"level":234},{},{"id":609,"data":1284,"type":240,"tunes":1285},{"text":611},{},{"id":614,"data":1287,"type":240,"tunes":1288},{"text":616},{},{"id":619,"data":1290,"type":240,"tunes":1291},{"text":621},{},{"id":624,"data":1293,"type":42,"tunes":1294},{"text":626,"level":234},{},{"id":629,"data":1296,"type":240,"tunes":1297},{"text":631},{},{"id":634,"data":1299,"type":240,"tunes":1300},{"text":636},{},{"id":639,"data":1302,"type":42,"tunes":1303},{"text":641,"level":234},{},{"id":644,"data":1305,"type":240,"tunes":1306},{"text":646},{},{"id":649,"data":1308,"type":240,"tunes":1309},{"text":651},{},{"id":654,"data":1311,"type":291,"tunes":1312},{"code":656},{},{"id":659,"data":1314,"type":240,"tunes":1315},{"text":661},{},{"id":664,"data":1317,"type":42,"tunes":1318},{"text":666,"level":234},{},{"id":669,"data":1320,"type":240,"tunes":1321},{"text":671},{},{"id":674,"data":1323,"type":240,"tunes":1324},{"text":676},{},{"id":679,"data":1326,"type":240,"tunes":1327},{"text":681},{},"Post erfolgreich abgerufen",{"items":1330,"source":1351,"manualIds":1352,"manualMatchedIds":1353},[1331,1338,1344],{"id":1332,"slug":1333,"title":1334,"excerpt":1335,"featuredImage":1336,"publishedAt":1337},"384","new-qwen-3-5-plus","Nouveau Qwen 3.5-Plus : l'IA open-source passe aux choses sérieuses","Découvrez les fonctionnalités et avantages révolutionnaires de Qwen 3.5-Plus d'Alibaba, une IA open-source qui change la donne pour les développeurs.","\u002Fuploads\u002F2026\u002F02\u002Fnew-qwen-3-5-plus-1771515512741-dcbi9p.webp","2026-02-19T10:23:00.000Z",{"id":1339,"slug":1340,"title":1341,"excerpt":1342,"featuredImage":1336,"publishedAt":1343},"445","qwen-3-6-in-production-release-runbook-ai-rollback-and-llmops-versioning","Qwen 3.6 en production : Runbook de déploiement, Rollback IA et Versionnage LLMOps","Qwen 3.6 n'est pas seulement une autre mise à jour de modèle. C'est à la fois un événement de déploiement, un scénario de rollback et un problème de versionnage. Cet article explique comment Qwen 3.6 doit être géré en production à travers la discipline LLMOps, la traçabilité des prompts et des modèles, le déploiement contrôlé et une préparation au rollback basée sur des preuves.","2026-05-04T02:49:00.000Z",{"id":1345,"slug":1346,"title":1347,"excerpt":1348,"featuredImage":1349,"publishedAt":1350},"486","source-of-truth-in-ai-systems-where-reliable-knowledge-actually-comes-from","Source de vérité dans les systèmes d’IA : d’où provient réellement une connaissance fiable","Une source de vérité définit quelle source fait autorité pour un fait ou un état spécifique. Découvrez en quoi elle diffère du RAG, de la provenance, de la mémoire, du contexte, des bases de données vectorielles et des systèmes d'enregistrement.","\u002Fuploads\u002F2026\u002F10\u002Fsource-of-truth-in-ai-systems-where-reliable-knowledge-actually-comes-from-1791479103235-6bq9em.webp","2026-10-08T13:02:00.000Z","fallback",[],[]]