[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"portal-settings:stajic:it":3,"public-menus:all":38,"post:what-is-rag-the-simplest-explanation-of-how-it-works:it":205,"related:post:what-is-rag-the-simplest-explanation-of-how-it-works:it:1":1798},{"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","it","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,[148,161,175,181,193],{"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-folder","custom",[],{"id":176,"title":177,"url":179,"target":61,"icon":172,"isActive":14,"type":173,"productId":10,"categoryId":10,"shopCategoryId":10,"articleId":10,"pageId":10,"portfolioId":10,"children":180},"item-28",{"de":178,"en":178,"es":178,"fr":178,"it":178,"ru":178,"sr":178,"zh":178},"Solr Suggester","\u002Fportfolio\u002Fsolr-fuzzy-suggester-und-solr-infix-suggester-abfrage-ueber-ajax-und-filterung",[],{"id":182,"title":183,"url":191,"target":61,"icon":172,"isActive":14,"type":173,"productId":10,"categoryId":10,"shopCategoryId":10,"articleId":10,"pageId":10,"portfolioId":10,"children":192},"item-27",{"de":184,"en":185,"es":186,"fr":187,"it":188,"ru":189,"sr":190,"zh":185},"Firmenwebseite SEO","Company Website SEO","Sitio web corporativo SEO","Site web d’entreprise SEO","Sito web aziendale SEO","Корпоративный сайт SEO","Пословна веб-страница SEO","\u002Fportfolio\u002Fseo-sem-branding-mobile-webseite-muenchen",[],{"id":194,"title":195,"url":203,"target":61,"icon":172,"isActive":14,"type":173,"productId":10,"categoryId":10,"shopCategoryId":10,"articleId":10,"pageId":10,"portfolioId":10,"children":204},"item-31",{"de":196,"en":197,"es":198,"fr":199,"it":200,"ru":201,"sr":202,"zh":197},"Digitalisierungsportal","Digitalization Portal","Portal de digitalización","Portail de numérisation","Portale di digitalizzazione","Портал цифровизации","Портал за дигитализацију","\u002Fportfolio\u002Fdigitalisierungsportal-archiv-museum-bibliothek-ead-lido-mets-mods",[],{"statusCode":4,"data":206,"message":1797},{"id":207,"title":208,"slug":209,"content":210,"contentJson":211,"excerpt":874,"featuredImage":875,"featuredImageAlt":876,"featuredImageCaption":10,"featuredImageTitle":10,"featuredImageCopyright":10,"featuredImageAuthor":10,"featuredImageSourceUrl":10,"featuredImageLicense":10,"featuredImageIsAiGenerated":43,"status":877,"publishedAt":878,"createdAt":879,"updatedAt":880,"seoLocalePaths":881,"categories":890,"author":915,"translations":920},"478","Cos'è il RAG? La spiegazione più semplice di come funziona","what-is-rag-the-simplest-explanation-of-how-it-works","\u003Cp>RAG sembra complicato perché il nome è complicato. L'idea non lo è. RAG significa semplicemente: prima che l'IA risponda, cerca prima le informazioni rilevanti da una fonte di conoscenza e fornisce tali informazioni al modello linguistico.\u003C\u002Fp>\n\u003Caside class=\"editorjs-callout editorjs-callout--info my-6 rounded-xl border p-5 border-blue-300 bg-blue-50 dark:border-blue-900 dark:bg-blue-950\u002F20\" role=\"note\">\u003Cstrong class=\"block mb-2 text-gray-900 dark:text-gray-100\">RAG in una frase\u003C\u002Fstrong>\u003Cdiv class=\"text-gray-700 dark:text-gray-200\">\u003Cstrong>RAG è il passaggio in cui un'IA cerca in una base di conoscenza informazioni utili prima che l'LLM scriva la risposta.\u003C\u002Fstrong>\u003C\u002Fdiv>\u003C\u002Faside>\n\u003Cp>Pensa a un LLM come a una persona intelligente seduta a una scrivania. RAG è il bibliotecario che porta la pagina giusta dal libro giusto. L'LLM poi legge quella pagina e ti risponde.\u003C\u002Fp>\n\u003Cnav class=\"editorjs-toc\" data-editorjs-toc=\"true\" aria-label=\"Contenuti\">\u003Cstrong class=\"editorjs-toc__title\">Contenuti\u003C\u002Fstrong>\u003Col class=\"editorjs-toc__list editorjs-toc__list--depth-0\">\u003Cli class=\"editorjs-toc__item\">\u003Ca href=\"#section-5\" class=\"editorjs-toc__link\">Prima: cosa fa l'LLM?\u003C\u002Fa>\u003C\u002Fli>\u003Cli class=\"editorjs-toc__item\">\u003Ca href=\"#section-10\" class=\"editorjs-toc__link\">Poi: cos'è la base di conoscenza?\u003C\u002Fa>\u003C\u002Fli>\u003Cli class=\"editorjs-toc__item\">\u003Ca href=\"#section-15\" class=\"editorjs-toc__link\">Quindi cosa fa effettivamente RAG?\u003C\u002Fa>\u003C\u002Fli>\u003Cli class=\"editorjs-toc__item\">\u003Ca href=\"#section-19\" class=\"editorjs-toc__link\">Un esempio molto semplice\u003C\u002Fa>\u003C\u002Fli>\u003Cli class=\"editorjs-toc__item\">\u003Ca href=\"#section-25\" class=\"editorjs-toc__link\">Ora la parte importante: RAG non è lo stato attuale\u003C\u002Fa>\u003C\u002Fli>\u003Cli class=\"editorjs-toc__item\">\u003Ca href=\"#section-30\" class=\"editorjs-toc__link\">Cos'è un database di stato?\u003C\u002Fa>\u003C\u002Fli>\u003Cli class=\"editorjs-toc__item\">\u003Ca href=\"#section-36\" class=\"editorjs-toc__link\">Come i tre pezzi lavorano insieme\u003C\u002Fa>\u003C\u002Fli>\u003Cli class=\"editorjs-toc__item\">\u003Ca href=\"#section-40\" class=\"editorjs-toc__link\">Il RAG utilizza sempre un database vettoriale?\u003C\u002Fa>\u003C\u002Fli>\u003Cli class=\"editorjs-toc__item\">\u003Ca href=\"#section-45\" class=\"editorjs-toc__link\">Che cos'è un embedding, in parole semplici?\u003C\u002Fa>\u003C\u002Fli>\u003Cli class=\"editorjs-toc__item\">\u003Ca href=\"#section-50\" class=\"editorjs-toc__link\">Anche il RAG non è memoria\u003C\u002Fa>\u003C\u002Fli>\u003Cli class=\"editorjs-toc__item\">\u003Ca href=\"#section-54\" class=\"editorjs-toc__link\">Un esempio reale di gioco: PUBG Ally\u003C\u002Fa>\u003C\u002Fli>\u003Cli class=\"editorjs-toc__item\">\u003Ca href=\"#section-60\" class=\"editorjs-toc__link\">Un esempio completo\u003C\u002Fa>\u003C\u002Fli>\u003Cli class=\"editorjs-toc__item\">\u003Ca href=\"#section-64\" class=\"editorjs-toc__link\">Perché usare RAG?\u003C\u002Fa>\u003C\u002Fli>\u003Cli class=\"editorjs-toc__item\">\u003Ca href=\"#section-68\" class=\"editorjs-toc__link\">Cosa non garantisce RAG\u003C\u002Fa>\u003C\u002Fli>\u003Cli class=\"editorjs-toc__item\">\u003Ca href=\"#section-72\" class=\"editorjs-toc__link\">Il modello mentale più facile da ricordare\u003C\u002Fa>\u003C\u002Fli>\u003Cli class=\"editorjs-toc__item\">\u003Ca href=\"#section-75\" class=\"editorjs-toc__link\">Conclusione\u003C\u002Fa>\u003C\u002Fli>\u003Cli class=\"editorjs-toc__item\">\u003Ca href=\"#section-79\" class=\"editorjs-toc__link\">FAQ\u003C\u002Fa>\u003C\u002Fli>\u003Cli class=\"editorjs-toc__item\">\u003Ca href=\"#section-81\" class=\"editorjs-toc__link\">Glossario\u003C\u002Fa>\u003C\u002Fli>\u003Cli class=\"editorjs-toc__item\">\u003Ca href=\"#section-83\" class=\"editorjs-toc__link\">Fonti primarie\u003C\u002Fa>\u003C\u002Fli>\u003C\u002Fol>\u003C\u002Fnav>\n\u003Ch2 id=\"section-5\">Prima: cosa fa l'LLM?\u003C\u002Fh2>\n\u003Cp>L'LLM è la parte che comprende il linguaggio e produce linguaggio. Può leggere la tua domanda, comprendere le istruzioni, confrontare informazioni, spiegare qualcosa e scrivere una risposta.\u003C\u002Fp>\n\u003Cp>Ma l'LLM non sa automaticamente cosa c'è attualmente nel database della tua azienda, nella tua sessione di gioco, nei tuoi documenti privati o in un file che hai creato cinque minuti fa.\u003C\u002Fp>\n\u003Cp>Sa solo ciò che è già all'interno del modello più qualsiasi informazione che l'applicazione gli fornisce nella richiesta corrente.\u003C\u002Fp>\n\u003Caside class=\"editorjs-callout editorjs-callout--note my-6 rounded-xl border p-5 border-gray-300 bg-gray-50 dark:border-gray-700 dark:bg-gray-900\u002F40\" role=\"note\">\u003Cstrong class=\"block mb-2 text-gray-900 dark:text-gray-100\">Regola semplice\u003C\u002Fstrong>\u003Cdiv class=\"text-gray-700 dark:text-gray-200\">L'LLM \u003Cstrong>pensa e scrive\u003C\u002Fstrong>. Non possiede automaticamente tutti i tuoi dati attuali.\u003C\u002Fdiv>\u003C\u002Faside>\n\u003Ch2 id=\"section-10\">Poi: cos'è la base di conoscenza?\u003C\u002Fh2>\n\u003Cp>Una base di conoscenza è semplicemente informazione che l'applicazione può cercare.\u003C\u002Fp>\n\u003Cp>Potrebbe contenere PDF, manuali, documentazione di prodotto, articoli di supporto, contratti, regole di gioco, dati sulle armi, documenti aziendali interni, record di database o altro testo.\u003C\u002Fp>\n\u003Cp>La base di conoscenza può essere locale sulla tua macchina. Può essere su un server. Può essere in un database vettoriale. Può anche essere costruita da file normali. RAG non significa Internet.\u003C\u002Fp>\n\u003Caside class=\"editorjs-callout editorjs-callout--success my-6 rounded-xl border p-5 border-emerald-300 bg-emerald-50 dark:border-emerald-900 dark:bg-emerald-950\u002F20\" role=\"note\">\u003Cstrong class=\"block mb-2 text-gray-900 dark:text-gray-100\">Importante\u003C\u002Fstrong>\u003Cdiv class=\"text-gray-700 dark:text-gray-200\">\u003Cstrong>RAG non richiede Internet.\u003C\u002Fstrong> Le informazioni possono essere completamente locali.\u003C\u002Fdiv>\u003C\u002Faside>\n\u003Ch2 id=\"section-15\">Quindi cosa fa effettivamente RAG?\u003C\u002Fh2>\n\u003Csection class=\"editorjs-process my-6\">\u003Ch3 class=\"mb-3 text-lg font-semibold\">L'intero processo RAG\u003C\u002Fh3>\u003Cdiv class=\"flex flex-col sm:flex-row gap-3\">\u003Cdiv class=\"editorjs-process__step min-w-0 flex-1 rounded-xl border border-gray-200 dark:border-gray-700 p-4\">\u003Cdiv class=\"text-xs font-semibold text-gray-500 dark:text-gray-400\">1\u003C\u002Fdiv>\u003Cdiv class=\"mt-1 font-semibold text-gray-900 dark:text-gray-100\">1. Fai una domanda\u003C\u002Fdiv>\u003Cdiv class=\"mt-1 text-sm text-gray-600 dark:text-gray-300\">Ad esempio: quali munizioni usa quest'arma?\u003C\u002Fdiv>\u003C\u002Fdiv>\u003Cdiv class=\"editorjs-process__arrow shrink-0 self-center text-xl text-gray-400 rotate-90 sm:rotate-0\" aria-hidden=\"true\">→\u003C\u002Fdiv>\u003Cdiv class=\"editorjs-process__step min-w-0 flex-1 rounded-xl border border-gray-200 dark:border-gray-700 p-4\">\u003Cdiv class=\"text-xs font-semibold text-gray-500 dark:text-gray-400\">2\u003C\u002Fdiv>\u003Cdiv class=\"mt-1 font-semibold text-gray-900 dark:text-gray-100\">2. RAG cerca nella base di conoscenza\u003C\u002Fdiv>\u003Cdiv class=\"mt-1 text-sm text-gray-600 dark:text-gray-300\">Il sistema cerca i piccoli pezzi di informazione più rilevanti per la tua domanda.\u003C\u002Fdiv>\u003C\u002Fdiv>\u003Cdiv class=\"editorjs-process__arrow shrink-0 self-center text-xl text-gray-400 rotate-90 sm:rotate-0\" aria-hidden=\"true\">→\u003C\u002Fdiv>\u003Cdiv class=\"editorjs-process__step min-w-0 flex-1 rounded-xl border border-gray-200 dark:border-gray-700 p-4\">\u003Cdiv class=\"text-xs font-semibold text-gray-500 dark:text-gray-400\">3\u003C\u002Fdiv>\u003Cdiv class=\"mt-1 font-semibold text-gray-900 dark:text-gray-100\">3. RAG fornisce quei pezzi all'LLM\u003C\u002Fdiv>\u003Cdiv class=\"mt-1 text-sm text-gray-600 dark:text-gray-300\">L'LLM riceve la domanda più le informazioni recuperate.\u003C\u002Fdiv>\u003C\u002Fdiv>\u003Cdiv class=\"editorjs-process__arrow shrink-0 self-center text-xl text-gray-400 rotate-90 sm:rotate-0\" aria-hidden=\"true\">→\u003C\u002Fdiv>\u003Cdiv class=\"editorjs-process__step min-w-0 flex-1 rounded-xl border border-gray-200 dark:border-gray-700 p-4\">\u003Cdiv class=\"text-xs font-semibold text-gray-500 dark:text-gray-400\">4\u003C\u002Fdiv>\u003Cdiv class=\"mt-1 font-semibold text-gray-900 dark:text-gray-100\">4. L'LLM scrive la risposta\u003C\u002Fdiv>\u003Cdiv class=\"mt-1 text-sm text-gray-600 dark:text-gray-300\">Usa le informazioni recuperate come contesto per la risposta.\u003C\u002Fdiv>\u003C\u002Fdiv>\u003C\u002Fdiv>\u003C\u002Fsection>\n\u003Cp>Questo è RAG.\u003C\u002Fp>\n\u003Cp>Il nome completo è Retrieval-Augmented Generation. Retrieval significa trovare le informazioni rilevanti. Augmented significa aggiungere quelle informazioni al contesto del modello. Generation significa che l'LLM scrive la risposta finale.\u003C\u002Fp>\n\u003Ch2 id=\"section-19\">Un esempio molto semplice\u003C\u002Fh2>\n\u003Cp>Immagina di avere una base di conoscenza locale su un gioco.\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\">La base di conoscenza contiene\u003C\u002Fth>\u003Cth class=\"border border-gray-300 px-4 py-2 text-left font-semibold\">Esempio\u003C\u002Fth>\u003C\u002Ftr>\u003C\u002Fthead>\u003Ctbody>\u003Ctr>\u003Ctd class=\"border border-gray-300 px-4 py-2\">Armi\u003C\u002Ftd>\u003Ctd class=\"border border-gray-300 px-4 py-2\">L'AKM usa munizioni da 7,62 mm\u003C\u002Ftd>\u003C\u002Ftr>\u003Ctr>\u003Ctd class=\"border border-gray-300 px-4 py-2\">Oggetti curativi\u003C\u002Ftd>\u003Ctd class=\"border border-gray-300 px-4 py-2\">Il Med Kit ripristina la salute\u003C\u002Ftd>\u003C\u002Ftr>\u003Ctr>\u003Ctd class=\"border border-gray-300 px-4 py-2\">Accessori\u003C\u002Ftd>\u003Ctd class=\"border border-gray-300 px-4 py-2\">Questo accessorio funziona con queste armi\u003C\u002Ftd>\u003C\u002Ftr>\u003Ctr>\u003Ctd class=\"border border-gray-300 px-4 py-2\">Regole della mappa\u003C\u002Ftd>\u003Ctd class=\"border border-gray-300 px-4 py-2\">Questa zona si comporta in questo modo\u003C\u002Ftd>\u003C\u002Ftr>\u003C\u002Ftbody>\u003C\u002Ftable>\u003C\u002Fdiv>\n\u003Cp>Chiedi: \"Quali munizioni usa l'AKM?\"\u003C\u002Fp>\n\u003Cp>RAG cerca nella base di conoscenza e trova la voce relativa all'AKM. Fornisce quel piccolo pezzo di informazione all'LLM. L'LLM poi risponde: \"L'AKM usa munizioni da 7,62 mm.\"\u003C\u002Fp>\n\u003Cp>L'LLM non aveva bisogno dell'intero database. RAG ha portato solo la parte utile.\u003C\u002Fp>\n\u003Ch2 id=\"section-25\">Ora la parte importante: RAG non è lo stato attuale\u003C\u002Fh2>\n\u003Cp>Qui è dove molte spiegazioni diventano confuse.\u003C\u002Fp>\n\u003Cp>RAG di solito fornisce conoscenza all'IA. Un sistema di stato fornisce all'IA fatti su ciò che è vero in questo momento.\u003C\u002Fp>\n\u003Csection class=\"editorjs-comparison my-6\">\u003Ch3 class=\"mb-3 text-lg font-semibold\">Conoscenza vs stato attuale\u003C\u002Fh3>\u003Cdiv class=\"overflow-x-auto\">\u003Ctable class=\"w-full border-collapse\">\u003Cthead>\u003Ctr>\u003Cth class=\"border border-gray-300 bg-gray-50 px-4 py-3 text-left dark:border-gray-700 dark:bg-gray-900\">\u003C\u002Fth>\u003Cth class=\"border border-gray-300 bg-gray-50 px-4 py-3 text-left font-semibold dark:border-gray-700 dark:bg-gray-900\">RAG \u002F conoscenza\u003C\u002Fth>\u003Cth class=\"border border-gray-300 bg-gray-50 px-4 py-3 text-left font-semibold dark:border-gray-700 dark:bg-gray-900\">Stato attuale\u003C\u002Fth>\u003C\u002Ftr>\u003C\u002Fthead>\u003Ctbody>\u003Ctr>\u003Cth class=\"border border-gray-300 px-4 py-3 text-left font-semibold dark:border-gray-700\">Arma\u003C\u002Fth>\u003Ctd class=\"border border-gray-300 px-4 py-3 dark:border-gray-700\">\u003C\u002Ftd>\u003Ctd class=\"border border-gray-300 px-4 py-3 dark:border-gray-700\">\u003C\u002Ftd>\u003C\u002Ftr>\u003Ctr>\u003Cth class=\"border border-gray-300 px-4 py-3 text-left font-semibold dark:border-gray-700\">Munizioni\u003C\u002Fth>\u003Ctd class=\"border border-gray-300 px-4 py-3 dark:border-gray-700\">\u003C\u002Ftd>\u003Ctd class=\"border border-gray-300 px-4 py-3 dark:border-gray-700\">\u003C\u002Ftd>\u003C\u002Ftr>\u003Ctr>\u003Cth class=\"border border-gray-300 px-4 py-3 text-left font-semibold dark:border-gray-700\">Salute\u003C\u002Fth>\u003Ctd class=\"border border-gray-300 px-4 py-3 dark:border-gray-700\">\u003C\u002Ftd>\u003Ctd class=\"border border-gray-300 px-4 py-3 dark:border-gray-700\">\u003C\u002Ftd>\u003C\u002Ftr>\u003Ctr>\u003Cth class=\"border border-gray-300 px-4 py-3 text-left font-semibold dark:border-gray-700\">Nemico\u003C\u002Fth>\u003Ctd class=\"border border-gray-300 px-4 py-3 dark:border-gray-700\">\u003C\u002Ftd>\u003Ctd class=\"border border-gray-300 px-4 py-3 dark:border-gray-700\">\u003C\u002Ftd>\u003C\u002Ftr>\u003C\u002Ftbody>\u003C\u002Ftable>\u003C\u002Fdiv>\u003C\u002Fsection>\n\u003Caside class=\"editorjs-callout editorjs-callout--warning my-6 rounded-xl border p-5 border-amber-300 bg-amber-50 dark:border-amber-900 dark:bg-amber-950\u002F20\" role=\"note\">\u003Cstrong class=\"block mb-2 text-gray-900 dark:text-gray-100\">Non mescolare questi due\u003C\u002Fstrong>\u003Cdiv class=\"text-gray-700 dark:text-gray-200\">RAG risponde: \u003Cstrong>Cosa è generalmente vero?\u003C\u002Fstrong>\u003Cbr>Lo stato risponde: \u003Cstrong>Cosa è vero in questo momento?\u003C\u002Fstrong>\u003C\u002Fdiv>\u003C\u002Faside>\n\u003Ch2 id=\"section-30\">Cos'è un database di stato?\u003C\u002Fh2>\n\u003Cp>Un database di stato o archivio di stato è semplicemente un luogo dove l'applicazione conserva i fatti attuali.\u003C\u002Fp>\n\u003Cp>In un gioco, il motore sa già cose come la tua salute, posizione, inventario, munizioni, missione corrente, oggetti vicini e stato dei nemici. Un sistema di IA può esporre parti selezionate di quello stato al modello.\u003C\u002Fp>\n\u003Cp>In un'applicazione aziendale, la stessa idea potrebbe essere un database di ordini, un record cliente, lo stato di un progetto o il valore corrente di un sensore.\u003C\u002Fp>\n\u003Cp>Lo stato viene creato dall'applicazione stessa man mano che accadono le cose. Se perdi salute, il gioco aggiorna il valore della salute. Se raccogli munizioni, l'inventario cambia. Se un ordine viene pagato, il sistema aziendale cambia lo stato dell'ordine.\u003C\u002Fp>\n\u003Caside class=\"editorjs-callout editorjs-callout--info my-6 rounded-xl border p-5 border-blue-300 bg-blue-50 dark:border-blue-900 dark:bg-blue-950\u002F20\" role=\"note\">\u003Cstrong class=\"block mb-2 text-gray-900 dark:text-gray-100\">Regola semplice\u003C\u002Fstrong>\u003Cdiv class=\"text-gray-700 dark:text-gray-200\">L'applicazione crea e aggiorna lo \u003Cstrong>stato\u003C\u002Fstrong>. RAG cerca la \u003Cstrong>conoscenza\u003C\u002Fstrong>. L'LLM usa entrambi per decidere cosa dire o fare.\u003C\u002Fdiv>\u003C\u002Faside>\n\u003Ch2 id=\"section-36\">Come i tre pezzi lavorano insieme\u003C\u002Fh2>\n\u003Csection class=\"editorjs-process my-6\">\u003Ch3 class=\"mb-3 text-lg font-semibold\">LLM + stato + RAG\u003C\u002Fh3>\u003Cdiv class=\"grid grid-cols-1 md:grid-cols-2 xl:grid-cols-3 gap-4\">\u003Cdiv class=\"editorjs-process__step min-w-0  rounded-xl border border-gray-200 dark:border-gray-700 p-4\">\u003Cdiv class=\"text-xs font-semibold text-gray-500 dark:text-gray-400\">1\u003C\u002Fdiv>\u003Cdiv class=\"mt-1 font-semibold text-gray-900 dark:text-gray-100\">1. Stato attuale\u003C\u002Fdiv>\u003Cdiv class=\"mt-1 text-sm text-gray-600 dark:text-gray-300\">L'applicazione comunica all'IA ciò che è vero in questo momento: salute 41%, AKM equipaggiato, 23 colpi.\u003C\u002Fdiv>\u003C\u002Fdiv>\u003Cdiv class=\"editorjs-process__step min-w-0  rounded-xl border border-gray-200 dark:border-gray-700 p-4\">\u003Cdiv class=\"text-xs font-semibold text-gray-500 dark:text-gray-400\">2\u003C\u002Fdiv>\u003Cdiv class=\"mt-1 font-semibold text-gray-900 dark:text-gray-100\">2. RAG\u003C\u002Fdiv>\u003Cdiv class=\"mt-1 text-sm text-gray-600 dark:text-gray-300\">Il sistema recupera conoscenze utili: come funziona l'arma, quale oggetto curativo è disponibile o una regola pertinente.\u003C\u002Fdiv>\u003C\u002Fdiv>\u003Cdiv class=\"editorjs-process__step min-w-0  rounded-xl border border-gray-200 dark:border-gray-700 p-4\">\u003Cdiv class=\"text-xs font-semibold text-gray-500 dark:text-gray-400\">3\u003C\u002Fdiv>\u003Cdiv class=\"mt-1 font-semibold text-gray-900 dark:text-gray-100\">3. LLM\u003C\u002Fdiv>\u003Cdiv class=\"mt-1 text-sm text-gray-600 dark:text-gray-300\">Il modello riceve la domanda, lo stato attuale e le conoscenze recuperate.\u003C\u002Fdiv>\u003C\u002Fdiv>\u003Cdiv class=\"editorjs-process__step min-w-0  rounded-xl border border-gray-200 dark:border-gray-700 p-4\">\u003Cdiv class=\"text-xs font-semibold text-gray-500 dark:text-gray-400\">4\u003C\u002Fdiv>\u003Cdiv class=\"mt-1 font-semibold text-gray-900 dark:text-gray-100\">4. Ragionamento\u003C\u002Fdiv>\u003Cdiv class=\"mt-1 text-sm text-gray-600 dark:text-gray-300\">L'LLM combina questi input e decide quale risposta o azione di alto livello ha senso.\u003C\u002Fdiv>\u003C\u002Fdiv>\u003Cdiv class=\"editorjs-process__step min-w-0  rounded-xl border border-gray-200 dark:border-gray-700 p-4\">\u003Cdiv class=\"text-xs font-semibold text-gray-500 dark:text-gray-400\">5\u003C\u002Fdiv>\u003Cdiv class=\"mt-1 font-semibold text-gray-900 dark:text-gray-100\">5. Applicazione\u003C\u002Fdiv>\u003Cdiv class=\"mt-1 text-sm text-gray-600 dark:text-gray-300\">Se è richiesta un'azione, l'applicazione o il motore di gioco la esegue e aggiorna nuovamente lo stato.\u003C\u002Fdiv>\u003C\u002Fdiv>\u003C\u002Fdiv>\u003C\u002Fsection>\n\u003Cp>Quindi l'architettura di base è:\u003C\u002Fp>\n\u003Caside class=\"editorjs-callout editorjs-callout--note my-6 rounded-xl border p-5 border-gray-300 bg-gray-50 dark:border-gray-700 dark:bg-gray-900\u002F40\" role=\"note\">\u003Cstrong class=\"block mb-2 text-gray-900 dark:text-gray-100\">L'architettura più semplice\u003C\u002Fstrong>\u003Cdiv class=\"text-gray-700 dark:text-gray-200\">\u003Cstrong>Stato = ciò che è vero ora\u003C\u002Fstrong>\u003Cbr>\u003Cstrong>RAG = conoscenze utili\u003C\u002Fstrong>\u003Cbr>\u003Cstrong>LLM = comprende, ragiona e scrive\u003C\u002Fstrong>\u003Cbr>\u003Cstrong>Applicazione = esegue l'azione reale\u003C\u002Fstrong>\u003C\u002Fdiv>\u003C\u002Faside>\n\u003Ch2 id=\"section-40\">Il RAG utilizza sempre un database vettoriale?\u003C\u002Fh2>\n\u003Cp>No.\u003C\u002Fp>\n\u003Cp>Un database vettoriale è un modo comune per costruire la ricerca semantica, ma non è la definizione di RAG.\u003C\u002Fp>\n\u003Cp>La parte importante è il recupero: il sistema trova informazioni esterne pertinenti e le aggiunge al contesto dell'LLM prima che venga generata la risposta.\u003C\u002Fp>\n\u003Cp>Il File Search di OpenAI, ad esempio, può funzionare con file archiviati in vector store. I file vengono suddivisi in parti più piccole in modo che il sistema possa recuperare le parti pertinenti a una domanda. Questa è una delle implementazioni della stessa idea di base.\u003C\u002Fp>\n\u003Ch2 id=\"section-45\">Che cos'è un embedding, in parole semplici?\u003C\u002Fh2>\n\u003Cp>Non è necessario comprendere gli embedding per capire il RAG.\u003C\u002Fp>\n\u003Cp>Ma la versione semplice è questa: un embedding è una rappresentazione numerica del significato. Aiuta un sistema di ricerca a trovare testo concettualmente simile anche quando le parole non sono esattamente le stesse.\u003C\u002Fp>\n\u003Cp>Ad esempio, una normale ricerca per parole chiave può cercare le parole esatte \"riparazione auto\". La ricerca semantica può anche capire che \"riparare il mio veicolo\" riguarda un argomento simile.\u003C\u002Fp>\n\u003Cp>Ciò rende gli embedding utili per il RAG, ma il RAG può anche utilizzare la ricerca per parole chiave, query su database o un ibrido di diversi metodi.\u003C\u002Fp>\n\u003Ch2 id=\"section-50\">Anche il RAG non è memoria\u003C\u002Fh2>\n\u003Cp>La memoria è un altro concetto che viene spesso confuso con il RAG.\u003C\u002Fp>\n\u003Cp>La memoria è solitamente l'informazione che il sistema conserva sulle interazioni precedenti o sugli eventi precedenti. Il RAG è il meccanismo utilizzato per recuperare le conoscenze pertinenti quando servono.\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\">Parte\u003C\u002Fth>\u003Cth class=\"border border-gray-300 px-4 py-2 text-left font-semibold\">Significato semplice\u003C\u002Fth>\u003C\u002Ftr>\u003C\u002Fthead>\u003Ctbody>\u003Ctr>\u003Ctd class=\"border border-gray-300 px-4 py-2\">LLM\u003C\u002Ftd>\u003Ctd class=\"border border-gray-300 px-4 py-2\">La parte che comprende e genera il linguaggio\u003C\u002Ftd>\u003C\u002Ftr>\u003Ctr>\u003Ctd class=\"border border-gray-300 px-4 py-2\">RAG\u003C\u002Ftd>\u003Ctd class=\"border border-gray-300 px-4 py-2\">La parte che cerca le conoscenze pertinenti prima della risposta\u003C\u002Ftd>\u003C\u002Ftr>\u003Ctr>\u003Ctd class=\"border border-gray-300 px-4 py-2\">Base di conoscenza\u003C\u002Ftd>\u003Ctd class=\"border border-gray-300 px-4 py-2\">Le informazioni che il RAG può cercare\u003C\u002Ftd>\u003C\u002Ftr>\u003Ctr>\u003Ctd class=\"border border-gray-300 px-4 py-2\">Stato\u003C\u002Ftd>\u003Ctd class=\"border border-gray-300 px-4 py-2\">Ciò che è vero in questo momento nell'applicazione o nel mondo\u003C\u002Ftd>\u003C\u002Ftr>\u003Ctr>\u003Ctd class=\"border border-gray-300 px-4 py-2\">Memoria\u003C\u002Ftd>\u003Ctd class=\"border border-gray-300 px-4 py-2\">Informazioni conservate da interazioni o eventi precedenti\u003C\u002Ftd>\u003C\u002Ftr>\u003Ctr>\u003Ctd class=\"border border-gray-300 px-4 py-2\">Strumento \u002F azione\u003C\u002Ftd>\u003Ctd class=\"border border-gray-300 px-4 py-2\">Qualcosa che l'IA è autorizzata a chiamare o a chiedere all'applicazione di fare\u003C\u002Ftd>\u003C\u002Ftr>\u003Ctr>\u003Ctd class=\"border border-gray-300 px-4 py-2\">Contesto\u003C\u002Ftd>\u003Ctd class=\"border border-gray-300 px-4 py-2\">Le informazioni attualmente poste davanti all'LLM per questa richiesta\u003C\u002Ftd>\u003C\u002Ftr>\u003C\u002Ftbody>\u003C\u002Ftable>\u003C\u002Fdiv>\n\u003Ch2 id=\"section-54\">Un esempio reale di gioco: PUBG Ally\u003C\u002Fh2>\n\u003Cp>PUBG Ally è un esempio utile perché rende visibile la differenza.\u003C\u002Fp>\n\u003Cp>KRAFTON descrive lo stato della partita in tempo reale come una fonte di verità separata. Il gioco espone i fatti correnti attraverso strumenti di osservazione: arma attuale, munizioni, salute, stato della zona sicura, oggetti vicini e situazione di combattimento.\u003C\u002Fp>\n\u003Cp>La ricerca di conoscenza è un compito diverso. Il sistema può utilizzare conoscenze curate su armi, accessori, oggetti e regole. L'ACE Game Agent SDK di NVIDIA espone anche una API RAG separata per recuperare conoscenza da database creati dagli sviluppatori.\u003C\u002Fp>\n\u003Cp>Questo ci dà una separazione netta: il motore di gioco dice cosa sta succedendo ora, il recupero fornisce conoscenza rilevante, e il modello linguistico decide cosa significano le informazioni.\u003C\u002Fp>\n\u003Caside class=\"editorjs-referral my-6\">\u003Ca href=\"https:\u002F\u002Ffigure.rocks\u002Fblog\u002Fpubg-ally-shows-why-ai-teammates-need-two-brains-fast-reflexes-and-slow-reasoning\" target=\"_blank\" rel=\"noopener noreferrer\" class=\"flex flex-col sm:flex-row gap-4 rounded-xl border border-gray-200 dark:border-gray-700 p-4 transition hover:border-primary-500\">\u003Cdiv class=\"min-w-0 flex-1\">\u003Cstrong class=\"block text-lg text-gray-900 dark:text-gray-100\">PUBG Ally mostra perché i compagni di squadra AI hanno bisogno di due cervelli: riflessi veloci e ragionamento lento\u003C\u002Fstrong>\u003Cp class=\"mt-2 text-sm text-gray-600 dark:text-gray-300\">Un esempio pratico di gioco che mostra come stato in tempo reale, ragionamento linguistico e controllo deterministico lato gioco possono lavorare insieme.\u003C\u002Fp>\u003Cspan class=\"mt-3 inline-flex text-sm font-medium text-primary-600 dark:text-primary-400\">Leggi l'articolo sull'architettura di PUBG Ally →\u003C\u002Fspan>\u003C\u002Fdiv>\u003C\u002Fa>\u003C\u002Faside>\n\u003Ch2 id=\"section-60\">Un esempio completo\u003C\u002Fh2>\n\u003Cp>Immagina di dire a un compagno di squadra AI: \"Ho poca salute. Dovremmo attaccare?\"\u003C\u002Fp>\n\u003Csection class=\"editorjs-process my-6\">\u003Ch3 class=\"mb-3 text-lg font-semibold\">Cosa succede dopo\u003C\u002Fh3>\u003Cdiv class=\"grid grid-cols-1 md:grid-cols-2 xl:grid-cols-3 gap-4\">\u003Cdiv class=\"editorjs-process__step min-w-0  rounded-xl border border-gray-200 dark:border-gray-700 p-4\">\u003Cdiv class=\"text-xs font-semibold text-gray-500 dark:text-gray-400\">1\u003C\u002Fdiv>\u003Cdiv class=\"mt-1 font-semibold text-gray-900 dark:text-gray-100\">Stato\u003C\u002Fdiv>\u003Cdiv class=\"mt-1 text-sm text-gray-600 dark:text-gray-300\">Il gioco riporta: salute 24%, un nemico nelle vicinanze, due oggetti curativi disponibili.\u003C\u002Fdiv>\u003C\u002Fdiv>\u003Cdiv class=\"editorjs-process__step min-w-0  rounded-xl border border-gray-200 dark:border-gray-700 p-4\">\u003Cdiv class=\"text-xs font-semibold text-gray-500 dark:text-gray-400\">2\u003C\u002Fdiv>\u003Cdiv class=\"mt-1 font-semibold text-gray-900 dark:text-gray-100\">RAG\u003C\u002Fdiv>\u003Cdiv class=\"mt-1 text-sm text-gray-600 dark:text-gray-300\">Il sistema di conoscenza recupera le regole rilevanti per l'oggetto curativo e forse informazioni sull'arma attuale o sulla meccanica tattica.\u003C\u002Fdiv>\u003C\u002Fdiv>\u003Cdiv class=\"editorjs-process__step min-w-0  rounded-xl border border-gray-200 dark:border-gray-700 p-4\">\u003Cdiv class=\"text-xs font-semibold text-gray-500 dark:text-gray-400\">3\u003C\u002Fdiv>\u003Cdiv class=\"mt-1 font-semibold text-gray-900 dark:text-gray-100\">LLM\u003C\u002Fdiv>\u003Cdiv class=\"mt-1 text-sm text-gray-600 dark:text-gray-300\">Il modello combina la tua richiesta, lo stato attuale e la conoscenza recuperata.\u003C\u002Fdiv>\u003C\u002Fdiv>\u003Cdiv class=\"editorjs-process__step min-w-0  rounded-xl border border-gray-200 dark:border-gray-700 p-4\">\u003Cdiv class=\"text-xs font-semibold text-gray-500 dark:text-gray-400\">4\u003C\u002Fdiv>\u003Cdiv class=\"mt-1 font-semibold text-gray-900 dark:text-gray-100\">Decisione\u003C\u002Fdiv>\u003Cdiv class=\"mt-1 text-sm text-gray-600 dark:text-gray-300\">Conclude che curarsi prima è più sicuro che attaccare immediatamente.\u003C\u002Fdiv>\u003C\u002Fdiv>\u003Cdiv class=\"editorjs-process__step min-w-0  rounded-xl border border-gray-200 dark:border-gray-700 p-4\">\u003Cdiv class=\"text-xs font-semibold text-gray-500 dark:text-gray-400\">5\u003C\u002Fdiv>\u003Cdiv class=\"mt-1 font-semibold text-gray-900 dark:text-gray-100\">Strumento \u002F motore di gioco\u003C\u002Fdiv>\u003Cdiv class=\"mt-1 text-sm text-gray-600 dark:text-gray-300\">L'agente richiede un'azione di gioco legale come spostarsi al riparo o usare l'oggetto curativo.\u003C\u002Fdiv>\u003C\u002Fdiv>\u003Cdiv class=\"editorjs-process__step min-w-0  rounded-xl border border-gray-200 dark:border-gray-700 p-4\">\u003Cdiv class=\"text-xs font-semibold text-gray-500 dark:text-gray-400\">6\u003C\u002Fdiv>\u003Cdiv class=\"mt-1 font-semibold text-gray-900 dark:text-gray-100\">Nuovo stato\u003C\u002Fdiv>\u003Cdiv class=\"mt-1 text-sm text-gray-600 dark:text-gray-300\">Il gioco esegue l'azione e riporta la situazione aggiornata all'agente.\u003C\u002Fdiv>\u003C\u002Fdiv>\u003C\u002Fdiv>\u003C\u002Fsection>\n\u003Cp>RAG non controllava il personaggio. Il database di stato non ragionava. L'LLM non modificava direttamente il gioco. Ogni parte aveva un compito.\u003C\u002Fp>\n\u003Ch2 id=\"section-64\">Perché usare RAG?\u003C\u002Fh2>\n\u003Cp>Perché mettere ogni documento, regola e record di database in ogni prompt sarebbe lento, costoso e spesso confuso.\u003C\u002Fp>\n\u003Cp>RAG permette al sistema di selezionare solo le informazioni utili per la domanda corrente.\u003C\u002Fp>\n\u003Cp>Permette anche di aggiornare la base di conoscenza senza riaddestrare l'intero modello linguistico. Cambia il documento o il database, ricostruisci o aggiorna l'indice quando necessario, e il prossimo recupero potrà usare le informazioni più recenti.\u003C\u002Fp>\n\u003Ch2 id=\"section-68\">Cosa non garantisce RAG\u003C\u002Fh2>\n\u003Cp>RAG può migliorare l'ancoraggio, ma non rende automaticamente corretta una risposta.\u003C\u002Fp>\n\u003Cp>Il passo di recupero può trovare il documento sbagliato. Il documento corretto può essere obsoleto. L'LLM può fraintendere prove valide. Oppure lo stato attuale può essere cambiato.\u003C\u002Fp>\n\u003Cp>Un sistema affidabile deve quindi validare separatamente il recupero, la freschezza dello stato e il ragionamento finale del modello.\u003C\u002Fp>\n\u003Ch2 id=\"section-72\">Il modello mentale più facile da ricordare\u003C\u002Fh2>\n\u003Csection class=\"editorjs-comparison my-6\">\u003Ch3 class=\"mb-3 text-lg font-semibold\">Pensa a un sistema di IA come a una persona alla scrivania\u003C\u002Fh3>\u003Cdiv class=\"overflow-x-auto\">\u003Ctable class=\"w-full border-collapse\">\u003Cthead>\u003Ctr>\u003Cth class=\"border border-gray-300 bg-gray-50 px-4 py-3 text-left dark:border-gray-700 dark:bg-gray-900\">\u003C\u002Fth>\u003Cth class=\"border border-gray-300 bg-gray-50 px-4 py-3 text-left font-semibold dark:border-gray-700 dark:bg-gray-900\">Analogia\u003C\u002Fth>\u003Cth class=\"border border-gray-300 bg-gray-50 px-4 py-3 text-left font-semibold dark:border-gray-700 dark:bg-gray-900\">Sistema di IA\u003C\u002Fth>\u003C\u002Ftr>\u003C\u002Fthead>\u003Ctbody>\u003Ctr>\u003Cth class=\"border border-gray-300 px-4 py-3 text-left font-semibold dark:border-gray-700\">Persona che pensa\u003C\u002Fth>\u003Ctd class=\"border border-gray-300 px-4 py-3 dark:border-gray-700\">\u003C\u002Ftd>\u003Ctd class=\"border border-gray-300 px-4 py-3 dark:border-gray-700\">\u003C\u002Ftd>\u003C\u002Ftr>\u003Ctr>\u003Cth class=\"border border-gray-300 px-4 py-3 text-left font-semibold dark:border-gray-700\">Trovare un libro di riferimento\u003C\u002Fth>\u003Ctd class=\"border border-gray-300 px-4 py-3 dark:border-gray-700\">\u003C\u002Ftd>\u003Ctd class=\"border border-gray-300 px-4 py-3 dark:border-gray-700\">\u003C\u002Ftd>\u003C\u002Ftr>\u003Ctr>\u003Cth class=\"border border-gray-300 px-4 py-3 text-left font-semibold dark:border-gray-700\">Libri sullo scaffale\u003C\u002Fth>\u003Ctd class=\"border border-gray-300 px-4 py-3 dark:border-gray-700\">\u003C\u002Ftd>\u003Ctd class=\"border border-gray-300 px-4 py-3 dark:border-gray-700\">\u003C\u002Ftd>\u003C\u002Ftr>\u003Ctr>\u003Cth class=\"border border-gray-300 px-4 py-3 text-left font-semibold dark:border-gray-700\">Cruscotto o pannello strumenti attuale\u003C\u002Fth>\u003Ctd class=\"border border-gray-300 px-4 py-3 dark:border-gray-700\">\u003C\u002Ftd>\u003Ctd class=\"border border-gray-300 px-4 py-3 dark:border-gray-700\">\u003C\u002Ftd>\u003C\u002Ftr>\u003Ctr>\u003Cth class=\"border border-gray-300 px-4 py-3 text-left font-semibold dark:border-gray-700\">Note delle riunioni precedenti\u003C\u002Fth>\u003Ctd class=\"border border-gray-300 px-4 py-3 dark:border-gray-700\">\u003C\u002Ftd>\u003Ctd class=\"border border-gray-300 px-4 py-3 dark:border-gray-700\">\u003C\u002Ftd>\u003C\u002Ftr>\u003Ctr>\u003Cth class=\"border border-gray-300 px-4 py-3 text-left font-semibold dark:border-gray-700\">Fare qualcosa nel mondo reale\u003C\u002Fth>\u003Ctd class=\"border border-gray-300 px-4 py-3 dark:border-gray-700\">\u003C\u002Ftd>\u003Ctd class=\"border border-gray-300 px-4 py-3 dark:border-gray-700\">\u003C\u002Ftd>\u003C\u002Ftr>\u003C\u002Ftbody>\u003C\u002Ftable>\u003C\u002Fdiv>\u003C\u002Fsection>\n\u003Caside class=\"editorjs-callout editorjs-callout--success my-6 rounded-xl border p-5 border-emerald-300 bg-emerald-50 dark:border-emerald-900 dark:bg-emerald-950\u002F20\" role=\"note\">\u003Cstrong class=\"block mb-2 text-gray-900 dark:text-gray-100\">Se ricordi solo questo\u003C\u002Fstrong>\u003Cdiv class=\"text-gray-700 dark:text-gray-200\">\u003Cstrong>LLM = cervello.\u003C\u002Fstrong>\u003Cbr>\u003Cstrong>RAG = bibliotecario.\u003C\u002Fstrong>\u003Cbr>\u003Cstrong>Base di conoscenza = biblioteca.\u003C\u002Fstrong>\u003Cbr>\u003Cstrong>Stato = ciò che dice il cruscotto in questo momento.\u003C\u002Fstrong>\u003Cbr>\u003Cstrong>Strumenti = le mani che possono effettivamente fare qualcosa.\u003C\u002Fstrong>\u003C\u002Fdiv>\u003C\u002Faside>\n\u003Ch2 id=\"section-75\">Conclusione\u003C\u002Fh2>\n\u003Cp>RAG è molto meno misterioso una volta separate le parti.\u003C\u002Fp>\n\u003Cp>L'LLM comprende e genera linguaggio. L'applicazione mantiene lo stato attuale. La base di conoscenza memorizza le informazioni. RAG trova la parte utile di tali informazioni e la inserisce nel contesto dell'LLM. Gli strumenti o l'applicazione eseguono azioni reali.\u003C\u002Fp>\n\u003Cp>Questa è l'architettura di base dietro molti assistenti e agenti di IA moderni.\u003C\u002Fp>\n\u003Ch2 id=\"section-79\">FAQ\u003C\u002Fh2>\n\u003Csection class=\"editorjs-faq my-6 rounded-xl border border-gray-200 p-5 dark:border-gray-700\">\u003Ch3 class=\"mb-3 text-lg font-semibold\">RAG in parole semplici\u003C\u002Fh3>\u003Cdiv id=\"faq1\" class=\"border-t border-gray-200 py-4 first:border-t-0 dark:border-gray-700\">\u003Ch4 class=\"font-semibold text-gray-900 dark:text-gray-100\">Cos'è RAG in termini semplici?\u003C\u002Fh4>\u003Cdiv class=\"mt-2 text-gray-600 dark:text-gray-300\">RAG è un passaggio in cui un'IA cerca informazioni rilevanti in una fonte di conoscenza prima che il modello linguistico scriva la sua risposta.\u003C\u002Fdiv>\u003C\u002Fdiv>\u003Cdiv id=\"faq2\" class=\"border-t border-gray-200 py-4 first:border-t-0 dark:border-gray-700\">\u003Ch4 class=\"font-semibold text-gray-900 dark:text-gray-100\">RAG ha bisogno di Internet?\u003C\u002Fh4>\u003Cdiv class=\"mt-2 text-gray-600 dark:text-gray-300\">No. La base di conoscenza può essere completamente locale sul tuo computer o server.\u003C\u002Fdiv>\u003C\u002Fdiv>\u003Cdiv id=\"faq3\" class=\"border-t border-gray-200 py-4 first:border-t-0 dark:border-gray-700\">\u003Ch4 class=\"font-semibold text-gray-900 dark:text-gray-100\">RAG è la stessa cosa di un database?\u003C\u002Fh4>\u003Cdiv class=\"mt-2 text-gray-600 dark:text-gray-300\">No. Il database o i file contengono le informazioni. RAG è il processo di recupero che trova la parte utile e la fornisce all'LLM.\u003C\u002Fdiv>\u003C\u002Fdiv>\u003Cdiv id=\"faq4\" class=\"border-t border-gray-200 py-4 first:border-t-0 dark:border-gray-700\">\u003Ch4 class=\"font-semibold text-gray-900 dark:text-gray-100\">RAG è la stessa cosa della memoria?\u003C\u002Fh4>\u003Cdiv class=\"mt-2 text-gray-600 dark:text-gray-300\">No. La memoria di solito memorizza interazioni o eventi precedenti. RAG recupera conoscenze rilevanti quando servono.\u003C\u002Fdiv>\u003C\u002Fdiv>\u003Cdiv id=\"faq5\" class=\"border-t border-gray-200 py-4 first:border-t-0 dark:border-gray-700\">\u003Ch4 class=\"font-semibold text-gray-900 dark:text-gray-100\">Lo stato attuale dell'applicazione fa parte di RAG?\u003C\u002Fh4>\u003Cdiv class=\"mt-2 text-gray-600 dark:text-gray-300\">Non necessariamente. Lo stato attuale di solito si ottiene direttamente dall'applicazione o da un archivio di stato. RAG è meglio inteso come recupero da una fonte di conoscenza.\u003C\u002Fdiv>\u003C\u002Fdiv>\u003Cdiv id=\"faq6\" class=\"border-t border-gray-200 py-4 first:border-t-0 dark:border-gray-700\">\u003Ch4 class=\"font-semibold text-gray-900 dark:text-gray-100\">RAG rende corrette le risposte dell'IA?\u003C\u002Fh4>\u003Cdiv class=\"mt-2 text-gray-600 dark:text-gray-300\">No. Può fornire prove migliori, ma il recupero può comunque essere errato o obsoleto e l'LLM può comunque ragionare in modo scorretto.\u003C\u002Fdiv>\u003C\u002Fdiv>\u003C\u002Fsection>\n\u003Ch2 id=\"section-81\">Glossario\u003C\u002Fh2>\n\u003Csection class=\"editorjs-glossary my-6 rounded-xl border border-gray-200 dark:border-gray-700 p-5\">\u003Ch3 class=\"mb-3 text-lg font-semibold\">I termini di base\u003C\u002Fh3>\u003Cdl>\u003Cdiv id=\"llm\" class=\"border-t border-gray-200 dark:border-gray-700 py-3 first:border-t-0\">\u003Cdt class=\"font-semibold text-gray-900 dark:text-gray-100\">LLM\u003C\u002Fdt>\u003Cdd class=\"mt-1 text-gray-600 dark:text-gray-300\">Un modello linguistico che comprende e genera testo e può ragionare sulle informazioni inserite nel suo contesto.\u003C\u002Fdd>\u003C\u002Fdiv>\u003Cdiv id=\"rag\" class=\"border-t border-gray-200 dark:border-gray-700 py-3 first:border-t-0\">\u003Cdt class=\"font-semibold text-gray-900 dark:text-gray-100\">RAG\u003C\u002Fdt>\u003Cdd class=\"mt-1 text-gray-600 dark:text-gray-300\">Retrieval-Augmented Generation: recuperare informazioni esterne rilevanti e aggiungerle al contesto del modello prima di generare una risposta.\u003C\u002Fdd>\u003C\u002Fdiv>\u003Cdiv id=\"knowledge-base\" class=\"border-t border-gray-200 dark:border-gray-700 py-3 first:border-t-0\">\u003Cdt class=\"font-semibold text-gray-900 dark:text-gray-100\">Base di conoscenza\u003C\u002Fdt>\u003Cdd class=\"mt-1 text-gray-600 dark:text-gray-300\">I file, i documenti, i record o altre informazioni che il recupero può cercare.\u003C\u002Fdd>\u003C\u002Fdiv>\u003Cdiv id=\"state\" class=\"border-t border-gray-200 dark:border-gray-700 py-3 first:border-t-0\">\u003Cdt class=\"font-semibold text-gray-900 dark:text-gray-100\">Stato\u003C\u002Fdt>\u003Cdd class=\"mt-1 text-gray-600 dark:text-gray-300\">I fatti attuali di un'applicazione, sistema o mondo in un particolare momento.\u003C\u002Fdd>\u003C\u002Fdiv>\u003Cdiv id=\"context\" class=\"border-t border-gray-200 dark:border-gray-700 py-3 first:border-t-0\">\u003Cdt class=\"font-semibold text-gray-900 dark:text-gray-100\">Contesto\u003C\u002Fdt>\u003Cdd class=\"mt-1 text-gray-600 dark:text-gray-300\">Le informazioni attualmente fornite al modello linguistico per una richiesta o un passaggio di ragionamento.\u003C\u002Fdd>\u003C\u002Fdiv>\u003Cdiv id=\"embedding\" class=\"border-t border-gray-200 dark:border-gray-700 py-3 first:border-t-0\">\u003Cdt class=\"font-semibold text-gray-900 dark:text-gray-100\">Embedding\u003C\u002Fdt>\u003Cdd class=\"mt-1 text-gray-600 dark:text-gray-300\">Una rappresentazione numerica del significato che può aiutare la ricerca semantica a trovare informazioni concettualmente simili.\u003C\u002Fdd>\u003C\u002Fdiv>\u003C\u002Fdl>\u003C\u002Fsection>\n\u003Ch2 id=\"section-83\">Fonti primarie\u003C\u002Fh2>\n\u003Ca href=\"https:\u002F\u002Fplatform.openai.com\u002Fdocs\u002Fapi-reference\u002Fvector-stores-files\" target=\"_blank\" rel=\"noopener noreferrer\" 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\">OpenAI — File di Vector Store\u003C\u002Fstrong>\u003Cp class=\"text-sm text-gray-600 dark:text-gray-400\">Documentazione ufficiale che mostra come i file possono essere allegati ai vector store, suddivisi in chunk e resi disponibili per il recupero tramite ricerca di file.\u003C\u002Fp>\u003C\u002Fa>\n\u003Ca href=\"https:\u002F\u002Fplatform.openai.com\u002Fdocs\u002Fquickstart\" target=\"_blank\" rel=\"noopener noreferrer\" 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\">OpenAI — Guida rapida per sviluppatori\u003C\u002Fstrong>\u003Cp class=\"text-sm text-gray-600 dark:text-gray-400\">Documentazione ufficiale OpenAI che descrive strumenti come la ricerca di file per dare ai modelli accesso a informazioni esterne.\u003C\u002Fp>\u003C\u002Fa>\n\u003Ca href=\"https:\u002F\u002Fdeveloper.nvidia.com\u002Face-for-games\" target=\"_blank\" rel=\"noopener noreferrer\" 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\">NVIDIA Developer — ACE per i giochi\u003C\u002Fstrong>\u003Cp class=\"text-sm text-gray-600 dark:text-gray-400\">Documentazione ufficiale NVIDIA che descrive API separate Agent, Chat e RAG per collegare i personaggi dei giochi allo stato del gioco, alla conoscenza contestuale e alle azioni guidate dal modello.\u003C\u002Fp>\u003C\u002Fa>\n\u003Ca href=\"https:\u002F\u002Fdeveloper.nvidia.com\u002Fblog\u002Fhow-krafton-built-pubg-ally-a-co-playable-character-powered-by-nvidia-ace\u002F\" target=\"_blank\" rel=\"noopener noreferrer\" 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\">NVIDIA Developer — Come KRAFTON ha costruito PUBG Ally\u003C\u002Fstrong>\u003Cp class=\"text-sm text-gray-600 dark:text-gray-400\">Spiegazione tecnica ufficiale che separa lo stato della partita in tempo reale dalla ricerca di conoscenza e dal ragionamento del modello linguistico.\u003C\u002Fp>\u003C\u002Fa>",{"time":212,"blocks":213,"version":873},1790377619048,[214,220,228,233,241,246,251,256,261,268,273,278,283,288,295,300,320,325,330,335,340,361,366,371,376,381,386,391,422,429,434,439,444,449,454,459,464,485,490,496,501,506,511,516,521,526,531,536,541,546,551,556,561,590,595,600,605,610,615,624,629,634,655,660,665,670,675,680,685,690,695,700,705,741,747,752,757,762,767,772,802,807,831,836,846,855,864],{"id":215,"data":216,"type":218,"tunes":219},"intro",{"text":217},"RAG sembra complicato perché il nome è complicato. L'idea non lo è. RAG significa semplicemente: prima che l'IA risponda, cerca prima le informazioni rilevanti da una fonte di conoscenza e fornisce tali informazioni al modello linguistico.","paragraph",{},{"id":221,"data":222,"type":226,"tunes":227},"one-sentence",{"body":223,"title":224,"variant":225},"\u003Cstrong>RAG è il passaggio in cui un'IA cerca in una base di conoscenza informazioni utili prima che l'LLM scriva la risposta.\u003C\u002Fstrong>","RAG in una frase","info","callout",{},{"id":229,"data":230,"type":218,"tunes":232},"analogy",{"text":231},"Pensa a un LLM come a una persona intelligente seduta a una scrivania. RAG è il bibliotecario che porta la pagina giusta dal libro giusto. L'LLM poi legge quella pagina e ti risponde.",{},{"id":234,"data":235,"type":239,"tunes":240},"toc",{"title":236,"maxLevel":237,"minLevel":238},"Contenuti",3,2,"tableOfContents",{},{"id":242,"data":243,"type":42,"tunes":245},"h-llm",{"text":244,"level":238},"Prima: cosa fa l'LLM?",{},{"id":247,"data":248,"type":218,"tunes":250},"p-llm-1",{"text":249},"L'LLM è la parte che comprende il linguaggio e produce linguaggio. Può leggere la tua domanda, comprendere le istruzioni, confrontare informazioni, spiegare qualcosa e scrivere una risposta.",{},{"id":252,"data":253,"type":218,"tunes":255},"p-llm-2",{"text":254},"Ma l'LLM non sa automaticamente cosa c'è attualmente nel database della tua azienda, nella tua sessione di gioco, nei tuoi documenti privati o in un file che hai creato cinque minuti fa.",{},{"id":257,"data":258,"type":218,"tunes":260},"p-llm-3",{"text":259},"Sa solo ciò che è già all'interno del modello più qualsiasi informazione che l'applicazione gli fornisce nella richiesta corrente.",{},{"id":262,"data":263,"type":226,"tunes":267},"llm-rule",{"body":264,"title":265,"variant":266},"L'LLM \u003Cstrong>pensa e scrive\u003C\u002Fstrong>. Non possiede automaticamente tutti i tuoi dati attuali.","Regola semplice","note",{},{"id":269,"data":270,"type":42,"tunes":272},"h-kb",{"text":271,"level":238},"Poi: cos'è la base di conoscenza?",{},{"id":274,"data":275,"type":218,"tunes":277},"p-kb-1",{"text":276},"Una base di conoscenza è semplicemente informazione che l'applicazione può cercare.",{},{"id":279,"data":280,"type":218,"tunes":282},"p-kb-2",{"text":281},"Potrebbe contenere PDF, manuali, documentazione di prodotto, articoli di supporto, contratti, regole di gioco, dati sulle armi, documenti aziendali interni, record di database o altro testo.",{},{"id":284,"data":285,"type":218,"tunes":287},"p-kb-3",{"text":286},"La base di conoscenza può essere locale sulla tua macchina. Può essere su un server. Può essere in un database vettoriale. Può anche essere costruita da file normali. RAG non significa Internet.",{},{"id":289,"data":290,"type":226,"tunes":294},"no-internet",{"body":291,"title":292,"variant":293},"\u003Cstrong>RAG non richiede Internet.\u003C\u002Fstrong> Le informazioni possono essere completamente locali.","Importante","success",{},{"id":296,"data":297,"type":42,"tunes":299},"h-rag",{"text":298,"level":238},"Quindi cosa fa effettivamente RAG?",{},{"id":301,"data":302,"type":318,"tunes":319},"rag-flow",{"steps":303,"title":316,"orientation":317},[304,307,310,313],{"label":305,"description":306},"1. Fai una domanda","Ad esempio: quali munizioni usa quest'arma?",{"label":308,"description":309},"2. RAG cerca nella base di conoscenza","Il sistema cerca i piccoli pezzi di informazione più rilevanti per la tua domanda.",{"label":311,"description":312},"3. RAG fornisce quei pezzi all'LLM","L'LLM riceve la domanda più le informazioni recuperate.",{"label":314,"description":315},"4. L'LLM scrive la risposta","Usa le informazioni recuperate come contesto per la risposta.","L'intero processo RAG","auto","processFlow",{},{"id":321,"data":322,"type":218,"tunes":324},"rag-that-is-it",{"text":323},"Questo è RAG.",{},{"id":326,"data":327,"type":218,"tunes":329},"rag-name",{"text":328},"Il nome completo è Retrieval-Augmented Generation. Retrieval significa trovare le informazioni rilevanti. Augmented significa aggiungere quelle informazioni al contesto del modello. Generation significa che l'LLM scrive la risposta finale.",{},{"id":331,"data":332,"type":42,"tunes":334},"h-example",{"text":333,"level":238},"Un esempio molto semplice",{},{"id":336,"data":337,"type":218,"tunes":339},"p-ex-1",{"text":338},"Immagina di avere una base di conoscenza locale su un gioco.",{},{"id":341,"data":342,"type":359,"tunes":360},"kb-table",{"content":343,"stretched":43,"withHeadings":14},[344,347,350,353,356],[345,346],"La base di conoscenza contiene","Esempio",[348,349],"Armi","L'AKM usa munizioni da 7,62 mm",[351,352],"Oggetti curativi","Il Med Kit ripristina la salute",[354,355],"Accessori","Questo accessorio funziona con queste armi",[357,358],"Regole della mappa","Questa zona si comporta in questo modo","table",{},{"id":362,"data":363,"type":218,"tunes":365},"p-ex-2",{"text":364},"Chiedi: \"Quali munizioni usa l'AKM?\"",{},{"id":367,"data":368,"type":218,"tunes":370},"p-ex-3",{"text":369},"RAG cerca nella base di conoscenza e trova la voce relativa all'AKM. Fornisce quel piccolo pezzo di informazione all'LLM. L'LLM poi risponde: \"L'AKM usa munizioni da 7,62 mm.\"",{},{"id":372,"data":373,"type":218,"tunes":375},"p-ex-4",{"text":374},"L'LLM non aveva bisogno dell'intero database. RAG ha portato solo la parte utile.",{},{"id":377,"data":378,"type":42,"tunes":380},"h-state",{"text":379,"level":238},"Ora la parte importante: RAG non è lo stato attuale",{},{"id":382,"data":383,"type":218,"tunes":385},"p-state-1",{"text":384},"Qui è dove molte spiegazioni diventano confuse.",{},{"id":387,"data":388,"type":218,"tunes":390},"p-state-2",{"text":389},"RAG di solito fornisce conoscenza all'IA. Un sistema di stato fornisce all'IA fatti su ciò che è vero in questo momento.",{},{"id":392,"data":393,"type":420,"tunes":421},"knowledge-state",{"rows":394,"title":412,"layout":359,"columns":413},[395,400,404,408],{"id":396,"label":397,"values":398},"weapon","Arma",[399,399],"",{"id":401,"label":402,"values":403},"ammo","Munizioni",[399,399],{"id":405,"label":406,"values":407},"health","Salute",[399,399],{"id":409,"label":410,"values":411},"enemy","Nemico",[399,399],"Conoscenza vs stato attuale",[414,417],{"id":415,"label":416},"knowledge","RAG \u002F conoscenza",{"id":418,"label":419},"state","Stato attuale","comparison",{},{"id":423,"data":424,"type":226,"tunes":428},"dont-mix",{"body":425,"title":426,"variant":427},"RAG risponde: \u003Cstrong>Cosa è generalmente vero?\u003C\u002Fstrong>\u003Cbr>Lo stato risponde: \u003Cstrong>Cosa è vero in questo momento?\u003C\u002Fstrong>","Non mescolare questi due","warning",{},{"id":430,"data":431,"type":42,"tunes":433},"h-state-db",{"text":432,"level":238},"Cos'è un database di stato?",{},{"id":435,"data":436,"type":218,"tunes":438},"p-statedb-1",{"text":437},"Un database di stato o archivio di stato è semplicemente un luogo dove l'applicazione conserva i fatti attuali.",{},{"id":440,"data":441,"type":218,"tunes":443},"p-statedb-2",{"text":442},"In un gioco, il motore sa già cose come la tua salute, posizione, inventario, munizioni, missione corrente, oggetti vicini e stato dei nemici. Un sistema di IA può esporre parti selezionate di quello stato al modello.",{},{"id":445,"data":446,"type":218,"tunes":448},"p-statedb-3",{"text":447},"In un'applicazione aziendale, la stessa idea potrebbe essere un database di ordini, un record cliente, lo stato di un progetto o il valore corrente di un sensore.",{},{"id":450,"data":451,"type":218,"tunes":453},"p-statedb-4",{"text":452},"Lo stato viene creato dall'applicazione stessa man mano che accadono le cose. Se perdi salute, il gioco aggiorna il valore della salute. Se raccogli munizioni, l'inventario cambia. Se un ordine viene pagato, il sistema aziendale cambia lo stato dell'ordine.",{},{"id":455,"data":456,"type":226,"tunes":458},"state-rule",{"body":457,"title":265,"variant":225},"L'applicazione crea e aggiorna lo \u003Cstrong>stato\u003C\u002Fstrong>. RAG cerca la \u003Cstrong>conoscenza\u003C\u002Fstrong>. L'LLM usa entrambi per decidere cosa dire o fare.",{},{"id":460,"data":461,"type":42,"tunes":463},"h-together",{"text":462,"level":238},"Come i tre pezzi lavorano insieme",{},{"id":465,"data":466,"type":318,"tunes":484},"together-flow",{"steps":467,"title":483,"orientation":317},[468,471,474,477,480],{"label":469,"description":470},"1. Stato attuale","L'applicazione comunica all'IA ciò che è vero in questo momento: salute 41%, AKM equipaggiato, 23 colpi.",{"label":472,"description":473},"2. RAG","Il sistema recupera conoscenze utili: come funziona l'arma, quale oggetto curativo è disponibile o una regola pertinente.",{"label":475,"description":476},"3. LLM","Il modello riceve la domanda, lo stato attuale e le conoscenze recuperate.",{"label":478,"description":479},"4. Ragionamento","L'LLM combina questi input e decide quale risposta o azione di alto livello ha senso.",{"label":481,"description":482},"5. Applicazione","Se è richiesta un'azione, l'applicazione o il motore di gioco la esegue e aggiorna nuovamente lo stato.","LLM + stato + RAG",{},{"id":486,"data":487,"type":218,"tunes":489},"p-arch-intro",{"text":488},"Quindi l'architettura di base è:",{},{"id":491,"data":492,"type":226,"tunes":495},"simple-architecture",{"body":493,"title":494,"variant":266},"\u003Cstrong>Stato = ciò che è vero ora\u003C\u002Fstrong>\u003Cbr>\u003Cstrong>RAG = conoscenze utili\u003C\u002Fstrong>\u003Cbr>\u003Cstrong>LLM = comprende, ragiona e scrive\u003C\u002Fstrong>\u003Cbr>\u003Cstrong>Applicazione = esegue l'azione reale\u003C\u002Fstrong>","L'architettura più semplice",{},{"id":497,"data":498,"type":42,"tunes":500},"h-vector",{"text":499,"level":238},"Il RAG utilizza sempre un database vettoriale?",{},{"id":502,"data":503,"type":218,"tunes":505},"p-vector-1",{"text":504},"No.",{},{"id":507,"data":508,"type":218,"tunes":510},"p-vector-2",{"text":509},"Un database vettoriale è un modo comune per costruire la ricerca semantica, ma non è la definizione di RAG.",{},{"id":512,"data":513,"type":218,"tunes":515},"p-vector-3",{"text":514},"La parte importante è il recupero: il sistema trova informazioni esterne pertinenti e le aggiunge al contesto dell'LLM prima che venga generata la risposta.",{},{"id":517,"data":518,"type":218,"tunes":520},"p-vector-4",{"text":519},"Il File Search di OpenAI, ad esempio, può funzionare con file archiviati in vector store. I file vengono suddivisi in parti più piccole in modo che il sistema possa recuperare le parti pertinenti a una domanda. Questa è una delle implementazioni della stessa idea di base.",{},{"id":522,"data":523,"type":42,"tunes":525},"h-embedding",{"text":524,"level":238},"Che cos'è un embedding, in parole semplici?",{},{"id":527,"data":528,"type":218,"tunes":530},"p-emb-1",{"text":529},"Non è necessario comprendere gli embedding per capire il RAG.",{},{"id":532,"data":533,"type":218,"tunes":535},"p-emb-2",{"text":534},"Ma la versione semplice è questa: un embedding è una rappresentazione numerica del significato. Aiuta un sistema di ricerca a trovare testo concettualmente simile anche quando le parole non sono esattamente le stesse.",{},{"id":537,"data":538,"type":218,"tunes":540},"p-emb-3",{"text":539},"Ad esempio, una normale ricerca per parole chiave può cercare le parole esatte \"riparazione auto\". La ricerca semantica può anche capire che \"riparare il mio veicolo\" riguarda un argomento simile.",{},{"id":542,"data":543,"type":218,"tunes":545},"p-emb-4",{"text":544},"Ciò rende gli embedding utili per il RAG, ma il RAG può anche utilizzare la ricerca per parole chiave, query su database o un ibrido di diversi metodi.",{},{"id":547,"data":548,"type":42,"tunes":550},"h-memory",{"text":549,"level":238},"Anche il RAG non è memoria",{},{"id":552,"data":553,"type":218,"tunes":555},"p-memory-1",{"text":554},"La memoria è un altro concetto che viene spesso confuso con il RAG.",{},{"id":557,"data":558,"type":218,"tunes":560},"p-memory-2",{"text":559},"La memoria è solitamente l'informazione che il sistema conserva sulle interazioni precedenti o sugli eventi precedenti. Il RAG è il meccanismo utilizzato per recuperare le conoscenze pertinenti quando servono.",{},{"id":562,"data":563,"type":359,"tunes":589},"parts-table",{"content":564,"stretched":43,"withHeadings":14},[565,568,571,574,577,580,583,586],[566,567],"Parte","Significato semplice",[569,570],"LLM","La parte che comprende e genera il linguaggio",[572,573],"RAG","La parte che cerca le conoscenze pertinenti prima della risposta",[575,576],"Base di conoscenza","Le informazioni che il RAG può cercare",[578,579],"Stato","Ciò che è vero in questo momento nell'applicazione o nel mondo",[581,582],"Memoria","Informazioni conservate da interazioni o eventi precedenti",[584,585],"Strumento \u002F azione","Qualcosa che l'IA è autorizzata a chiamare o a chiedere all'applicazione di fare",[587,588],"Contesto","Le informazioni attualmente poste davanti all'LLM per questa richiesta",{},{"id":591,"data":592,"type":42,"tunes":594},"h-pubg",{"text":593,"level":238},"Un esempio reale di gioco: PUBG Ally",{},{"id":596,"data":597,"type":218,"tunes":599},"p-pubg-1",{"text":598},"PUBG Ally è un esempio utile perché rende visibile la differenza.",{},{"id":601,"data":602,"type":218,"tunes":604},"p-pubg-2",{"text":603},"KRAFTON descrive lo stato della partita in tempo reale come una fonte di verità separata. Il gioco espone i fatti correnti attraverso strumenti di osservazione: arma attuale, munizioni, salute, stato della zona sicura, oggetti vicini e situazione di combattimento.",{},{"id":606,"data":607,"type":218,"tunes":609},"p-pubg-3",{"text":608},"La ricerca di conoscenza è un compito diverso. Il sistema può utilizzare conoscenze curate su armi, accessori, oggetti e regole. L'ACE Game Agent SDK di NVIDIA espone anche una API RAG separata per recuperare conoscenza da database creati dagli sviluppatori.",{},{"id":611,"data":612,"type":218,"tunes":614},"p-pubg-4",{"text":613},"Questo ci dà una separazione netta: il motore di gioco dice cosa sta succedendo ora, il recupero fornisce conoscenza rilevante, e il modello linguistico decide cosa significano le informazioni.",{},{"id":616,"data":617,"type":622,"tunes":623},"ref-pubg",{"url":618,"title":619,"excerpt":620,"ctaLabel":621},"https:\u002F\u002Ffigure.rocks\u002Fblog\u002Fpubg-ally-shows-why-ai-teammates-need-two-brains-fast-reflexes-and-slow-reasoning","PUBG Ally mostra perché i compagni di squadra AI hanno bisogno di due cervelli: riflessi veloci e ragionamento lento","Un esempio pratico di gioco che mostra come stato in tempo reale, ragionamento linguistico e controllo deterministico lato gioco possono lavorare insieme.","Leggi l'articolo sull'architettura di PUBG Ally","referralArticle",{},{"id":625,"data":626,"type":42,"tunes":628},"h-complete",{"text":627,"level":238},"Un esempio completo",{},{"id":630,"data":631,"type":218,"tunes":633},"p-complete-1",{"text":632},"Immagina di dire a un compagno di squadra AI: \"Ho poca salute. Dovremmo attaccare?\"",{},{"id":635,"data":636,"type":318,"tunes":654},"complete-flow",{"steps":637,"title":653,"orientation":317},[638,640,642,644,647,650],{"label":578,"description":639},"Il gioco riporta: salute 24%, un nemico nelle vicinanze, due oggetti curativi disponibili.",{"label":572,"description":641},"Il sistema di conoscenza recupera le regole rilevanti per l'oggetto curativo e forse informazioni sull'arma attuale o sulla meccanica tattica.",{"label":569,"description":643},"Il modello combina la tua richiesta, lo stato attuale e la conoscenza recuperata.",{"label":645,"description":646},"Decisione","Conclude che curarsi prima è più sicuro che attaccare immediatamente.",{"label":648,"description":649},"Strumento \u002F motore di gioco","L'agente richiede un'azione di gioco legale come spostarsi al riparo o usare l'oggetto curativo.",{"label":651,"description":652},"Nuovo stato","Il gioco esegue l'azione e riporta la situazione aggiornata all'agente.","Cosa succede dopo",{},{"id":656,"data":657,"type":218,"tunes":659},"p-complete-2",{"text":658},"RAG non controllava il personaggio. Il database di stato non ragionava. L'LLM non modificava direttamente il gioco. Ogni parte aveva un compito.",{},{"id":661,"data":662,"type":42,"tunes":664},"h-why",{"text":663,"level":238},"Perché usare RAG?",{},{"id":666,"data":667,"type":218,"tunes":669},"p-why-1",{"text":668},"Perché mettere ogni documento, regola e record di database in ogni prompt sarebbe lento, costoso e spesso confuso.",{},{"id":671,"data":672,"type":218,"tunes":674},"p-why-2",{"text":673},"RAG permette al sistema di selezionare solo le informazioni utili per la domanda corrente.",{},{"id":676,"data":677,"type":218,"tunes":679},"p-why-3",{"text":678},"Permette anche di aggiornare la base di conoscenza senza riaddestrare l'intero modello linguistico. Cambia il documento o il database, ricostruisci o aggiorna l'indice quando necessario, e il prossimo recupero potrà usare le informazioni più recenti.",{},{"id":681,"data":682,"type":42,"tunes":684},"h-not-guarantee",{"text":683,"level":238},"Cosa non garantisce RAG",{},{"id":686,"data":687,"type":218,"tunes":689},"p-not-1",{"text":688},"RAG può migliorare l'ancoraggio, ma non rende automaticamente corretta una risposta.",{},{"id":691,"data":692,"type":218,"tunes":694},"p-not-2",{"text":693},"Il passo di recupero può trovare il documento sbagliato. Il documento corretto può essere obsoleto. L'LLM può fraintendere prove valide. Oppure lo stato attuale può essere cambiato.",{},{"id":696,"data":697,"type":218,"tunes":699},"p-not-3",{"text":698},"Un sistema affidabile deve quindi validare separatamente il recupero, la freschezza dello stato e il ragionamento finale del modello.",{},{"id":701,"data":702,"type":42,"tunes":704},"h-mental",{"text":703,"level":238},"Il modello mentale più facile da ricordare",{},{"id":706,"data":707,"type":420,"tunes":740},"mental-table",{"rows":708,"title":733,"layout":359,"columns":734},[709,713,717,721,725,729],{"id":710,"label":711,"values":712},"brain","Persona che pensa",[399,399],{"id":714,"label":715,"values":716},"library","Trovare un libro di riferimento",[399,399],{"id":718,"label":719,"values":720},"books","Libri sullo scaffale",[399,399],{"id":722,"label":723,"values":724},"dashboard","Cruscotto o pannello strumenti attuale",[399,399],{"id":726,"label":727,"values":728},"notes","Note delle riunioni precedenti",[399,399],{"id":730,"label":731,"values":732},"hands","Fare qualcosa nel mondo reale",[399,399],"Pensa a un sistema di IA come a una persona alla scrivania",[735,737],{"id":229,"label":736},"Analogia",{"id":738,"label":739},"system","Sistema di IA",{},{"id":742,"data":743,"type":226,"tunes":746},"remember",{"body":744,"title":745,"variant":293},"\u003Cstrong>LLM = cervello.\u003C\u002Fstrong>\u003Cbr>\u003Cstrong>RAG = bibliotecario.\u003C\u002Fstrong>\u003Cbr>\u003Cstrong>Base di conoscenza = biblioteca.\u003C\u002Fstrong>\u003Cbr>\u003Cstrong>Stato = ciò che dice il cruscotto in questo momento.\u003C\u002Fstrong>\u003Cbr>\u003Cstrong>Strumenti = le mani che possono effettivamente fare qualcosa.\u003C\u002Fstrong>","Se ricordi solo questo",{},{"id":748,"data":749,"type":42,"tunes":751},"h-conclusion",{"text":750,"level":238},"Conclusione",{},{"id":753,"data":754,"type":218,"tunes":756},"p-conc-1",{"text":755},"RAG è molto meno misterioso una volta separate le parti.",{},{"id":758,"data":759,"type":218,"tunes":761},"p-conc-2",{"text":760},"L'LLM comprende e genera linguaggio. L'applicazione mantiene lo stato attuale. La base di conoscenza memorizza le informazioni. RAG trova la parte utile di tali informazioni e la inserisce nel contesto dell'LLM. Gli strumenti o l'applicazione eseguono azioni reali.",{},{"id":763,"data":764,"type":218,"tunes":766},"p-conc-3",{"text":765},"Questa è l'architettura di base dietro molti assistenti e agenti di IA moderni.",{},{"id":768,"data":769,"type":42,"tunes":771},"h-faq",{"text":770,"level":238},"FAQ",{},{"id":773,"data":774,"type":773,"tunes":801},"faq",{"items":775,"title":800},[776,780,784,788,792,796],{"id":777,"answer":778,"question":779},"faq1","RAG è un passaggio in cui un'IA cerca informazioni rilevanti in una fonte di conoscenza prima che il modello linguistico scriva la sua risposta.","Cos'è RAG in termini semplici?",{"id":781,"answer":782,"question":783},"faq2","No. La base di conoscenza può essere completamente locale sul tuo computer o server.","RAG ha bisogno di Internet?",{"id":785,"answer":786,"question":787},"faq3","No. Il database o i file contengono le informazioni. RAG è il processo di recupero che trova la parte utile e la fornisce all'LLM.","RAG è la stessa cosa di un database?",{"id":789,"answer":790,"question":791},"faq4","No. La memoria di solito memorizza interazioni o eventi precedenti. RAG recupera conoscenze rilevanti quando servono.","RAG è la stessa cosa della memoria?",{"id":793,"answer":794,"question":795},"faq5","Non necessariamente. Lo stato attuale di solito si ottiene direttamente dall'applicazione o da un archivio di stato. RAG è meglio inteso come recupero da una fonte di conoscenza.","Lo stato attuale dell'applicazione fa parte di RAG?",{"id":797,"answer":798,"question":799},"faq6","No. Può fornire prove migliori, ma il recupero può comunque essere errato o obsoleto e l'LLM può comunque ragionare in modo scorretto.","RAG rende corrette le risposte dell'IA?","RAG in parole semplici",{},{"id":803,"data":804,"type":42,"tunes":806},"h-glossary",{"text":805,"level":238},"Glossario",{},{"id":808,"data":809,"type":808,"tunes":830},"glossary",{"title":810,"entries":811},"I termini di base",[812,815,818,821,823,826],{"term":569,"anchor":813,"definition":814},"llm","Un modello linguistico che comprende e genera testo e può ragionare sulle informazioni inserite nel suo contesto.",{"term":572,"anchor":816,"definition":817},"rag","Retrieval-Augmented Generation: recuperare informazioni esterne rilevanti e aggiungerle al contesto del modello prima di generare una risposta.",{"term":575,"anchor":819,"definition":820},"knowledge-base","I file, i documenti, i record o altre informazioni che il recupero può cercare.",{"term":578,"anchor":418,"definition":822},"I fatti attuali di un'applicazione, sistema o mondo in un particolare momento.",{"term":587,"anchor":824,"definition":825},"context","Le informazioni attualmente fornite al modello linguistico per una richiesta o un passaggio di ragionamento.",{"term":827,"anchor":828,"definition":829},"Embedding","embedding","Una rappresentazione numerica del significato che può aiutare la ricerca semantica a trovare informazioni concettualmente simili.",{},{"id":832,"data":833,"type":42,"tunes":835},"h-sources",{"text":834,"level":238},"Fonti primarie",{},{"id":837,"data":838,"type":844,"tunes":845},"src-openai-vector",{"link":839,"meta":840},"https:\u002F\u002Fplatform.openai.com\u002Fdocs\u002Fapi-reference\u002Fvector-stores-files",{"image":841,"title":842,"description":843},{"url":399},"OpenAI — File di Vector Store","Documentazione ufficiale che mostra come i file possono essere allegati ai vector store, suddivisi in chunk e resi disponibili per il recupero tramite ricerca di file.","linkTool",{},{"id":847,"data":848,"type":844,"tunes":854},"src-openai-quickstart",{"link":849,"meta":850},"https:\u002F\u002Fplatform.openai.com\u002Fdocs\u002Fquickstart",{"image":851,"title":852,"description":853},{"url":399},"OpenAI — Guida rapida per sviluppatori","Documentazione ufficiale OpenAI che descrive strumenti come la ricerca di file per dare ai modelli accesso a informazioni esterne.",{},{"id":856,"data":857,"type":844,"tunes":863},"src-nvidia-ace",{"link":858,"meta":859},"https:\u002F\u002Fdeveloper.nvidia.com\u002Face-for-games",{"image":860,"title":861,"description":862},{"url":399},"NVIDIA Developer — ACE per i giochi","Documentazione ufficiale NVIDIA che descrive API separate Agent, Chat e RAG per collegare i personaggi dei giochi allo stato del gioco, alla conoscenza contestuale e alle azioni guidate dal modello.",{},{"id":865,"data":866,"type":844,"tunes":872},"src-nvidia-pubg",{"link":867,"meta":868},"https:\u002F\u002Fdeveloper.nvidia.com\u002Fblog\u002Fhow-krafton-built-pubg-ally-a-co-playable-character-powered-by-nvidia-ace\u002F",{"image":869,"title":870,"description":871},{"url":399},"NVIDIA Developer — Come KRAFTON ha costruito PUBG Ally","Spiegazione tecnica ufficiale che separa lo stato della partita in tempo reale dalla ricerca di conoscenza e dal ragionamento del modello linguistico.",{},"2.31","RAG sembra complicato, ma l'idea è semplice: prima che un'IA risponda, cerca prima informazioni utili da una fonte di conoscenza e fornisce tali informazioni al modello linguistico. Questa guida spiega RAG, LLM, stato, memoria e strumenti utilizzando un semplice modello mentale.","\u002Fuploads\u002F2026\u002F09\u002Fwhat-is-rag-the-simplest-explanation-of-how-it-works-1790377492124-khjagt.webp","what-is-rag-the-simplest-explanation-of-how-it-works-1790377492124-khjagt","PUBLISHED","2026-09-25T19:03:00.000Z","2026-09-25T23:03:13.651Z","2026-09-25T23:41:17.272Z",{"en":882,"de":883,"sr":884,"es":885,"fr":886,"it":887,"ru":888,"zh":889},"\u002Fblog\u002Fwhat-is-rag-the-simplest-explanation-of-how-it-works","\u002Fde\u002Fblog\u002Fwhat-is-rag-the-simplest-explanation-of-how-it-works","\u002Fsr\u002Fblog\u002Fwhat-is-rag-the-simplest-explanation-of-how-it-works","\u002Fes\u002Fblog\u002Fwhat-is-rag-the-simplest-explanation-of-how-it-works","\u002Ffr\u002Fblog\u002Fwhat-is-rag-the-simplest-explanation-of-how-it-works","\u002Fit\u002Fblog\u002Fwhat-is-rag-the-simplest-explanation-of-how-it-works","\u002Fru\u002Fblog\u002Fwhat-is-rag-the-simplest-explanation-of-how-it-works","\u002Fzh\u002Fblog\u002Fwhat-is-rag-the-simplest-explanation-of-how-it-works",[891,895,899,903,907,911],{"id":892,"name":893,"slug":894},46,"Panoramica","overview",{"id":896,"name":897,"slug":898},57,"Limiti dei dati","data-boundaries",{"id":900,"name":901,"slug":902},51,"Anti-pattern","anti-patterns",{"id":904,"name":905,"slug":906},58,"Valutazione e gate di qualità","evaluation",{"id":908,"name":909,"slug":910},56,"Portafoglio casi d’uso","use-case-portfolio",{"id":912,"name":913,"slug":914},60,"Controlli costo e latenza","cost-and-latency",{"id":916,"login":917,"email":918,"displayName":919},"20","rooth8233","aleksandar@stajic.de","Aleksandar Stajić",[921,1450],{"lang":922,"title":923,"content":924,"contentJson":925,"excerpt":1449},"en","What Is RAG? The Simplest Explanation of How It Works","{\"time\":1790377494031,\"blocks\":[{\"id\":\"intro\",\"type\":\"paragraph\",\"data\":{\"text\":\"RAG sounds complicated because the name is complicated. The idea is not. RAG simply means: before the AI answers, it first looks up relevant information from a knowledge source and gives that information to the language model.\"},\"tunes\":{}},{\"id\":\"one-sentence\",\"type\":\"callout\",\"data\":{\"variant\":\"info\",\"title\":\"RAG in one sentence\",\"body\":\"\u003Cstrong>RAG is the step where an AI searches a knowledge base for useful information before the LLM writes the answer.\u003C\u002Fstrong>\"},\"tunes\":{}},{\"id\":\"analogy\",\"type\":\"paragraph\",\"data\":{\"text\":\"Think of an LLM as a smart person sitting at a desk. RAG is the librarian who brings the right page from the right book. The LLM then reads that page and answers you.\"},\"tunes\":{}},{\"id\":\"toc\",\"type\":\"tableOfContents\",\"data\":{\"title\":\"Contents\",\"minLevel\":2,\"maxLevel\":3},\"tunes\":{}},{\"id\":\"h-llm\",\"type\":\"header\",\"data\":{\"text\":\"First: what does the LLM do?\",\"level\":2},\"tunes\":{}},{\"id\":\"p-llm-1\",\"type\":\"paragraph\",\"data\":{\"text\":\"The LLM is the part that understands language and produces language. It can read your question, understand instructions, compare information, explain something and write an answer.\"},\"tunes\":{}},{\"id\":\"p-llm-2\",\"type\":\"paragraph\",\"data\":{\"text\":\"But the LLM does not automatically know what is currently inside your company database, your game session, your private documents or a file you created five minutes ago.\"},\"tunes\":{}},{\"id\":\"p-llm-3\",\"type\":\"paragraph\",\"data\":{\"text\":\"It only knows what is already inside the model plus whatever information the application gives it in the current request.\"},\"tunes\":{}},{\"id\":\"llm-rule\",\"type\":\"callout\",\"data\":{\"variant\":\"note\",\"title\":\"Simple rule\",\"body\":\"The LLM \u003Cstrong>thinks and writes\u003C\u002Fstrong>. It does not automatically own all of your current data.\"},\"tunes\":{}},{\"id\":\"h-kb\",\"type\":\"header\",\"data\":{\"text\":\"Then: what is the knowledge base?\",\"level\":2},\"tunes\":{}},{\"id\":\"p-kb-1\",\"type\":\"paragraph\",\"data\":{\"text\":\"A knowledge base is simply information the application can search.\"},\"tunes\":{}},{\"id\":\"p-kb-2\",\"type\":\"paragraph\",\"data\":{\"text\":\"It could contain PDFs, manuals, product documentation, support articles, contracts, game rules, weapon data, internal company documents, database records or other text.\"},\"tunes\":{}},{\"id\":\"p-kb-3\",\"type\":\"paragraph\",\"data\":{\"text\":\"The knowledge base can be local on your own machine. It can be on a server. It can be in a vector database. It can also be built from normal files. RAG does not mean Internet.\"},\"tunes\":{}},{\"id\":\"no-internet\",\"type\":\"callout\",\"data\":{\"variant\":\"success\",\"title\":\"Important\",\"body\":\"\u003Cstrong>RAG does not require the Internet.\u003C\u002Fstrong> The information can be completely local.\"},\"tunes\":{}},{\"id\":\"h-rag\",\"type\":\"header\",\"data\":{\"text\":\"So what does RAG actually do?\",\"level\":2},\"tunes\":{}},{\"id\":\"rag-flow\",\"type\":\"processFlow\",\"data\":{\"title\":\"The whole RAG process\",\"orientation\":\"auto\",\"steps\":[{\"label\":\"1. You ask a question\",\"description\":\"For example: Which ammunition does this weapon use?\"},{\"label\":\"2. RAG searches the knowledge base\",\"description\":\"The system looks for the small pieces of information most relevant to your question.\"},{\"label\":\"3. RAG gives those pieces to the LLM\",\"description\":\"The LLM receives the question plus the retrieved information.\"},{\"label\":\"4. The LLM writes the answer\",\"description\":\"It uses the retrieved information as context for the response.\"}]},\"tunes\":{}},{\"id\":\"rag-that-is-it\",\"type\":\"paragraph\",\"data\":{\"text\":\"That is RAG.\"},\"tunes\":{}},{\"id\":\"rag-name\",\"type\":\"paragraph\",\"data\":{\"text\":\"The full name is Retrieval-Augmented Generation. Retrieval means finding the relevant information. Augmented means adding that information to the model's context. Generation means the LLM writes the final answer.\"},\"tunes\":{}},{\"id\":\"h-example\",\"type\":\"header\",\"data\":{\"text\":\"A very simple example\",\"level\":2},\"tunes\":{}},{\"id\":\"p-ex-1\",\"type\":\"paragraph\",\"data\":{\"text\":\"Imagine you have a local knowledge base about a game.\"},\"tunes\":{}},{\"id\":\"kb-table\",\"type\":\"table\",\"data\":{\"withHeadings\":true,\"stretched\":false,\"content\":[[\"Knowledge base contains\",\"Example\"],[\"Weapons\",\"AKM uses 7.62 mm ammunition\"],[\"Healing items\",\"Med Kit restores health\"],[\"Attachments\",\"This attachment works with these weapons\"],[\"Map rules\",\"This zone behaves in this way\"]]},\"tunes\":{}},{\"id\":\"p-ex-2\",\"type\":\"paragraph\",\"data\":{\"text\":\"You ask: “Which ammunition does the AKM use?”\"},\"tunes\":{}},{\"id\":\"p-ex-3\",\"type\":\"paragraph\",\"data\":{\"text\":\"RAG searches the knowledge base and finds the entry about the AKM. It gives that small piece of information to the LLM. The LLM then answers: “The AKM uses 7.62 mm ammunition.”\"},\"tunes\":{}},{\"id\":\"p-ex-4\",\"type\":\"paragraph\",\"data\":{\"text\":\"The LLM did not need the entire database. RAG only brought the useful part.\"},\"tunes\":{}},{\"id\":\"h-state\",\"type\":\"header\",\"data\":{\"text\":\"Now the important part: RAG is not the current state\",\"level\":2},\"tunes\":{}},{\"id\":\"p-state-1\",\"type\":\"paragraph\",\"data\":{\"text\":\"This is where many explanations become confusing.\"},\"tunes\":{}},{\"id\":\"p-state-2\",\"type\":\"paragraph\",\"data\":{\"text\":\"RAG usually gives the AI knowledge. A state system gives the AI facts about what is true right now.\"},\"tunes\":{}},{\"id\":\"knowledge-state\",\"type\":\"comparison\",\"data\":{\"title\":\"Knowledge vs current state\",\"layout\":\"table\",\"columns\":[{\"id\":\"knowledge\",\"label\":\"RAG \u002F knowledge\"},{\"id\":\"state\",\"label\":\"Current state\"}],\"rows\":[{\"id\":\"weapon\",\"label\":\"Weapon\",\"values\":[\"\",\"\"]},{\"id\":\"ammo\",\"label\":\"Ammunition\",\"values\":[\"\",\"\"]},{\"id\":\"health\",\"label\":\"Health\",\"values\":[\"\",\"\"]},{\"id\":\"enemy\",\"label\":\"Enemy\",\"values\":[\"\",\"\"]}]},\"tunes\":{}},{\"id\":\"dont-mix\",\"type\":\"callout\",\"data\":{\"variant\":\"warning\",\"title\":\"Do not mix these two\",\"body\":\"RAG answers: \u003Cstrong>What is generally true?\u003C\u002Fstrong>\u003Cbr>State answers: \u003Cstrong>What is true right now?\u003C\u002Fstrong>\"},\"tunes\":{}},{\"id\":\"h-state-db\",\"type\":\"header\",\"data\":{\"text\":\"What is a state database?\",\"level\":2},\"tunes\":{}},{\"id\":\"p-statedb-1\",\"type\":\"paragraph\",\"data\":{\"text\":\"A state database or state store is simply a place where the application keeps current facts.\"},\"tunes\":{}},{\"id\":\"p-statedb-2\",\"type\":\"paragraph\",\"data\":{\"text\":\"In a game, the engine already knows things such as your health, position, inventory, ammunition, current mission, nearby objects and enemy status. An AI system can expose selected parts of that state to the model.\"},\"tunes\":{}},{\"id\":\"p-statedb-3\",\"type\":\"paragraph\",\"data\":{\"text\":\"In a business application, the same idea could be an order database, a customer record, a project status or the current value of a sensor.\"},\"tunes\":{}},{\"id\":\"p-statedb-4\",\"type\":\"paragraph\",\"data\":{\"text\":\"The state is created by the application itself as things happen. If you lose health, the game updates the health value. If you pick up ammunition, the inventory changes. If an order is paid, the business system changes the order status.\"},\"tunes\":{}},{\"id\":\"state-rule\",\"type\":\"callout\",\"data\":{\"variant\":\"info\",\"title\":\"Simple rule\",\"body\":\"The application creates and updates \u003Cstrong>state\u003C\u002Fstrong>. RAG searches \u003Cstrong>knowledge\u003C\u002Fstrong>. The LLM uses both to decide what to say or do.\"},\"tunes\":{}},{\"id\":\"h-together\",\"type\":\"header\",\"data\":{\"text\":\"How the three pieces work together\",\"level\":2},\"tunes\":{}},{\"id\":\"together-flow\",\"type\":\"processFlow\",\"data\":{\"title\":\"LLM + state + RAG\",\"orientation\":\"auto\",\"steps\":[{\"label\":\"1. Current state\",\"description\":\"The application tells the AI what is true now: health 41%, AKM equipped, 23 rounds.\"},{\"label\":\"2. RAG\",\"description\":\"The system retrieves useful knowledge: how the weapon works, which healing item is available, or a relevant rule.\"},{\"label\":\"3. LLM\",\"description\":\"The model receives the question, current state and retrieved knowledge.\"},{\"label\":\"4. Reasoning\",\"description\":\"The LLM combines those inputs and decides what answer or high-level action makes sense.\"},{\"label\":\"5. Application\",\"description\":\"If an action is required, the application or game engine executes it and updates the state again.\"}]},\"tunes\":{}},{\"id\":\"p-arch-intro\",\"type\":\"paragraph\",\"data\":{\"text\":\"So the basic architecture is:\"},\"tunes\":{}},{\"id\":\"simple-architecture\",\"type\":\"callout\",\"data\":{\"variant\":\"note\",\"title\":\"The simplest architecture\",\"body\":\"\u003Cstrong>State = what is true now\u003C\u002Fstrong>\u003Cbr>\u003Cstrong>RAG = useful knowledge\u003C\u002Fstrong>\u003Cbr>\u003Cstrong>LLM = understands, reasons and writes\u003C\u002Fstrong>\u003Cbr>\u003Cstrong>Application = performs the real action\u003C\u002Fstrong>\"},\"tunes\":{}},{\"id\":\"h-vector\",\"type\":\"header\",\"data\":{\"text\":\"Does RAG always use a vector database?\",\"level\":2},\"tunes\":{}},{\"id\":\"p-vector-1\",\"type\":\"paragraph\",\"data\":{\"text\":\"No.\"},\"tunes\":{}},{\"id\":\"p-vector-2\",\"type\":\"paragraph\",\"data\":{\"text\":\"A vector database is a common way to build semantic search, but it is not the definition of RAG.\"},\"tunes\":{}},{\"id\":\"p-vector-3\",\"type\":\"paragraph\",\"data\":{\"text\":\"The important part is retrieval: the system finds relevant external information and adds it to the LLM's context before the answer is generated.\"},\"tunes\":{}},{\"id\":\"p-vector-4\",\"type\":\"paragraph\",\"data\":{\"text\":\"OpenAI's File Search, for example, can work with files stored in vector stores. Files are chunked into smaller pieces so the system can retrieve the parts that are relevant to a question. That is one implementation of the same basic idea.\"},\"tunes\":{}},{\"id\":\"h-embedding\",\"type\":\"header\",\"data\":{\"text\":\"What is an embedding, in plain English?\",\"level\":2},\"tunes\":{}},{\"id\":\"p-emb-1\",\"type\":\"paragraph\",\"data\":{\"text\":\"You do not need to understand embeddings to understand RAG.\"},\"tunes\":{}},{\"id\":\"p-emb-2\",\"type\":\"paragraph\",\"data\":{\"text\":\"But the simple version is this: an embedding is a numerical representation of meaning. It helps a search system find text that is conceptually similar even when the words are not exactly the same.\"},\"tunes\":{}},{\"id\":\"p-emb-3\",\"type\":\"paragraph\",\"data\":{\"text\":\"For example, a normal keyword search may look for the exact words “car repair.” Semantic search can also understand that “fix my vehicle” is about a similar topic.\"},\"tunes\":{}},{\"id\":\"p-emb-4\",\"type\":\"paragraph\",\"data\":{\"text\":\"That makes embeddings useful for RAG, but RAG can also use keyword search, database queries or a hybrid of several methods.\"},\"tunes\":{}},{\"id\":\"h-memory\",\"type\":\"header\",\"data\":{\"text\":\"RAG is not memory either\",\"level\":2},\"tunes\":{}},{\"id\":\"p-memory-1\",\"type\":\"paragraph\",\"data\":{\"text\":\"Memory is another concept that is often mixed together with RAG.\"},\"tunes\":{}},{\"id\":\"p-memory-2\",\"type\":\"paragraph\",\"data\":{\"text\":\"Memory is usually information the system keeps about previous interactions or previous events. RAG is the mechanism used to retrieve relevant knowledge when it is needed.\"},\"tunes\":{}},{\"id\":\"parts-table\",\"type\":\"table\",\"data\":{\"withHeadings\":true,\"stretched\":false,\"content\":[[\"Part\",\"Simple meaning\"],[\"LLM\",\"The part that understands and generates language\"],[\"RAG\",\"The part that looks up relevant knowledge before the answer\"],[\"Knowledge base\",\"The information RAG can search\"],[\"State\",\"What is true right now in the application or world\"],[\"Memory\",\"Information kept from previous interactions or events\"],[\"Tool \u002F action\",\"Something the AI is allowed to call or ask the application to do\"],[\"Context\",\"The information currently placed in front of the LLM for this request\"]]},\"tunes\":{}},{\"id\":\"h-pubg\",\"type\":\"header\",\"data\":{\"text\":\"A real game example: PUBG Ally\",\"level\":2},\"tunes\":{}},{\"id\":\"p-pubg-1\",\"type\":\"paragraph\",\"data\":{\"text\":\"PUBG Ally is a useful example because it makes the difference visible.\"},\"tunes\":{}},{\"id\":\"p-pubg-2\",\"type\":\"paragraph\",\"data\":{\"text\":\"KRAFTON describes live match state as a separate source of truth. The game exposes current facts through observation tools: current weapon, ammunition, health, safe-zone status, nearby items and combat situation.\"},\"tunes\":{}},{\"id\":\"p-pubg-3\",\"type\":\"paragraph\",\"data\":{\"text\":\"Knowledge lookup is a different job. The system can use curated knowledge about weapons, attachments, items and rules. NVIDIA's ACE Game Agent SDK also exposes a separate RAG API for retrieving knowledge from developer-built databases.\"},\"tunes\":{}},{\"id\":\"p-pubg-4\",\"type\":\"paragraph\",\"data\":{\"text\":\"That gives us the clean separation: the game engine says what is happening now, retrieval provides relevant knowledge, and the language model decides what the information means.\"},\"tunes\":{}},{\"id\":\"ref-pubg\",\"type\":\"referralArticle\",\"data\":{\"url\":\"https:\u002F\u002Ffigure.rocks\u002Fblog\u002Fpubg-ally-shows-why-ai-teammates-need-two-brains-fast-reflexes-and-slow-reasoning\",\"title\":\"PUBG Ally Shows Why AI Teammates Need Two Brains: Fast Reflexes and Slow Reasoning\",\"excerpt\":\"A practical game example showing how live state, language reasoning and deterministic game-side control can work together.\",\"ctaLabel\":\"Read the PUBG Ally architecture article\"},\"tunes\":{}},{\"id\":\"h-complete\",\"type\":\"header\",\"data\":{\"text\":\"One complete example\",\"level\":2},\"tunes\":{}},{\"id\":\"p-complete-1\",\"type\":\"paragraph\",\"data\":{\"text\":\"Imagine you tell an AI teammate: “I am low on health. Should we attack?”\"},\"tunes\":{}},{\"id\":\"complete-flow\",\"type\":\"processFlow\",\"data\":{\"title\":\"What happens next\",\"orientation\":\"auto\",\"steps\":[{\"label\":\"State\",\"description\":\"The game reports: health 24%, one enemy nearby, two healing items available.\"},{\"label\":\"RAG\",\"description\":\"The knowledge system retrieves the relevant rules for the healing item and perhaps information about the current weapon or tactical mechanic.\"},{\"label\":\"LLM\",\"description\":\"The model combines your request, the current state and the retrieved knowledge.\"},{\"label\":\"Decision\",\"description\":\"It concludes that healing first is safer than attacking immediately.\"},{\"label\":\"Tool \u002F game engine\",\"description\":\"The agent requests a legal game action such as moving to cover or using the healing item.\"},{\"label\":\"New state\",\"description\":\"The game executes the action and reports the updated situation back to the agent.\"}]},\"tunes\":{}},{\"id\":\"p-complete-2\",\"type\":\"paragraph\",\"data\":{\"text\":\"RAG did not control the character. The state database did not reason. The LLM did not directly change the game. Each part had one job.\"},\"tunes\":{}},{\"id\":\"h-why\",\"type\":\"header\",\"data\":{\"text\":\"Why use RAG at all?\",\"level\":2},\"tunes\":{}},{\"id\":\"p-why-1\",\"type\":\"paragraph\",\"data\":{\"text\":\"Because putting every document, rule and database record into every prompt would be slow, expensive and often confusing.\"},\"tunes\":{}},{\"id\":\"p-why-2\",\"type\":\"paragraph\",\"data\":{\"text\":\"RAG lets the system select only the information that is useful for the current question.\"},\"tunes\":{}},{\"id\":\"p-why-3\",\"type\":\"paragraph\",\"data\":{\"text\":\"It also lets you update the knowledge base without retraining the entire language model. Change the document or database, rebuild or refresh the index when necessary, and the next retrieval can use the newer information.\"},\"tunes\":{}},{\"id\":\"h-not-guarantee\",\"type\":\"header\",\"data\":{\"text\":\"What RAG does not guarantee\",\"level\":2},\"tunes\":{}},{\"id\":\"p-not-1\",\"type\":\"paragraph\",\"data\":{\"text\":\"RAG can improve grounding, but it does not make an answer automatically correct.\"},\"tunes\":{}},{\"id\":\"p-not-2\",\"type\":\"paragraph\",\"data\":{\"text\":\"The retrieval step can find the wrong document. The correct document can be outdated. The LLM can misunderstand good evidence. Or the current state can have changed.\"},\"tunes\":{}},{\"id\":\"p-not-3\",\"type\":\"paragraph\",\"data\":{\"text\":\"A reliable system therefore has to validate retrieval, state freshness and the model's final reasoning separately.\"},\"tunes\":{}},{\"id\":\"h-mental\",\"type\":\"header\",\"data\":{\"text\":\"The easiest mental model to remember\",\"level\":2},\"tunes\":{}},{\"id\":\"mental-table\",\"type\":\"comparison\",\"data\":{\"title\":\"Think of an AI system like a person at a desk\",\"layout\":\"table\",\"columns\":[{\"id\":\"analogy\",\"label\":\"Analogy\"},{\"id\":\"system\",\"label\":\"AI system\"}],\"rows\":[{\"id\":\"brain\",\"label\":\"Person thinking\",\"values\":[\"\",\"\"]},{\"id\":\"library\",\"label\":\"Finding a reference book\",\"values\":[\"\",\"\"]},{\"id\":\"books\",\"label\":\"Books on the shelf\",\"values\":[\"\",\"\"]},{\"id\":\"dashboard\",\"label\":\"Current dashboard or instrument panel\",\"values\":[\"\",\"\"]},{\"id\":\"notes\",\"label\":\"Notes from earlier meetings\",\"values\":[\"\",\"\"]},{\"id\":\"hands\",\"label\":\"Doing something in the real world\",\"values\":[\"\",\"\"]}]},\"tunes\":{}},{\"id\":\"remember\",\"type\":\"callout\",\"data\":{\"variant\":\"success\",\"title\":\"If you remember only this\",\"body\":\"\u003Cstrong>LLM = brain.\u003C\u002Fstrong>\u003Cbr>\u003Cstrong>RAG = librarian.\u003C\u002Fstrong>\u003Cbr>\u003Cstrong>Knowledge base = library.\u003C\u002Fstrong>\u003Cbr>\u003Cstrong>State = what the dashboard says right now.\u003C\u002Fstrong>\u003Cbr>\u003Cstrong>Tools = the hands that can actually do something.\u003C\u002Fstrong>\"},\"tunes\":{}},{\"id\":\"h-conclusion\",\"type\":\"header\",\"data\":{\"text\":\"Conclusion\",\"level\":2},\"tunes\":{}},{\"id\":\"p-conc-1\",\"type\":\"paragraph\",\"data\":{\"text\":\"RAG is much less mysterious once the parts are separated.\"},\"tunes\":{}},{\"id\":\"p-conc-2\",\"type\":\"paragraph\",\"data\":{\"text\":\"The LLM understands and generates language. The application maintains current state. The knowledge base stores information. RAG finds the useful part of that information and puts it into the LLM's context. Tools or the application perform real actions.\"},\"tunes\":{}},{\"id\":\"p-conc-3\",\"type\":\"paragraph\",\"data\":{\"text\":\"That is the basic architecture behind many modern AI assistants and agents.\"},\"tunes\":{}},{\"id\":\"h-faq\",\"type\":\"header\",\"data\":{\"text\":\"FAQ\",\"level\":2},\"tunes\":{}},{\"id\":\"faq\",\"type\":\"faq\",\"data\":{\"title\":\"RAG in plain English\",\"items\":[{\"id\":\"faq1\",\"question\":\"What is RAG in simple terms?\",\"answer\":\"RAG is a step where an AI searches a knowledge source for relevant information before the language model writes its answer.\"},{\"id\":\"faq2\",\"question\":\"Does RAG need the Internet?\",\"answer\":\"No. The knowledge base can be completely local on your computer or server.\"},{\"id\":\"faq3\",\"question\":\"Is RAG the same as a database?\",\"answer\":\"No. The database or files contain the information. RAG is the retrieval process that finds the useful part and gives it to the LLM.\"},{\"id\":\"faq4\",\"question\":\"Is RAG the same as memory?\",\"answer\":\"No. Memory usually stores previous interactions or events. RAG retrieves relevant knowledge when it is needed.\"},{\"id\":\"faq5\",\"question\":\"Is current application state part of RAG?\",\"answer\":\"Not necessarily. Current state is usually obtained directly from the application or a state store. RAG is better understood as retrieval from a knowledge source.\"},{\"id\":\"faq6\",\"question\":\"Does RAG make AI answers correct?\",\"answer\":\"No. It can provide better evidence, but retrieval can still be wrong or outdated and the LLM can still reason incorrectly.\"}]},\"tunes\":{}},{\"id\":\"h-glossary\",\"type\":\"header\",\"data\":{\"text\":\"Glossary\",\"level\":2},\"tunes\":{}},{\"id\":\"glossary\",\"type\":\"glossary\",\"data\":{\"title\":\"The basic terms\",\"entries\":[{\"term\":\"LLM\",\"definition\":\"A language model that understands and generates text and can reason over information placed in its context.\",\"anchor\":\"llm\"},{\"term\":\"RAG\",\"definition\":\"Retrieval-Augmented Generation: retrieving relevant external information and adding it to the model's context before generating an answer.\",\"anchor\":\"rag\"},{\"term\":\"Knowledge base\",\"definition\":\"The files, documents, records or other information that retrieval can search.\",\"anchor\":\"knowledge-base\"},{\"term\":\"State\",\"definition\":\"The current facts of an application, system or world at a particular moment.\",\"anchor\":\"state\"},{\"term\":\"Context\",\"definition\":\"The information currently supplied to the language model for one request or reasoning step.\",\"anchor\":\"context\"},{\"term\":\"Embedding\",\"definition\":\"A numerical representation of meaning that can help semantic search find conceptually similar information.\",\"anchor\":\"embedding\"}]},\"tunes\":{}},{\"id\":\"h-sources\",\"type\":\"header\",\"data\":{\"text\":\"Primary sources\",\"level\":2},\"tunes\":{}},{\"id\":\"src-openai-vector\",\"type\":\"linkTool\",\"data\":{\"link\":\"https:\u002F\u002Fplatform.openai.com\u002Fdocs\u002Fapi-reference\u002Fvector-stores-files\",\"meta\":{\"image\":{\"url\":\"\"},\"title\":\"OpenAI — Vector Store Files\",\"description\":\"Official documentation showing how files can be attached to vector stores, chunked and made available to file-search retrieval.\"}},\"tunes\":{}},{\"id\":\"src-openai-quickstart\",\"type\":\"linkTool\",\"data\":{\"link\":\"https:\u002F\u002Fplatform.openai.com\u002Fdocs\u002Fquickstart\",\"meta\":{\"image\":{\"url\":\"\"},\"title\":\"OpenAI — Developer Quickstart\",\"description\":\"Official OpenAI documentation describing tools such as file search for giving models access to external information.\"}},\"tunes\":{}},{\"id\":\"src-nvidia-ace\",\"type\":\"linkTool\",\"data\":{\"link\":\"https:\u002F\u002Fdeveloper.nvidia.com\u002Face-for-games\",\"meta\":{\"image\":{\"url\":\"\"},\"title\":\"NVIDIA Developer — ACE for Games\",\"description\":\"Official NVIDIA documentation describing separate Agent, Chat and RAG APIs for connecting game characters to game state, contextual knowledge and model-driven actions.\"}},\"tunes\":{}},{\"id\":\"src-nvidia-pubg\",\"type\":\"linkTool\",\"data\":{\"link\":\"https:\u002F\u002Fdeveloper.nvidia.com\u002Fblog\u002Fhow-krafton-built-pubg-ally-a-co-playable-character-powered-by-nvidia-ace\u002F\",\"meta\":{\"image\":{\"url\":\"\"},\"title\":\"NVIDIA Developer — How KRAFTON Built PUBG Ally\",\"description\":\"Official technical explanation separating live match state from knowledge lookup and language-model reasoning.\"}},\"tunes\":{}}],\"version\":\"2.31.6\"}",{"time":926,"blocks":927,"version":1448},1790377494031,[928,932,937,941,945,949,953,957,961,966,970,974,978,982,987,991,1008,1012,1016,1020,1024,1043,1047,1051,1055,1059,1063,1067,1089,1094,1098,1102,1106,1110,1114,1118,1122,1140,1144,1149,1153,1156,1160,1164,1168,1172,1176,1180,1184,1188,1192,1196,1200,1226,1230,1234,1238,1242,1246,1252,1256,1260,1280,1284,1288,1292,1296,1300,1304,1308,1312,1316,1320,1348,1353,1357,1361,1365,1369,1372,1395,1399,1416,1420,1427,1434,1441],{"id":215,"data":929,"type":218,"tunes":931},{"text":930},"RAG sounds complicated because the name is complicated. The idea is not. RAG simply means: before the AI answers, it first looks up relevant information from a knowledge source and gives that information to the language model.",{},{"id":221,"data":933,"type":226,"tunes":936},{"body":934,"title":935,"variant":225},"\u003Cstrong>RAG is the step where an AI searches a knowledge base for useful information before the LLM writes the answer.\u003C\u002Fstrong>","RAG in one sentence",{},{"id":229,"data":938,"type":218,"tunes":940},{"text":939},"Think of an LLM as a smart person sitting at a desk. RAG is the librarian who brings the right page from the right book. The LLM then reads that page and answers you.",{},{"id":234,"data":942,"type":239,"tunes":944},{"title":943,"maxLevel":237,"minLevel":238},"Contents",{},{"id":242,"data":946,"type":42,"tunes":948},{"text":947,"level":238},"First: what does the LLM do?",{},{"id":247,"data":950,"type":218,"tunes":952},{"text":951},"The LLM is the part that understands language and produces language. It can read your question, understand instructions, compare information, explain something and write an answer.",{},{"id":252,"data":954,"type":218,"tunes":956},{"text":955},"But the LLM does not automatically know what is currently inside your company database, your game session, your private documents or a file you created five minutes ago.",{},{"id":257,"data":958,"type":218,"tunes":960},{"text":959},"It only knows what is already inside the model plus whatever information the application gives it in the current request.",{},{"id":262,"data":962,"type":226,"tunes":965},{"body":963,"title":964,"variant":266},"The LLM \u003Cstrong>thinks and writes\u003C\u002Fstrong>. It does not automatically own all of your current data.","Simple rule",{},{"id":269,"data":967,"type":42,"tunes":969},{"text":968,"level":238},"Then: what is the knowledge base?",{},{"id":274,"data":971,"type":218,"tunes":973},{"text":972},"A knowledge base is simply information the application can search.",{},{"id":279,"data":975,"type":218,"tunes":977},{"text":976},"It could contain PDFs, manuals, product documentation, support articles, contracts, game rules, weapon data, internal company documents, database records or other text.",{},{"id":284,"data":979,"type":218,"tunes":981},{"text":980},"The knowledge base can be local on your own machine. It can be on a server. It can be in a vector database. It can also be built from normal files. RAG does not mean Internet.",{},{"id":289,"data":983,"type":226,"tunes":986},{"body":984,"title":985,"variant":293},"\u003Cstrong>RAG does not require the Internet.\u003C\u002Fstrong> The information can be completely local.","Important",{},{"id":296,"data":988,"type":42,"tunes":990},{"text":989,"level":238},"So what does RAG actually do?",{},{"id":301,"data":992,"type":318,"tunes":1007},{"steps":993,"title":1006,"orientation":317},[994,997,1000,1003],{"label":995,"description":996},"1. You ask a question","For example: Which ammunition does this weapon use?",{"label":998,"description":999},"2. RAG searches the knowledge base","The system looks for the small pieces of information most relevant to your question.",{"label":1001,"description":1002},"3. RAG gives those pieces to the LLM","The LLM receives the question plus the retrieved information.",{"label":1004,"description":1005},"4. The LLM writes the answer","It uses the retrieved information as context for the response.","The whole RAG process",{},{"id":321,"data":1009,"type":218,"tunes":1011},{"text":1010},"That is RAG.",{},{"id":326,"data":1013,"type":218,"tunes":1015},{"text":1014},"The full name is Retrieval-Augmented Generation. Retrieval means finding the relevant information. Augmented means adding that information to the model's context. Generation means the LLM writes the final answer.",{},{"id":331,"data":1017,"type":42,"tunes":1019},{"text":1018,"level":238},"A very simple example",{},{"id":336,"data":1021,"type":218,"tunes":1023},{"text":1022},"Imagine you have a local knowledge base about a game.",{},{"id":341,"data":1025,"type":359,"tunes":1042},{"content":1026,"stretched":43,"withHeadings":14},[1027,1030,1033,1036,1039],[1028,1029],"Knowledge base contains","Example",[1031,1032],"Weapons","AKM uses 7.62 mm ammunition",[1034,1035],"Healing items","Med Kit restores health",[1037,1038],"Attachments","This attachment works with these weapons",[1040,1041],"Map rules","This zone behaves in this way",{},{"id":362,"data":1044,"type":218,"tunes":1046},{"text":1045},"You ask: “Which ammunition does the AKM use?”",{},{"id":367,"data":1048,"type":218,"tunes":1050},{"text":1049},"RAG searches the knowledge base and finds the entry about the AKM. It gives that small piece of information to the LLM. The LLM then answers: “The AKM uses 7.62 mm ammunition.”",{},{"id":372,"data":1052,"type":218,"tunes":1054},{"text":1053},"The LLM did not need the entire database. RAG only brought the useful part.",{},{"id":377,"data":1056,"type":42,"tunes":1058},{"text":1057,"level":238},"Now the important part: RAG is not the current state",{},{"id":382,"data":1060,"type":218,"tunes":1062},{"text":1061},"This is where many explanations become confusing.",{},{"id":387,"data":1064,"type":218,"tunes":1066},{"text":1065},"RAG usually gives the AI knowledge. A state system gives the AI facts about what is true right now.",{},{"id":392,"data":1068,"type":420,"tunes":1088},{"rows":1069,"title":1082,"layout":359,"columns":1083},[1070,1073,1076,1079],{"id":396,"label":1071,"values":1072},"Weapon",[399,399],{"id":401,"label":1074,"values":1075},"Ammunition",[399,399],{"id":405,"label":1077,"values":1078},"Health",[399,399],{"id":409,"label":1080,"values":1081},"Enemy",[399,399],"Knowledge vs current state",[1084,1086],{"id":415,"label":1085},"RAG \u002F knowledge",{"id":418,"label":1087},"Current state",{},{"id":423,"data":1090,"type":226,"tunes":1093},{"body":1091,"title":1092,"variant":427},"RAG answers: \u003Cstrong>What is generally true?\u003C\u002Fstrong>\u003Cbr>State answers: \u003Cstrong>What is true right now?\u003C\u002Fstrong>","Do not mix these two",{},{"id":430,"data":1095,"type":42,"tunes":1097},{"text":1096,"level":238},"What is a state database?",{},{"id":435,"data":1099,"type":218,"tunes":1101},{"text":1100},"A state database or state store is simply a place where the application keeps current facts.",{},{"id":440,"data":1103,"type":218,"tunes":1105},{"text":1104},"In a game, the engine already knows things such as your health, position, inventory, ammunition, current mission, nearby objects and enemy status. An AI system can expose selected parts of that state to the model.",{},{"id":445,"data":1107,"type":218,"tunes":1109},{"text":1108},"In a business application, the same idea could be an order database, a customer record, a project status or the current value of a sensor.",{},{"id":450,"data":1111,"type":218,"tunes":1113},{"text":1112},"The state is created by the application itself as things happen. If you lose health, the game updates the health value. If you pick up ammunition, the inventory changes. If an order is paid, the business system changes the order status.",{},{"id":455,"data":1115,"type":226,"tunes":1117},{"body":1116,"title":964,"variant":225},"The application creates and updates \u003Cstrong>state\u003C\u002Fstrong>. RAG searches \u003Cstrong>knowledge\u003C\u002Fstrong>. The LLM uses both to decide what to say or do.",{},{"id":460,"data":1119,"type":42,"tunes":1121},{"text":1120,"level":238},"How the three pieces work together",{},{"id":465,"data":1123,"type":318,"tunes":1139},{"steps":1124,"title":1138,"orientation":317},[1125,1128,1130,1132,1135],{"label":1126,"description":1127},"1. Current state","The application tells the AI what is true now: health 41%, AKM equipped, 23 rounds.",{"label":472,"description":1129},"The system retrieves useful knowledge: how the weapon works, which healing item is available, or a relevant rule.",{"label":475,"description":1131},"The model receives the question, current state and retrieved knowledge.",{"label":1133,"description":1134},"4. Reasoning","The LLM combines those inputs and decides what answer or high-level action makes sense.",{"label":1136,"description":1137},"5. Application","If an action is required, the application or game engine executes it and updates the state again.","LLM + state + RAG",{},{"id":486,"data":1141,"type":218,"tunes":1143},{"text":1142},"So the basic architecture is:",{},{"id":491,"data":1145,"type":226,"tunes":1148},{"body":1146,"title":1147,"variant":266},"\u003Cstrong>State = what is true now\u003C\u002Fstrong>\u003Cbr>\u003Cstrong>RAG = useful knowledge\u003C\u002Fstrong>\u003Cbr>\u003Cstrong>LLM = understands, reasons and writes\u003C\u002Fstrong>\u003Cbr>\u003Cstrong>Application = performs the real action\u003C\u002Fstrong>","The simplest architecture",{},{"id":497,"data":1150,"type":42,"tunes":1152},{"text":1151,"level":238},"Does RAG always use a vector database?",{},{"id":502,"data":1154,"type":218,"tunes":1155},{"text":504},{},{"id":507,"data":1157,"type":218,"tunes":1159},{"text":1158},"A vector database is a common way to build semantic search, but it is not the definition of RAG.",{},{"id":512,"data":1161,"type":218,"tunes":1163},{"text":1162},"The important part is retrieval: the system finds relevant external information and adds it to the LLM's context before the answer is generated.",{},{"id":517,"data":1165,"type":218,"tunes":1167},{"text":1166},"OpenAI's File Search, for example, can work with files stored in vector stores. Files are chunked into smaller pieces so the system can retrieve the parts that are relevant to a question. That is one implementation of the same basic idea.",{},{"id":522,"data":1169,"type":42,"tunes":1171},{"text":1170,"level":238},"What is an embedding, in plain English?",{},{"id":527,"data":1173,"type":218,"tunes":1175},{"text":1174},"You do not need to understand embeddings to understand RAG.",{},{"id":532,"data":1177,"type":218,"tunes":1179},{"text":1178},"But the simple version is this: an embedding is a numerical representation of meaning. It helps a search system find text that is conceptually similar even when the words are not exactly the same.",{},{"id":537,"data":1181,"type":218,"tunes":1183},{"text":1182},"For example, a normal keyword search may look for the exact words “car repair.” Semantic search can also understand that “fix my vehicle” is about a similar topic.",{},{"id":542,"data":1185,"type":218,"tunes":1187},{"text":1186},"That makes embeddings useful for RAG, but RAG can also use keyword search, database queries or a hybrid of several methods.",{},{"id":547,"data":1189,"type":42,"tunes":1191},{"text":1190,"level":238},"RAG is not memory either",{},{"id":552,"data":1193,"type":218,"tunes":1195},{"text":1194},"Memory is another concept that is often mixed together with RAG.",{},{"id":557,"data":1197,"type":218,"tunes":1199},{"text":1198},"Memory is usually information the system keeps about previous interactions or previous events. RAG is the mechanism used to retrieve relevant knowledge when it is needed.",{},{"id":562,"data":1201,"type":359,"tunes":1225},{"content":1202,"stretched":43,"withHeadings":14},[1203,1206,1208,1210,1213,1216,1219,1222],[1204,1205],"Part","Simple meaning",[569,1207],"The part that understands and generates language",[572,1209],"The part that looks up relevant knowledge before the answer",[1211,1212],"Knowledge base","The information RAG can search",[1214,1215],"State","What is true right now in the application or world",[1217,1218],"Memory","Information kept from previous interactions or events",[1220,1221],"Tool \u002F action","Something the AI is allowed to call or ask the application to do",[1223,1224],"Context","The information currently placed in front of the LLM for this request",{},{"id":591,"data":1227,"type":42,"tunes":1229},{"text":1228,"level":238},"A real game example: PUBG Ally",{},{"id":596,"data":1231,"type":218,"tunes":1233},{"text":1232},"PUBG Ally is a useful example because it makes the difference visible.",{},{"id":601,"data":1235,"type":218,"tunes":1237},{"text":1236},"KRAFTON describes live match state as a separate source of truth. The game exposes current facts through observation tools: current weapon, ammunition, health, safe-zone status, nearby items and combat situation.",{},{"id":606,"data":1239,"type":218,"tunes":1241},{"text":1240},"Knowledge lookup is a different job. The system can use curated knowledge about weapons, attachments, items and rules. NVIDIA's ACE Game Agent SDK also exposes a separate RAG API for retrieving knowledge from developer-built databases.",{},{"id":611,"data":1243,"type":218,"tunes":1245},{"text":1244},"That gives us the clean separation: the game engine says what is happening now, retrieval provides relevant knowledge, and the language model decides what the information means.",{},{"id":616,"data":1247,"type":622,"tunes":1251},{"url":618,"title":1248,"excerpt":1249,"ctaLabel":1250},"PUBG Ally Shows Why AI Teammates Need Two Brains: Fast Reflexes and Slow Reasoning","A practical game example showing how live state, language reasoning and deterministic game-side control can work together.","Read the PUBG Ally architecture article",{},{"id":625,"data":1253,"type":42,"tunes":1255},{"text":1254,"level":238},"One complete example",{},{"id":630,"data":1257,"type":218,"tunes":1259},{"text":1258},"Imagine you tell an AI teammate: “I am low on health. Should we attack?”",{},{"id":635,"data":1261,"type":318,"tunes":1279},{"steps":1262,"title":1278,"orientation":317},[1263,1265,1267,1269,1272,1275],{"label":1214,"description":1264},"The game reports: health 24%, one enemy nearby, two healing items available.",{"label":572,"description":1266},"The knowledge system retrieves the relevant rules for the healing item and perhaps information about the current weapon or tactical mechanic.",{"label":569,"description":1268},"The model combines your request, the current state and the retrieved knowledge.",{"label":1270,"description":1271},"Decision","It concludes that healing first is safer than attacking immediately.",{"label":1273,"description":1274},"Tool \u002F game engine","The agent requests a legal game action such as moving to cover or using the healing item.",{"label":1276,"description":1277},"New state","The game executes the action and reports the updated situation back to the agent.","What happens next",{},{"id":656,"data":1281,"type":218,"tunes":1283},{"text":1282},"RAG did not control the character. The state database did not reason. The LLM did not directly change the game. Each part had one job.",{},{"id":661,"data":1285,"type":42,"tunes":1287},{"text":1286,"level":238},"Why use RAG at all?",{},{"id":666,"data":1289,"type":218,"tunes":1291},{"text":1290},"Because putting every document, rule and database record into every prompt would be slow, expensive and often confusing.",{},{"id":671,"data":1293,"type":218,"tunes":1295},{"text":1294},"RAG lets the system select only the information that is useful for the current question.",{},{"id":676,"data":1297,"type":218,"tunes":1299},{"text":1298},"It also lets you update the knowledge base without retraining the entire language model. Change the document or database, rebuild or refresh the index when necessary, and the next retrieval can use the newer information.",{},{"id":681,"data":1301,"type":42,"tunes":1303},{"text":1302,"level":238},"What RAG does not guarantee",{},{"id":686,"data":1305,"type":218,"tunes":1307},{"text":1306},"RAG can improve grounding, but it does not make an answer automatically correct.",{},{"id":691,"data":1309,"type":218,"tunes":1311},{"text":1310},"The retrieval step can find the wrong document. The correct document can be outdated. The LLM can misunderstand good evidence. Or the current state can have changed.",{},{"id":696,"data":1313,"type":218,"tunes":1315},{"text":1314},"A reliable system therefore has to validate retrieval, state freshness and the model's final reasoning separately.",{},{"id":701,"data":1317,"type":42,"tunes":1319},{"text":1318,"level":238},"The easiest mental model to remember",{},{"id":706,"data":1321,"type":420,"tunes":1347},{"rows":1322,"title":1341,"layout":359,"columns":1342},[1323,1326,1329,1332,1335,1338],{"id":710,"label":1324,"values":1325},"Person thinking",[399,399],{"id":714,"label":1327,"values":1328},"Finding a reference book",[399,399],{"id":718,"label":1330,"values":1331},"Books on the shelf",[399,399],{"id":722,"label":1333,"values":1334},"Current dashboard or instrument panel",[399,399],{"id":726,"label":1336,"values":1337},"Notes from earlier meetings",[399,399],{"id":730,"label":1339,"values":1340},"Doing something in the real world",[399,399],"Think of an AI system like a person at a desk",[1343,1345],{"id":229,"label":1344},"Analogy",{"id":738,"label":1346},"AI system",{},{"id":742,"data":1349,"type":226,"tunes":1352},{"body":1350,"title":1351,"variant":293},"\u003Cstrong>LLM = brain.\u003C\u002Fstrong>\u003Cbr>\u003Cstrong>RAG = librarian.\u003C\u002Fstrong>\u003Cbr>\u003Cstrong>Knowledge base = library.\u003C\u002Fstrong>\u003Cbr>\u003Cstrong>State = what the dashboard says right now.\u003C\u002Fstrong>\u003Cbr>\u003Cstrong>Tools = the hands that can actually do something.\u003C\u002Fstrong>","If you remember only this",{},{"id":748,"data":1354,"type":42,"tunes":1356},{"text":1355,"level":238},"Conclusion",{},{"id":753,"data":1358,"type":218,"tunes":1360},{"text":1359},"RAG is much less mysterious once the parts are separated.",{},{"id":758,"data":1362,"type":218,"tunes":1364},{"text":1363},"The LLM understands and generates language. The application maintains current state. The knowledge base stores information. RAG finds the useful part of that information and puts it into the LLM's context. Tools or the application perform real actions.",{},{"id":763,"data":1366,"type":218,"tunes":1368},{"text":1367},"That is the basic architecture behind many modern AI assistants and agents.",{},{"id":768,"data":1370,"type":42,"tunes":1371},{"text":770,"level":238},{},{"id":773,"data":1373,"type":773,"tunes":1394},{"items":1374,"title":1393},[1375,1378,1381,1384,1387,1390],{"id":777,"answer":1376,"question":1377},"RAG is a step where an AI searches a knowledge source for relevant information before the language model writes its answer.","What is RAG in simple terms?",{"id":781,"answer":1379,"question":1380},"No. The knowledge base can be completely local on your computer or server.","Does RAG need the Internet?",{"id":785,"answer":1382,"question":1383},"No. The database or files contain the information. RAG is the retrieval process that finds the useful part and gives it to the LLM.","Is RAG the same as a database?",{"id":789,"answer":1385,"question":1386},"No. Memory usually stores previous interactions or events. RAG retrieves relevant knowledge when it is needed.","Is RAG the same as memory?",{"id":793,"answer":1388,"question":1389},"Not necessarily. Current state is usually obtained directly from the application or a state store. RAG is better understood as retrieval from a knowledge source.","Is current application state part of RAG?",{"id":797,"answer":1391,"question":1392},"No. It can provide better evidence, but retrieval can still be wrong or outdated and the LLM can still reason incorrectly.","Does RAG make AI answers correct?","RAG in plain English",{},{"id":803,"data":1396,"type":42,"tunes":1398},{"text":1397,"level":238},"Glossary",{},{"id":808,"data":1400,"type":808,"tunes":1415},{"title":1401,"entries":1402},"The basic terms",[1403,1405,1407,1409,1411,1413],{"term":569,"anchor":813,"definition":1404},"A language model that understands and generates text and can reason over information placed in its context.",{"term":572,"anchor":816,"definition":1406},"Retrieval-Augmented Generation: retrieving relevant external information and adding it to the model's context before generating an answer.",{"term":1211,"anchor":819,"definition":1408},"The files, documents, records or other information that retrieval can search.",{"term":1214,"anchor":418,"definition":1410},"The current facts of an application, system or world at a particular moment.",{"term":1223,"anchor":824,"definition":1412},"The information currently supplied to the language model for one request or reasoning step.",{"term":827,"anchor":828,"definition":1414},"A numerical representation of meaning that can help semantic search find conceptually similar information.",{},{"id":832,"data":1417,"type":42,"tunes":1419},{"text":1418,"level":238},"Primary sources",{},{"id":837,"data":1421,"type":844,"tunes":1426},{"link":839,"meta":1422},{"image":1423,"title":1424,"description":1425},{"url":399},"OpenAI — Vector Store Files","Official documentation showing how files can be attached to vector stores, chunked and made available to file-search retrieval.",{},{"id":847,"data":1428,"type":844,"tunes":1433},{"link":849,"meta":1429},{"image":1430,"title":1431,"description":1432},{"url":399},"OpenAI — Developer Quickstart","Official OpenAI documentation describing tools such as file search for giving models access to external information.",{},{"id":856,"data":1435,"type":844,"tunes":1440},{"link":858,"meta":1436},{"image":1437,"title":1438,"description":1439},{"url":399},"NVIDIA Developer — ACE for Games","Official NVIDIA documentation describing separate Agent, Chat and RAG APIs for connecting game characters to game state, contextual knowledge and model-driven actions.",{},{"id":865,"data":1442,"type":844,"tunes":1447},{"link":867,"meta":1443},{"image":1444,"title":1445,"description":1446},{"url":399},"NVIDIA Developer — How KRAFTON Built PUBG Ally","Official technical explanation separating live match state from knowledge lookup and language-model reasoning.",{},"2.31.6","RAG sounds complicated, but the idea is simple: before an AI answers, it first looks up useful information from a knowledge source and gives that information to the language model. This guide explains RAG, LLMs, state, memory and tools using one simple mental model.",{"lang":7,"title":208,"content":210,"contentJson":1451,"excerpt":874},{"time":212,"blocks":1452,"version":873},[1453,1456,1459,1462,1465,1468,1471,1474,1477,1480,1483,1486,1489,1492,1495,1498,1506,1509,1512,1515,1518,1527,1530,1533,1536,1539,1542,1545,1560,1563,1566,1569,1572,1575,1578,1581,1584,1593,1596,1599,1602,1605,1608,1611,1614,1617,1620,1623,1626,1629,1632,1635,1638,1650,1653,1656,1659,1662,1665,1668,1671,1674,1684,1687,1690,1693,1696,1699,1702,1705,1708,1711,1714,1733,1736,1739,1742,1745,1748,1751,1761,1764,1774,1777,1782,1787,1792],{"id":215,"data":1454,"type":218,"tunes":1455},{"text":217},{},{"id":221,"data":1457,"type":226,"tunes":1458},{"body":223,"title":224,"variant":225},{},{"id":229,"data":1460,"type":218,"tunes":1461},{"text":231},{},{"id":234,"data":1463,"type":239,"tunes":1464},{"title":236,"maxLevel":237,"minLevel":238},{},{"id":242,"data":1466,"type":42,"tunes":1467},{"text":244,"level":238},{},{"id":247,"data":1469,"type":218,"tunes":1470},{"text":249},{},{"id":252,"data":1472,"type":218,"tunes":1473},{"text":254},{},{"id":257,"data":1475,"type":218,"tunes":1476},{"text":259},{},{"id":262,"data":1478,"type":226,"tunes":1479},{"body":264,"title":265,"variant":266},{},{"id":269,"data":1481,"type":42,"tunes":1482},{"text":271,"level":238},{},{"id":274,"data":1484,"type":218,"tunes":1485},{"text":276},{},{"id":279,"data":1487,"type":218,"tunes":1488},{"text":281},{},{"id":284,"data":1490,"type":218,"tunes":1491},{"text":286},{},{"id":289,"data":1493,"type":226,"tunes":1494},{"body":291,"title":292,"variant":293},{},{"id":296,"data":1496,"type":42,"tunes":1497},{"text":298,"level":238},{},{"id":301,"data":1499,"type":318,"tunes":1505},{"steps":1500,"title":316,"orientation":317},[1501,1502,1503,1504],{"label":305,"description":306},{"label":308,"description":309},{"label":311,"description":312},{"label":314,"description":315},{},{"id":321,"data":1507,"type":218,"tunes":1508},{"text":323},{},{"id":326,"data":1510,"type":218,"tunes":1511},{"text":328},{},{"id":331,"data":1513,"type":42,"tunes":1514},{"text":333,"level":238},{},{"id":336,"data":1516,"type":218,"tunes":1517},{"text":338},{},{"id":341,"data":1519,"type":359,"tunes":1526},{"content":1520,"stretched":43,"withHeadings":14},[1521,1522,1523,1524,1525],[345,346],[348,349],[351,352],[354,355],[357,358],{},{"id":362,"data":1528,"type":218,"tunes":1529},{"text":364},{},{"id":367,"data":1531,"type":218,"tunes":1532},{"text":369},{},{"id":372,"data":1534,"type":218,"tunes":1535},{"text":374},{},{"id":377,"data":1537,"type":42,"tunes":1538},{"text":379,"level":238},{},{"id":382,"data":1540,"type":218,"tunes":1541},{"text":384},{},{"id":387,"data":1543,"type":218,"tunes":1544},{"text":389},{},{"id":392,"data":1546,"type":420,"tunes":1559},{"rows":1547,"title":412,"layout":359,"columns":1556},[1548,1550,1552,1554],{"id":396,"label":397,"values":1549},[399,399],{"id":401,"label":402,"values":1551},[399,399],{"id":405,"label":406,"values":1553},[399,399],{"id":409,"label":410,"values":1555},[399,399],[1557,1558],{"id":415,"label":416},{"id":418,"label":419},{},{"id":423,"data":1561,"type":226,"tunes":1562},{"body":425,"title":426,"variant":427},{},{"id":430,"data":1564,"type":42,"tunes":1565},{"text":432,"level":238},{},{"id":435,"data":1567,"type":218,"tunes":1568},{"text":437},{},{"id":440,"data":1570,"type":218,"tunes":1571},{"text":442},{},{"id":445,"data":1573,"type":218,"tunes":1574},{"text":447},{},{"id":450,"data":1576,"type":218,"tunes":1577},{"text":452},{},{"id":455,"data":1579,"type":226,"tunes":1580},{"body":457,"title":265,"variant":225},{},{"id":460,"data":1582,"type":42,"tunes":1583},{"text":462,"level":238},{},{"id":465,"data":1585,"type":318,"tunes":1592},{"steps":1586,"title":483,"orientation":317},[1587,1588,1589,1590,1591],{"label":469,"description":470},{"label":472,"description":473},{"label":475,"description":476},{"label":478,"description":479},{"label":481,"description":482},{},{"id":486,"data":1594,"type":218,"tunes":1595},{"text":488},{},{"id":491,"data":1597,"type":226,"tunes":1598},{"body":493,"title":494,"variant":266},{},{"id":497,"data":1600,"type":42,"tunes":1601},{"text":499,"level":238},{},{"id":502,"data":1603,"type":218,"tunes":1604},{"text":504},{},{"id":507,"data":1606,"type":218,"tunes":1607},{"text":509},{},{"id":512,"data":1609,"type":218,"tunes":1610},{"text":514},{},{"id":517,"data":1612,"type":218,"tunes":1613},{"text":519},{},{"id":522,"data":1615,"type":42,"tunes":1616},{"text":524,"level":238},{},{"id":527,"data":1618,"type":218,"tunes":1619},{"text":529},{},{"id":532,"data":1621,"type":218,"tunes":1622},{"text":534},{},{"id":537,"data":1624,"type":218,"tunes":1625},{"text":539},{},{"id":542,"data":1627,"type":218,"tunes":1628},{"text":544},{},{"id":547,"data":1630,"type":42,"tunes":1631},{"text":549,"level":238},{},{"id":552,"data":1633,"type":218,"tunes":1634},{"text":554},{},{"id":557,"data":1636,"type":218,"tunes":1637},{"text":559},{},{"id":562,"data":1639,"type":359,"tunes":1649},{"content":1640,"stretched":43,"withHeadings":14},[1641,1642,1643,1644,1645,1646,1647,1648],[566,567],[569,570],[572,573],[575,576],[578,579],[581,582],[584,585],[587,588],{},{"id":591,"data":1651,"type":42,"tunes":1652},{"text":593,"level":238},{},{"id":596,"data":1654,"type":218,"tunes":1655},{"text":598},{},{"id":601,"data":1657,"type":218,"tunes":1658},{"text":603},{},{"id":606,"data":1660,"type":218,"tunes":1661},{"text":608},{},{"id":611,"data":1663,"type":218,"tunes":1664},{"text":613},{},{"id":616,"data":1666,"type":622,"tunes":1667},{"url":618,"title":619,"excerpt":620,"ctaLabel":621},{},{"id":625,"data":1669,"type":42,"tunes":1670},{"text":627,"level":238},{},{"id":630,"data":1672,"type":218,"tunes":1673},{"text":632},{},{"id":635,"data":1675,"type":318,"tunes":1683},{"steps":1676,"title":653,"orientation":317},[1677,1678,1679,1680,1681,1682],{"label":578,"description":639},{"label":572,"description":641},{"label":569,"description":643},{"label":645,"description":646},{"label":648,"description":649},{"label":651,"description":652},{},{"id":656,"data":1685,"type":218,"tunes":1686},{"text":658},{},{"id":661,"data":1688,"type":42,"tunes":1689},{"text":663,"level":238},{},{"id":666,"data":1691,"type":218,"tunes":1692},{"text":668},{},{"id":671,"data":1694,"type":218,"tunes":1695},{"text":673},{},{"id":676,"data":1697,"type":218,"tunes":1698},{"text":678},{},{"id":681,"data":1700,"type":42,"tunes":1701},{"text":683,"level":238},{},{"id":686,"data":1703,"type":218,"tunes":1704},{"text":688},{},{"id":691,"data":1706,"type":218,"tunes":1707},{"text":693},{},{"id":696,"data":1709,"type":218,"tunes":1710},{"text":698},{},{"id":701,"data":1712,"type":42,"tunes":1713},{"text":703,"level":238},{},{"id":706,"data":1715,"type":420,"tunes":1732},{"rows":1716,"title":733,"layout":359,"columns":1729},[1717,1719,1721,1723,1725,1727],{"id":710,"label":711,"values":1718},[399,399],{"id":714,"label":715,"values":1720},[399,399],{"id":718,"label":719,"values":1722},[399,399],{"id":722,"label":723,"values":1724},[399,399],{"id":726,"label":727,"values":1726},[399,399],{"id":730,"label":731,"values":1728},[399,399],[1730,1731],{"id":229,"label":736},{"id":738,"label":739},{},{"id":742,"data":1734,"type":226,"tunes":1735},{"body":744,"title":745,"variant":293},{},{"id":748,"data":1737,"type":42,"tunes":1738},{"text":750,"level":238},{},{"id":753,"data":1740,"type":218,"tunes":1741},{"text":755},{},{"id":758,"data":1743,"type":218,"tunes":1744},{"text":760},{},{"id":763,"data":1746,"type":218,"tunes":1747},{"text":765},{},{"id":768,"data":1749,"type":42,"tunes":1750},{"text":770,"level":238},{},{"id":773,"data":1752,"type":773,"tunes":1760},{"items":1753,"title":800},[1754,1755,1756,1757,1758,1759],{"id":777,"answer":778,"question":779},{"id":781,"answer":782,"question":783},{"id":785,"answer":786,"question":787},{"id":789,"answer":790,"question":791},{"id":793,"answer":794,"question":795},{"id":797,"answer":798,"question":799},{},{"id":803,"data":1762,"type":42,"tunes":1763},{"text":805,"level":238},{},{"id":808,"data":1765,"type":808,"tunes":1773},{"title":810,"entries":1766},[1767,1768,1769,1770,1771,1772],{"term":569,"anchor":813,"definition":814},{"term":572,"anchor":816,"definition":817},{"term":575,"anchor":819,"definition":820},{"term":578,"anchor":418,"definition":822},{"term":587,"anchor":824,"definition":825},{"term":827,"anchor":828,"definition":829},{},{"id":832,"data":1775,"type":42,"tunes":1776},{"text":834,"level":238},{},{"id":837,"data":1778,"type":844,"tunes":1781},{"link":839,"meta":1779},{"image":1780,"title":842,"description":843},{"url":399},{},{"id":847,"data":1783,"type":844,"tunes":1786},{"link":849,"meta":1784},{"image":1785,"title":852,"description":853},{"url":399},{},{"id":856,"data":1788,"type":844,"tunes":1791},{"link":858,"meta":1789},{"image":1790,"title":861,"description":862},{"url":399},{},{"id":865,"data":1793,"type":844,"tunes":1796},{"link":867,"meta":1794},{"image":1795,"title":870,"description":871},{"url":399},{},"Post erfolgreich abgerufen",{"items":1799,"source":1882,"manualIds":1883,"manualMatchedIds":1884},[1800,1807,1814,1821,1826,1833,1840,1847,1854,1861,1868,1875],{"id":1801,"slug":1802,"title":1803,"excerpt":1804,"featuredImage":1805,"publishedAt":1806},"472","why-more-context-can-make-ai-answers-worse","Perché più contesto può peggiorare le risposte dell'IA","Una finestra di contesto più ampia non garantisce una risposta migliore. Questo articolo spiega come la diluizione del segnale, le prove contrastanti, lo stato obsoleto, la sensibilità alla posizione e la compressione con perdita possano ridurre l'affidabilità dell'IA—e introduce un pratico Context Pressure Test.","\u002Fuploads\u002F2026\u002F09\u002Fwhy-more-context-can-make-ai-answers-worse-1790351615793-2ntv2v.webp","2026-09-25T11:51:00.000Z",{"id":1808,"slug":1809,"title":1810,"excerpt":1811,"featuredImage":1812,"publishedAt":1813},"470","what-should-an-ai-agent-remember-forget-recompute-or-retrieve-again","Cosa dovrebbe ricordare, dimenticare, ricalcolare o recuperare di nuovo un agente IA?","Gli agenti a lunga esecuzione non dovrebbero ricordare tutto. Questo articolo fornisce un modello pratico di ciclo di vita per decidere cosa appartiene alla memoria durevole, cosa dovrebbe essere recuperato di nuovo, cosa è più sicuro ricalcolare e cosa dovrebbe scadere o essere sostituito.","\u002Fuploads\u002F2026\u002F09\u002Fwhat-should-an-ai-agent-remember-forget-recompute-or-retrieve-again-1790351131087-iehz28.webp","2026-09-25T09:43:00.000Z",{"id":1815,"slug":1816,"title":1817,"excerpt":1818,"featuredImage":1819,"publishedAt":1820},"477","computer-use-agents-why-a-successful-demo-can-still-be-an-unreliable-system","Agenti per l'uso del computer: perché una demo di successo può comunque essere un sistema inaffidabile","Gli agenti computer-use possono ora completare impressionanti flussi di lavoro su browser e desktop, ma una singola esecuzione riuscita dimostra la capacità—non l'affidabilità. Questo articolo mostra come testare la ripetibilità, la robustezza ambientale, il controllo a lungo orizzonte, la consapevolezza dello stato, la verifica dei risultati e la gestione sicura degli obiettivi.","\u002Fuploads\u002F2026\u002F09\u002Fcomputer-use-agents-why-a-successful-demo-can-still-be-an-unreliable-system-1790352854690-75qnrg.webp","2026-09-25T12:13:00.000Z",{"id":1822,"slug":1823,"title":1823,"excerpt":10,"featuredImage":1824,"publishedAt":1825},"369","git-with-automatic-upload-and-synchronization-to-a-production-server","\u002Fuploads\u002F2024\u002F05\u002Fstep-by-step-guide-illustration-showing-the-process-of-setting-up-Git-with-auto-upload-and-synchronization-to-a-production-server-large.webp","2024-05-28T22:48:00.000Z",{"id":1827,"slug":1828,"title":1829,"excerpt":1830,"featuredImage":1831,"publishedAt":1832},"473","openai-agents-api-vs-agents-sdk-vs-responses-api-what-should-you-build-on-in-2026","OpenAI Agents API vs Agents SDK vs Responses API: Su cosa dovresti sviluppare nel 2026?","Lo stack di agenti di OpenAI è cambiato a settembre 2026. Questa guida all'architettura separa Agents API, Agents SDK, Responses API e Codex SDK in base alla proprietà del runtime—in modo che i team possano scegliere il giusto confine di controllo invece di confrontare i nomi dei prodotti.","\u002Fuploads\u002F2026\u002F09\u002Fopenai-agents-api-vs-agents-sdk-vs-responses-api-what-should-you-build-on-in-2026-1790351846714-zi7lus.webp","2026-09-25T11:56:00.000Z",{"id":1834,"slug":1835,"title":1836,"excerpt":1837,"featuredImage":1838,"publishedAt":1839},"476","mcp-vs-a2a-vs-ucp-vs-ap2-vs-a2ui-the-agent-protocol-stack-explained","MCP vs A2A vs UCP vs AP2 vs A2UI: Lo stack di protocolli degli agenti spiegato","MCP, A2A, UCP, AP2 e A2UI sono spesso presentati come standard per agenti concorrenti. Per lo più risolvono problemi di interoperabilità diversi. Questa guida mappa ciascun protocollo sul confine che effettivamente standardizza—e mostra come possano lavorare insieme in un unico sistema di produzione.","\u002Fuploads\u002F2026\u002F09\u002Fmcp-vs-a2a-vs-ucp-vs-ap2-vs-a2ui-the-agent-protocol-stack-explained-1790352625869-2ezle0.webp","2026-09-25T12:09:00.000Z",{"id":1841,"slug":1842,"title":1843,"excerpt":1844,"featuredImage":1845,"publishedAt":1846},"467","the-answer-validity-boundary-the-missing-layer-between-relevance-and-reliable-ai-answers","Il confine della validità della risposta: il livello mancante tra rilevanza e risposte AI affidabili","Una fonte può essere pertinente, autorevole e comunque errata per la domanda posta. Il livello mancante è l'applicabilità: le condizioni alle quali una risposta è valida e i cambiamenti che ne impongono una riconsiderazione. Questo articolo introduce l'Answer Validity Boundary come modello di progettazione delle fonti per esseri umani, ricerca AI e sistemi RAG.","\u002Fuploads\u002F2026\u002F09\u002Fthe-answer-validity-boundary-the-missing-layer-between-relevance-and-reliable-ai-answers-1790272901306-1g5jly.webp","2026-09-24T11:59:00.000Z",{"id":1848,"slug":1849,"title":1850,"excerpt":1851,"featuredImage":1852,"publishedAt":1853},"460","ai-agent-reliability-why-the-final-answer-is-not-enough","Affidabilità degli Agenti AI: Perché la Risposta Finale Non è Sufficiente","Un output corretto non dimostra un ragionamento corretto, un'esecuzione sicura o un sistema affidabile.","\u002Fuploads\u002F2026\u002F09\u002Fai-agent-reliability-why-the-final-answer-is-not-enough-1788955466306-pl0qhz.webp","2026-09-09T04:01:00.000Z",{"id":1855,"slug":1856,"title":1857,"excerpt":1858,"featuredImage":1859,"publishedAt":1860},"469","rag-failed-but-which-layer-actually-failed-a-diagnostic-method","RAG non riuscito — Ma quale livello ha effettivamente fallito? Un metodo diagnostico","Quando una risposta RAG è sbagliata, dare la colpa al recupero o al modello è troppo vago. Questo metodo diagnostico isola la copertura delle fonti, la costruzione della query, il recupero, il ranking, l'assemblaggio del contesto, la generazione, l'attribuzione delle evidenze e l'aggiornamento—così il guasto effettivo può essere riprodotto e corretto.","\u002Fuploads\u002F2026\u002F09\u002Frag-failed-but-which-layer-actually-failed-a-diagnostic-method-1790350847177-pior4c.webp","2026-09-24T19:39:00.000Z",{"id":1862,"slug":1863,"title":1864,"excerpt":1865,"featuredImage":1866,"publishedAt":1867},"466","the-gpu-is-not-the-product-future-proof-private-ai-architecture","La GPU non è il prodotto: architettura di IA privata a prova di futuro","L'infrastruttura di IA privata non dovrebbe essere progettata attorno a una sola GPU o a un solo modello. Un approccio più resiliente combina GPU veloci per l'inferenza, sistemi di IA ricchi di memoria, nodi di IA fisica e modelli cloud di frontiera opzionali dietro un livello di routing consapevole delle capacità.","\u002Fuploads\u002F2026\u002F09\u002Fthe-gpu-is-not-the-product-future-proof-private-ai-architecture-1790140878812-8hsl39.webp","2026-09-23T01:19:00.000Z",{"id":1869,"slug":1870,"title":1871,"excerpt":1872,"featuredImage":1873,"publishedAt":1874},"459","ollama-is-not-the-product-building-production-ready-open-llm-applications","Ollama non è il prodotto: costruire applicazioni Open-LLM pronte per la produzione","Eseguire un modello locale con Ollama è facile. Costruire un'applicazione Open-LLM pronta per la produzione è più difficile: richiede RAG, controllo degli accessi, astrazione del provider, valutazione, logging, disciplina di deployment e un livello applicativo controllato attorno al modello.","\u002Fuploads\u002F2026\u002F06\u002Follama-is-not-the-product-building-production-ready-open-llm-applications-1782679361640-h0usqf.webp","2026-06-28T16:39:00.000Z",{"id":1876,"slug":1877,"title":1878,"excerpt":1879,"featuredImage":1880,"publishedAt":1881},"364","tipps-fuer-die-verbesserung-der-seo-suchmaschinenoptimierung","Padroneggiare il Flusso di Lavoro SEO: Strategie di Ottimizzazione Essenziali per la Crescita Organica","Un flusso di lavoro SEO strutturato è fondamentale per una crescita organica sostenibile. Scopri le dieci strategie fondamentali, dalla ricerca di parole chiave e dall'ottimizzazione tecnica alla qualità dei contenuti e all'analisi delle prestazioni.","\u002Fuploads\u002F2026\u002F03\u002Ftipps-fuer-die-verbesserung-der-seo-suchmaschinenoptimierung-1774866098131-hwkzrg.webp","2024-01-26T06:35:00.000Z","fallback",[],[]]