{"id":5645,"date":"2025-12-26T17:04:33","date_gmt":"2025-12-26T14:04:33","guid":{"rendered":"https:\/\/datakapital.com\/blog\/?p=5645"},"modified":"2025-12-27T22:37:10","modified_gmt":"2025-12-27T19:37:10","slug":"adaptif-akilli-ogretimde-rag-mcp-ve-acp-mimarisi","status":"publish","type":"post","link":"https:\/\/datakapital.com\/blog\/adaptif-akilli-ogretimde-rag-mcp-ve-acp-mimarisi\/","title":{"rendered":"Adaptif Ak\u0131ll\u0131 \u00d6\u011fretimde RAG, MCP ve ACP Mimarisi"},"content":{"rendered":"<p><em><strong>Makale \u0130ncelemesi: \u201cLeveraging RAG with ACP &amp; MCP for Adaptive Intelligent Tutoring\u201d (Horia Alexandru Modran, 2025) &#8211; Adaptif Ak\u0131ll\u0131 \u00d6\u011fretimde RAG, MCP ve ACP Mimarisi<\/strong><\/em><\/p>\n<p>Horia Alexandru Modran&#8217;\u0131n bu \u00e7al\u0131\u015fmas\u0131, ak\u0131ll\u0131 \u00f6\u011fretim sistemlerinde denetlenebilirlik ve ba\u011flam y\u00f6netimi gibi kritik ve s\u00fcregelen sorunlara, protokol tabanl\u0131 bir mimari \u00f6nererek m\u00fcdahale etmektedir. Y\u00fcksek \u00f6\u011frenimde kullan\u0131lan mevcut LLM tabanl\u0131 \u00f6\u011fretim sistemleri, do\u011fruluk ve g\u00fcvenilirlik a\u00e7\u0131s\u0131ndan ciddi zorluklarla kar\u015f\u0131 kar\u015f\u0131yad\u0131r. Makalenin de hakl\u0131 olarak i\u015faret etti\u011fi gibi, bu sistemlerin en temel sorunlar\u0131; hal\u00fcsinasyon, \u00e7ok turlu diyaloglarda ba\u011flam\u0131 koruyamama ve \u00fcretilen \u00e7\u0131kt\u0131lar\u0131n denetlenebilirli\u011finin zay\u0131f olmas\u0131d\u0131r. Yazar, bu temel sorunlara \u00e7\u00f6z\u00fcm olarak, standart bir RAG mimarisini, Agent Communication Protocol (ACP) ve Model Context Protocol (MCP) ad\u0131 verilen iki yeni protokolle entegre etmeyi \u00f6nermektedir. Temel hedef, sistem bile\u015fenleri aras\u0131nda standartla\u015fm\u0131\u015f bir ileti\u015fim ve ba\u011flam y\u00f6netimi katman\u0131 olu\u015fturarak birlikte \u00e7al\u0131\u015fabilir ve pedagojik olarak uyumlu bir ak\u0131ll\u0131 \u00f6\u011fretim sistemi in\u015fa etmektir..<\/p>\n<p>Makale, ACP\u2013MCP\u2013RAG ad\u0131n\u0131 verdi\u011fi hibrit bir sistem mimarisi sunmaktad\u0131r. Bu mimarinin temel bile\u015fenleri ve i\u015fleyi\u015f mant\u0131\u011f\u0131 \u015fu \u015fekildedir:<\/p>\n<ul>\n<li>RouterAgent: Kullan\u0131c\u0131 sorgular\u0131n\u0131 alarak ilgili dersin metin veya multimodal i\u00e7eriklerine sahip ajan\u0131na y\u00f6nlendirir.<\/li>\n<li>MCP Sunucusu: Oturum, g\u00f6rev ve ders seviyesindeki ba\u011flam\u0131 (\u00f6\u011frenme hedefleri, \u00f6\u011frencinin \u00f6nceki hatalar\u0131, pedagojik k\u0131s\u0131tlamalar vb.) merkezi olarak saklar ve y\u00f6netir.<\/li>\n<li>E\u011fitim \u00d6\u011fretim Ajan\u0131: Pedagojik politikalar\u0131 (\u00f6rne\u011fin, do\u011frudan cevap vermek yerine ipucu verme stratejisi) uygulayarak son kullan\u0131c\u0131ya sunulacak yan\u0131t\u0131 denetler.<\/li>\n<\/ul>\n<p>Sistem, kullan\u0131c\u0131 sorgusunu ald\u0131ktan sonra ilk olarak vekt\u00f6r veritaban\u0131ndan anlamsal olarak ili\u015fkili aday belgeleri \u00e7eker. Ard\u0131ndan, mimarinin ana teknik mekanizmas\u0131 olan MCP destekli ba\u011flam f\u00fczyonu ve re-ranking ad\u0131m\u0131n\u0131 uygular. Bu ad\u0131m, aday belgeleri yaln\u0131zca anlamsal benzerli\u011fe g\u00f6re de\u011fil, ayn\u0131 zamanda MCP&#8217;den gelen pedagojik hedefler, \u00f6\u011frencinin ge\u00e7mi\u015f hatalar\u0131 ve e\u011fitmen etiketleri gibi ba\u011flamsal sinyallere g\u00f6re yeniden puanlayarak, pedagojik olarak en uygun bilginin LLM&#8217;e sunulmas\u0131n\u0131 ama\u00e7lamaktad\u0131r. En y\u00fcksek puan\u0131 alan belgeler, pedagojik talimatlarla birle\u015ftirilerek nihai yan\u0131t\u0131 \u00fcretmesi i\u00e7in LLM&#8217;e g\u00f6nderilir.<\/p>\n<p><img fetchpriority=\"high\" decoding=\"async\" class=\"alignnone size-full wp-image-5646\" src=\"https:\/\/datakapital.com\/blog\/wp-content\/uploads\/2025\/12\/Adaptif-Akilli-Ogretimde-RAG-MCP-ve-ACP-Mimarisi.jpg\" alt=\"RAG Mimarisi\" width=\"611\" height=\"215\" title=\"\" srcset=\"https:\/\/datakapital.com\/blog\/wp-content\/uploads\/2025\/12\/Adaptif-Akilli-Ogretimde-RAG-MCP-ve-ACP-Mimarisi.jpg 611w, https:\/\/datakapital.com\/blog\/wp-content\/uploads\/2025\/12\/Adaptif-Akilli-Ogretimde-RAG-MCP-ve-ACP-Mimarisi-300x106.jpg 300w, https:\/\/datakapital.com\/blog\/wp-content\/uploads\/2025\/12\/Adaptif-Akilli-Ogretimde-RAG-MCP-ve-ACP-Mimarisi-150x53.jpg 150w, https:\/\/datakapital.com\/blog\/wp-content\/uploads\/2025\/12\/Adaptif-Akilli-Ogretimde-RAG-MCP-ve-ACP-Mimarisi-450x158.jpg 450w\" sizes=\"(max-width: 611px) 100vw, 611px\" \/><\/p>\n<p>Yazar, \u00f6nerdi\u011fi mimarinin etkinli\u011fini basit bir RAG sistemi ile kar\u015f\u0131la\u015ft\u0131ran nicel sonu\u00e7lar sunmaktad\u0131r. Makalede \u00f6ne \u00e7\u0131kan temel bulgular \u015funlard\u0131r:<\/p>\n<ul>\n<li>Eri\u015fim Performans\u0131: En do\u011fru belgenin ilk s\u0131rada getirilme olas\u0131l\u0131\u011f\u0131n\u0131 \u00f6l\u00e7en Recall@1 metri\u011fi, temel modeldeki 0.42 seviyesinden \u00f6nerilen y\u00f6ntemle 0.64&#8217;e y\u00fckselmi\u015ftir. MRR metri\u011fi ise 0.52&#8217;den 0.73&#8217;e \u00e7\u0131km\u0131\u015ft\u0131r.<\/li>\n<li>\u00dcretim Kalitesi: \u00dcretilen yan\u0131tlardaki kaynaklara ba\u011fl\u0131l\u0131k oran\u0131 (citation fidelity) 0.68&#8217;den 0.86&#8217;ya \u00e7\u0131km\u0131\u015ft\u0131r. Ayn\u0131 zamanda, kaynaklarla desteklenmeyen say\u0131sal iddialar\u0131n oran\u0131 %12&#8217;den %3&#8217;e d\u00fc\u015f\u00fcr\u00fclm\u00fc\u015ft\u00fcr.<\/li>\n<li>\u0130nsan De\u011ferlendirmesi: 36 \u00f6\u011frenci ve 12 e\u011fitmenin kat\u0131ld\u0131\u011f\u0131 de\u011ferlendirmede, sistemin \u00fcretti\u011fi yan\u0131tlar\u0131n olgusal do\u011frulu\u011fu (factuality) 5 \u00fczerinden 4.4 puan, pedagojik uygunlu\u011fu ise 4.5 puan alm\u0131\u015ft\u0131r.<\/li>\n<\/ul>\n<p>Makalede genel olarak ba\u015far\u0131l\u0131 sonu\u00e7lar elde edilmi\u015f olsa da baz\u0131 noktalar soru i\u015fareti bar\u0131nd\u0131rmaktad\u0131r. Bunlardan ilki re-ranking form\u00fcl\u00fcd\u00fcr. Makalenin teknik kalbi olan yeniden s\u0131ralama form\u00fcl\u00fc [1] modern bilgi eri\u015fim sistemlerinde kullan\u0131lan learning-to-rank gibi yakla\u015f\u0131mlar\u0131n aksine, manuel olarak ayarlanm\u0131\u015f statik katsay\u0131lara dayanan ilkel ve keyfi bir y\u00f6ntemdir. Yazarlar, bu katsay\u0131lar\u0131n se\u00e7imine veya farkl\u0131 ders materyalleri i\u00e7in nas\u0131l optimize edilece\u011fine dair matematiksel veya ampirik dayanak sunmamaktad\u0131r. Bu durum, y\u00f6ntemin temelindeki mekanizmay\u0131 bilimsel bir temelden \u00e7ok, sezgisel bir kural setine indirgemekte ve metodolojik bir zay\u0131fl\u0131k te\u015fkil etmektedir.<\/p>\n<p><img decoding=\"async\" class=\"alignnone size-full wp-image-5647\" src=\"https:\/\/datakapital.com\/blog\/wp-content\/uploads\/2025\/12\/Rag-Mimarisi.jpg\" alt=\"RAG Form\u00fcl\u00fc\" width=\"526\" height=\"57\" title=\"\" srcset=\"https:\/\/datakapital.com\/blog\/wp-content\/uploads\/2025\/12\/Rag-Mimarisi.jpg 526w, https:\/\/datakapital.com\/blog\/wp-content\/uploads\/2025\/12\/Rag-Mimarisi-300x33.jpg 300w, https:\/\/datakapital.com\/blog\/wp-content\/uploads\/2025\/12\/Rag-Mimarisi-150x16.jpg 150w, https:\/\/datakapital.com\/blog\/wp-content\/uploads\/2025\/12\/Rag-Mimarisi-450x49.jpg 450w\" sizes=\"(max-width: 526px) 100vw, 526px\" \/><\/p>\n<p>Bir ba\u015fka zay\u0131f nokta ise \u00f6rneklem uzay\u0131d\u0131r. \u00c7al\u0131\u015fma, &#8220;y\u00fcksek \u00f6\u011frenim i\u00e7in \u00f6l\u00e7eklenebilir&#8221; bir mimari \u00f6nerdi\u011fini iddia etse de, t\u00fcm deneyler yaln\u0131zca iki ders ve olduk\u00e7a s\u0131n\u0131rl\u0131 say\u0131da kat\u0131l\u0131mc\u0131 (36 \u00f6\u011frenci, 12 e\u011fitmen) ile y\u00fcr\u00fct\u00fclm\u00fc\u015ft\u00fcr. \u00d6nerilen mimarinin, ger\u00e7ek bir \u00fcniversite ortam\u0131n\u0131n i\u00e7erdi\u011fi binlerce da\u011f\u0131n\u0131k belge, farkl\u0131 formatlar ve karma\u015f\u0131k veri yap\u0131lar\u0131 kar\u015f\u0131s\u0131ndaki genellenebilirli\u011fi ve dayan\u0131kl\u0131l\u0131\u011f\u0131 bu \u00f6l\u00e7ekte kan\u0131tlanamaz. Bununla birlikte makale, sistemin multimodal \u00e7oklu ortam yeteneklerini bir yenilik olarak sunmaktad\u0131r. Ancak teknik detaylar incelendi\u011finde, sistemin diyagramlar\u0131 veya \u015femalar\u0131 anlamsal olarak i\u015flemedi\u011fi, bunun yerine b\u00fcy\u00fck \u00f6l\u00e7\u00fcde metne indirgeme y\u00f6ntemlerine dayand\u0131\u011f\u0131 g\u00f6r\u00fclmektedir. Bu yakla\u015f\u0131m, karma\u015f\u0131k m\u00fchendislik \u015femalar\u0131nda veya el yaz\u0131s\u0131 notlarda s\u0131k\u00e7a kar\u015f\u0131la\u015f\u0131lan OCR hatalar\u0131na kar\u015f\u0131 sistemi son derece k\u0131r\u0131lgan hale getirmekte ve multimodalite iddias\u0131n\u0131 zay\u0131flatmaktad\u0131r.<\/p>\n<p>T\u00fcm metodolojik ele\u015ftirilere ra\u011fmen, makalenin kavramsal d\u00fczeydeki katk\u0131s\u0131 g\u00f6z ard\u0131 edilemez. Ajanlar aras\u0131 ileti\u015fim ve ba\u011flam y\u00f6netimini standart protokollerle tan\u0131mlama fikri, ak\u0131ll\u0131 sistemlerin en b\u00fcy\u00fck sorunlar\u0131ndan olan \u015feffaf ve denetlenebilir olma problemlerine y\u00f6nelik \u00f6zg\u00fcn ve de\u011ferli bir yakla\u015f\u0131md\u0131r. Sistemin neden belirli bir yan\u0131t \u00fcretti\u011finin izlenebilir hale gelmesi, \u00f6zellikle e\u011fitim gibi hassas bir alanda kritik \u00f6neme sahiptir. Yazarlar, \u00f6nerdikleri mimarinin etkinli\u011fini, end\u00fcstri standard\u0131 kabul edilen HyDE veya Cohere Rerank gibi geli\u015fmi\u015f y\u00f6ntemler yerine, yaln\u0131zca kosin\u00fcs benzerli\u011fine dayanan ve kas\u0131tl\u0131 olarak zay\u0131f se\u00e7ilmi\u015f bir \u2018saf RAG\u2019 ile k\u0131yaslam\u0131\u015ft\u0131r. Bu metodolojik tercih, sunulan performans art\u0131\u015f\u0131n\u0131n b\u00fcy\u00fckl\u00fc\u011f\u00fcn\u00fc yapay olarak \u015fi\u015firme riski ta\u015f\u0131maktad\u0131r.<\/p>\n<p>Makalenin arg\u00fcmanlar\u0131, sorgulanmadan do\u011fru kabul edilen birka\u00e7 temel varsay\u0131ma dayanmaktad\u0131r. Bunlardan ilki e\u011fitmen etiketlemesine dayal\u0131 bir varsay\u0131md\u0131r. Sistemin ba\u015far\u0131s\u0131, b\u00fcy\u00fck \u00f6l\u00e7\u00fcde e\u011fitmenlerin ders materyallerini manuel olarak etiketlemesine ve pedagojik kurallar\u0131 tan\u0131mlamas\u0131na ba\u011fl\u0131d\u0131r. Ancak makale, akademisyenlerin bu yo\u011fun ve zaman al\u0131c\u0131 i\u015f y\u00fck\u00fcn\u00fc \u00fcstlenmek i\u00e7in gerekli zaman ve motivasyona sahip olup olmad\u0131\u011f\u0131n\u0131 tamamen g\u00f6z ard\u0131 etmektedir. Bu pratik maliyet, sistemin ger\u00e7ek d\u00fcnyada benimsenmesinin \u00f6n\u00fcndeki en b\u00fcy\u00fck engeldir. Bununla birlikte bu manuel e\u011fitmen etiketlemesi darbo\u011faz\u0131, finansal alanda do\u011fal bir avantaja d\u00f6n\u00fc\u015fmektedir. E\u011fitim materyallerinin aksine, finansal raporlar uluslararas\u0131 standartlarda ve y\u00fcksek oranda yap\u0131sal formatlarda \u00fcretilmektedir. Bu durum, makalede \u00f6nerilen MCP mimarisinin ihtiya\u00e7 duydu\u011fu metadata\u2019lar\u0131n, manuel insan eme\u011fine gerek kalmadan otomatik olarak \u00e7\u0131kar\u0131labilmesine olanak tan\u0131r. Dolay\u0131s\u0131yla bu mimari, finansal konseptlerde e\u011fitime k\u0131yasla \u00e7ok daha h\u0131zl\u0131 \u00f6l\u00e7eklenebilir.<\/p>\n<p>Makale, hal\u00fcsinasyon oran\u0131ndaki d\u00fc\u015f\u00fc\u015f\u00fc kendi mimarisine ba\u011flamaktad\u0131r. Ancak deneylerde kullan\u0131lan temel modelin <a href=\"https:\/\/chatlyai.app\/\" target=\"_blank\" rel=\"noopener\">(Claude Sonnet 4.5)<\/a> zaten bu konuda g\u00fc\u00e7l\u00fc oldu\u011fu bilinmektedir. Azalman\u0131n \u00f6nerilen mimariden mi, yoksa yaln\u0131zca g\u00fc\u00e7l\u00fc temel model ve etkili istem m\u00fchendisli\u011finden mi kaynakland\u0131\u011f\u0131 belirsiz kalmaktad\u0131r.<\/p>\n<p>Geli\u015ftirilebilecek noktalara de\u011finmek gerekirse;<a href=\"https:\/\/datakapital.com\/blog\/lasso-regresyonu-ridge-regresyonu-ve-elastic-net\/\"> mevcut do\u011frusal fonksiyonun<\/a> yerini alacak ve ba\u011flam-kan\u0131t etkile\u015fimlerini daha zengin bir \u015fekilde modelleyecek, veriden \u00f6\u011frenen bir re-ranker geli\u015ftirilmelidir. Mimarinin etkinli\u011fini ve \u00f6l\u00e7eklenebilirli\u011fini do\u011frulamak i\u00e7in, \u00e7ok daha geni\u015f ve \u00e7e\u015fitli veri setleri \u00fczerinde uzun d\u00f6nemli \u00e7al\u0131\u015fmalar y\u00fcr\u00fct\u00fclmelidir. \u00d6nerilen protokollerin getirdi\u011fi m\u00fchendislik y\u00fck\u00fcn\u00fc, <a href=\"https:\/\/datakapital.com\/blog\/chatgptnin-sentiment-analizde-kullanimi\/\">long-context LLM&#8217;ler<\/a> ile olu\u015fturulan daha basit RAG yakla\u015f\u0131mlar\u0131na kar\u015f\u0131 etkinlik ve verimlilik a\u00e7\u0131s\u0131ndan de\u011ferlendiren k\u0131yaslamal\u0131 bir analiz ger\u00e7ekle\u015ftirilmelidir. MCP&#8217;nin, oturumlar aras\u0131 kal\u0131c\u0131 ki\u015fiselle\u015ftirme sa\u011flamak amac\u0131yla zamansal \u00f6\u011frenci modellerini ve diyalog ge\u00e7mi\u015flerini temsil edecek \u015fekilde geni\u015fletilmesi gerekmektedir. Manuel etiketleme y\u00fck\u00fcn\u00fc azaltmak ve yakla\u015f\u0131m\u0131 \u00f6l\u00e7eklendirmek i\u00e7in, e\u011fitmen etiketlemesinin yar\u0131 denetimli veya aktif \u00f6\u011frenme y\u00f6ntemleriyle otomatikle\u015ftirilmesi \u00fczerine \u00e7al\u0131\u015fmalar yap\u0131lmal\u0131d\u0131r.<\/p>\n<p>Sonu\u00e7 olarak makale, ak\u0131ll\u0131 \u00f6\u011fretim sistemlerinde denetlenebilirlik ve \u015feffafl\u0131k i\u00e7in protokollerin \u00f6nemini vurgulayarak literat\u00fcre de\u011ferli bir kavramsal \u00e7er\u00e7eve sunmaktad\u0131r. ACP ve MCP&#8217;nin standartla\u015ft\u0131r\u0131lm\u0131\u015f bir katman olarak \u00f6nerilmesi, gelecekteki mod\u00fcler ve g\u00fcvenilir sistemler i\u00e7in bir vizyon ortaya koymaktad\u0131r. Yazar\u0131n \u00f6nerdi\u011fi ACP ve MCP protokolleri, e\u011fitimde pedagojik uygunluk i\u00e7in kullan\u0131l\u0131rken, finansal uygulamalarda reg\u00fclasyon uyumlulu\u011fu i\u00e7in kritik bir altyap\u0131 sunmaktad\u0131r. Finansal dan\u0131\u015fmanl\u0131kta bir yapay zeka modelinin neden belirli bir yat\u0131r\u0131m tavsiyesi veya analizi \u00fcretti\u011finin saniye saniye izlenebilir olmas\u0131 k\u0131ymetlidir. Bu \u00e7al\u0131\u015fmadaki RouterAgent ve MCP sunucusu yap\u0131s\u0131, finansal tavsiyelerin kullan\u0131c\u0131n\u0131n risk profili ile uyumlu olup olmad\u0131\u011f\u0131n\u0131 denetleyen otomatik bir compliance ajan\u0131 olarak yeniden kurgulanmaya son derece m\u00fcsaittir.<\/p>\n","protected":false},"excerpt":{"rendered":"<p>Makale \u0130ncelemesi: \u201cLeveraging RAG with ACP &amp; MCP for Adaptive Intelligent Tutoring\u201d (Horia Alexandru Modran, 2025) &#8211; Adaptif Ak\u0131ll\u0131 \u00d6\u011fretimde RAG, MCP ve ACP Mimarisi Horia Alexandru Modran&#8217;\u0131n bu \u00e7al\u0131\u015fmas\u0131, ak\u0131ll\u0131 \u00f6\u011fretim sistemlerinde denetlenebilirlik ve ba\u011flam y\u00f6netimi gibi kritik ve s\u00fcregelen sorunlara, protokol tabanl\u0131 bir mimari \u00f6nererek m\u00fcdahale etmektedir. Y\u00fcksek \u00f6\u011frenimde kullan\u0131lan mevcut LLM tabanl\u0131<\/p>\n","protected":false},"author":21,"featured_media":5646,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[38,15,520],"tags":[548,549,547,546],"class_list":{"0":"post-5645","1":"post","2":"type-post","3":"status-publish","4":"format-standard","5":"has-post-thumbnail","7":"category-python-ile-veri-isleme","8":"category-sentiment-analizi","9":"category-yapay-zeka","10":"tag-acp","11":"tag-akilli-ogretim-sistemleri","12":"tag-mcp","13":"tag-rag"},"better_featured_image":{"id":5646,"alt_text":"RAG Mimarisi","caption":"","description":"","media_type":"image","media_details":{"width":611,"height":215,"file":"2025\/12\/Adaptif-Akilli-Ogretimde-RAG-MCP-ve-ACP-Mimarisi.jpg","filesize":34075,"sizes":{"medium":{"file":"Adaptif-Akilli-Ogretimde-RAG-MCP-ve-ACP-Mimarisi-300x106.jpg","width":300,"height":106,"mime-type":"image\/jpeg","filesize":5949,"source_url":"https:\/\/datakapital.com\/blog\/wp-content\/uploads\/2025\/12\/Adaptif-Akilli-Ogretimde-RAG-MCP-ve-ACP-Mimarisi-300x106.jpg"},"thumbnail":{"file":"Adaptif-Akilli-Ogretimde-RAG-MCP-ve-ACP-Mimarisi-150x150.jpg","width":150,"height":150,"mime-type":"image\/jpeg","filesize":5028,"source_url":"https:\/\/datakapital.com\/blog\/wp-content\/uploads\/2025\/12\/Adaptif-Akilli-Ogretimde-RAG-MCP-ve-ACP-Mimarisi-150x150.jpg"},"bunyad-small":{"file":"Adaptif-Akilli-Ogretimde-RAG-MCP-ve-ACP-Mimarisi-150x53.jpg","width":150,"height":53,"mime-type":"image\/jpeg","filesize":2113,"source_url":"https:\/\/datakapital.com\/blog\/wp-content\/uploads\/2025\/12\/Adaptif-Akilli-Ogretimde-RAG-MCP-ve-ACP-Mimarisi-150x53.jpg"},"bunyad-medium":{"file":"Adaptif-Akilli-Ogretimde-RAG-MCP-ve-ACP-Mimarisi-450x158.jpg","width":450,"height":158,"mime-type":"image\/jpeg","filesize":11585,"source_url":"https:\/\/datakapital.com\/blog\/wp-content\/uploads\/2025\/12\/Adaptif-Akilli-Ogretimde-RAG-MCP-ve-ACP-Mimarisi-450x158.jpg"}},"image_meta":{"aperture":"0","credit":"","camera":"","caption":"","created_timestamp":"0","copyright":"","focal_length":"0","iso":"0","shutter_speed":"0","title":"","orientation":"0","keywords":[]}},"post":5645,"source_url":"https:\/\/datakapital.com\/blog\/wp-content\/uploads\/2025\/12\/Adaptif-Akilli-Ogretimde-RAG-MCP-ve-ACP-Mimarisi.jpg"},"amp_enabled":true,"_links":{"self":[{"href":"https:\/\/datakapital.com\/blog\/wp-json\/wp\/v2\/posts\/5645","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/datakapital.com\/blog\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/datakapital.com\/blog\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/datakapital.com\/blog\/wp-json\/wp\/v2\/users\/21"}],"replies":[{"embeddable":true,"href":"https:\/\/datakapital.com\/blog\/wp-json\/wp\/v2\/comments?post=5645"}],"version-history":[{"count":3,"href":"https:\/\/datakapital.com\/blog\/wp-json\/wp\/v2\/posts\/5645\/revisions"}],"predecessor-version":[{"id":5651,"href":"https:\/\/datakapital.com\/blog\/wp-json\/wp\/v2\/posts\/5645\/revisions\/5651"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/datakapital.com\/blog\/wp-json\/wp\/v2\/media\/5646"}],"wp:attachment":[{"href":"https:\/\/datakapital.com\/blog\/wp-json\/wp\/v2\/media?parent=5645"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/datakapital.com\/blog\/wp-json\/wp\/v2\/categories?post=5645"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/datakapital.com\/blog\/wp-json\/wp\/v2\/tags?post=5645"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}