{"id":373,"date":"2025-02-04T23:01:00","date_gmt":"2025-02-04T23:01:00","guid":{"rendered":"http:\/\/blog.firatyasar.com\/?p=373"},"modified":"2025-03-29T23:05:32","modified_gmt":"2025-03-29T23:05:32","slug":"agentic-ai-extensions-function-ve-data-store-kullanim-senaryolari","status":"publish","type":"post","link":"https:\/\/blog.firatyasar.com\/?p=373","title":{"rendered":"Agentic AI &#8211; Extensions, Function ve Data Store Kullan\u0131m Senaryolar\u0131"},"content":{"rendered":"\n<p>Bir agent\u2019\u0131n d\u0131\u015f d\u00fcnya ile etkili bir \u015fekilde etkile\u015fim kurabilmesi i\u00e7in <strong>do\u011fru ara\u00e7lar\u0131 (tools)<\/strong> kullanmas\u0131 gerekir. Bu ara\u00e7lar, sadece d\u0131\u015f sistemlere eri\u015fimi de\u011fil, ayn\u0131 zamanda modelin s\u0131n\u0131rl\u0131 e\u011fitim bilgisini geni\u015fletmesini ve g\u00f6revlerini ba\u015far\u0131yla yerine getirmesini sa\u011flar. Agent mimarisinde en s\u0131k kullan\u0131lan \u00fc\u00e7 ara\u00e7 tipi \u015funlard\u0131r:<\/p>\n\n\n\n<ul><li><strong>Extensions<\/strong><\/li><li><strong>Function Calling<\/strong><\/li><li><strong>Data Stores<\/strong><\/li><\/ul>\n\n\n\n<p>Bu ara\u00e7lar\u0131n hangi senaryoda ne \u015fekilde tercih edilece\u011fini anlamak i\u00e7in a\u015fa\u011f\u0131daki tabloyu inceleyebiliriz:<\/p>\n\n\n\n<figure class=\"wp-block-table\"><table><thead><tr><th>\u00d6zellik \/ Kullan\u0131m<\/th><th><strong>Extensions<\/strong><\/th><th><strong>Function Calling<\/strong><\/th><th><strong>Data Stores<\/strong><\/th><\/tr><\/thead><tbody><tr><td><strong>\u00c7al\u0131\u015fma Yeri<\/strong><\/td><td>Agent-Side Execution<\/td><td>Client-Side Execution<\/td><td>Agent-Side Execution<\/td><\/tr><tr><td><strong>Kullan\u0131m Senaryosu<\/strong><\/td><td>&#8211; Geli\u015ftirici, agent\u2019\u0131n API endpoint&#8217;leriyle do\u011frudan etkile\u015fime ge\u00e7mesini ister. &#8211; Haz\u0131r Extension\u2019lar kullan\u0131ld\u0131\u011f\u0131nda faydal\u0131d\u0131r (\u00f6rne\u011fin: Vertex Search, Code Interpreter). &#8211; \u00c7ok ad\u0131ml\u0131 planlama ve API \u00e7a\u011fr\u0131lar\u0131 (\u00f6r. bir \u00f6nceki ad\u0131m\u0131n \u00e7\u0131kt\u0131s\u0131na g\u00f6re sonraki API se\u00e7imi).<\/td><td>&#8211; G\u00fcvenlik veya kimlik do\u011frulama k\u0131s\u0131tlar\u0131 nedeniyle agent API&#8217;ye do\u011frudan eri\u015femez. &#8211; Ger\u00e7ek zamanl\u0131 \u00e7a\u011fr\u0131lar\u0131n m\u00fcmk\u00fcn olmad\u0131\u011f\u0131 zamanlama k\u0131s\u0131tlar\u0131 vard\u0131r (\u00f6r. batch i\u015flemler, human-in-the-loop senaryolar\u0131). &#8211; API internet \u00fczerinden eri\u015filebilir de\u011filse ya da Google sistemleri taraf\u0131ndan eri\u015filemiyorsa.<\/td><td>&#8211; Geli\u015ftirici, RAG (Retrieval Augmented Generation) yakla\u015f\u0131m\u0131n\u0131 uygulamak ister. &#8211; A\u015fa\u011f\u0131daki veri t\u00fcrleriyle \u00e7al\u0131\u015fmak i\u00e7in uygundur: \u2022 Web sitesi i\u00e7erikleri \u2022 PDF, Word, Excel, CSV gibi yap\u0131land\u0131r\u0131lm\u0131\u015f veriler \u2022 HTML, TXT gibi yap\u0131land\u0131r\u0131lmam\u0131\u015f metinler \u2022 \u0130li\u015fkisel\/veritaban\u0131 sistemleri<\/td><\/tr><\/tbody><\/table><\/figure>\n\n\n\n<h2>Agent\u2019lar\u0131n Gelece\u011fi: Karma\u015f\u0131k G\u00f6revler, Zincirleme Kararlar<\/h2>\n\n\n\n<p>G\u00fcn\u00fcm\u00fczde agent\u2019lar, dil modellerinin s\u0131n\u0131rlar\u0131n\u0131 a\u015farak <strong>ger\u00e7ek zamanl\u0131 bilgiye eri\u015febilir<\/strong>, <strong>ger\u00e7ek d\u00fcnyada eylemler \u00f6nerebilir<\/strong>, hatta <strong>karma\u015f\u0131k g\u00f6revleri planlay\u0131p otomatik olarak y\u00fcr\u00fctebilir<\/strong> hale gelmi\u015ftir. Bu yaln\u0131zca bir ba\u015flang\u0131\u00e7t\u0131r; gelecekte bu sistemler \u00e7ok daha geli\u015fmi\u015f yeteneklerle donat\u0131lacakt\u0131r.<\/p>\n\n\n\n<p>\u00d6zellikle dikkat \u00e7eken bir yakla\u015f\u0131m olan <strong>\u201cagent chaining\u201d<\/strong>, birden fazla dil modelinin veya agent\u2019\u0131n birbiriyle koordineli \u015fekilde \u00e7al\u0131\u015fmas\u0131n\u0131 sa\u011flayarak, g\u00f6revler aras\u0131nda ge\u00e7i\u015f yapmalar\u0131na ve g\u00f6revleri par\u00e7alayarak y\u00f6netmelerine olanak tan\u0131r.<\/p>\n\n\n\n<h2>Orkestrasyon Katman\u0131:<\/h2>\n\n\n\n<p>Bir agent\u2019\u0131n karar alma ve hareket etme s\u00fcre\u00e7lerinin merkezinde, <strong>orchestration layer<\/strong> yer al\u0131r. Bu katman; ak\u0131l y\u00fcr\u00fctme, planlama, karar verme ve hareket etme s\u00fcre\u00e7lerini yap\u0131land\u0131ran bir t\u00fcr <strong>bili\u015fsel mimaridir<\/strong>.<\/p>\n\n\n\n<p>Bu yap\u0131y\u0131 destekleyen en \u00f6nemli reasoning framework\u2019leri \u015funlard\u0131r:<\/p>\n\n\n\n<ul><li><strong>ReAct<\/strong>: Ak\u0131l y\u00fcr\u00fctme + eylem \u00fcretme<\/li><li><strong>Chain-of-Thought (CoT)<\/strong>: Ad\u0131m ad\u0131m d\u00fc\u015f\u00fcnce zinciri<\/li><li><strong>Tree-of-Thoughts (ToT)<\/strong>: Paralel d\u00fc\u015f\u00fcnce dallar\u0131 ile problem \u00e7\u00f6zme<\/li><\/ul>\n\n\n\n<p>Orchestration layer, bu framework\u2019ler yard\u0131m\u0131yla verileri i\u015fler, kararlar al\u0131r ve agent\u2019\u0131n aksiyonlar\u0131n\u0131 y\u00f6nlendirir.<\/p>\n\n\n\n<h2>Tools: D\u0131\u015f D\u00fcnyaya A\u00e7\u0131lan Kap\u0131lar<\/h2>\n\n\n\n<p>Agent\u2019lar i\u00e7in ara\u00e7lar (tools), d\u0131\u015f d\u00fcnyayla ba\u011flant\u0131 kurmalar\u0131n\u0131 sa\u011flayan <strong>anahtar bile\u015fenlerdir<\/strong>. Her bir ara\u00e7 farkl\u0131 g\u00f6revleri \u00fcstlenir:<\/p>\n\n\n\n<ul><li><strong>Extensions<\/strong>: Agent ile harici API\u2019ler aras\u0131nda k\u00f6pr\u00fc kurar, do\u011frudan API \u00e7a\u011fr\u0131lar\u0131 ve ger\u00e7ek zamanl\u0131 veri alma i\u015flevi sunar.<\/li><li><strong>Function Calling<\/strong>: Fonksiyon parametrelerini model \u00fcretir, ancak \u00e7al\u0131\u015ft\u0131rma i\u015flemi istemci taraf\u0131nda yap\u0131l\u0131r. Geli\u015ftiriciye esneklik ve kontrol sa\u011flar.<\/li><li><strong>Data Stores<\/strong>: Yap\u0131land\u0131r\u0131lm\u0131\u015f veya yap\u0131land\u0131r\u0131lmam\u0131\u015f veri kaynaklar\u0131na eri\u015fim sa\u011flar. Agent\u2019\u0131n bilgi taban\u0131n\u0131 geni\u015fleterek RAG tabanl\u0131 uygulamalarda kritik rol oynar.<\/li><\/ul>\n\n\n\n<h2>Sonu\u00e7<\/h2>\n\n\n\n<p>Extensions, Function Calling ve Data Stores, her biri farkl\u0131 ama\u00e7lara hizmet eden ve agent\u2019lar\u0131n yeteneklerini geni\u015fleten ara\u00e7lard\u0131r. Do\u011fru senaryoda do\u011fru arac\u0131 se\u00e7mek, bir agent\u2019\u0131n performans\u0131 ile ba\u015far\u0131s\u0131 aras\u0131ndaki fark\u0131 belirleyebilir. Gelece\u011fin agent sistemleri; daha fazla ara\u00e7la donat\u0131lm\u0131\u015f, daha derin reasoning yeteneklerine sahip ve \u00e7ok daha karma\u015f\u0131k g\u00f6revleri otonom olarak yerine getirebilen sistemler olacak.<\/p>\n","protected":false},"excerpt":{"rendered":"<p>Bir agent\u2019\u0131n d\u0131\u015f d\u00fcnya ile etkili bir \u015fekilde etkile\u015fim kurabilmesi i\u00e7in do\u011fru ara\u00e7lar\u0131 (tools) kullanmas\u0131 gerekir. Bu ara\u00e7lar, sadece d\u0131\u015f sistemlere eri\u015fimi de\u011fil, ayn\u0131 zamanda modelin s\u0131n\u0131rl\u0131 e\u011fitim bilgisini geni\u015fletmesini ve g\u00f6revlerini ba\u015far\u0131yla yerine getirmesini sa\u011flar. Agent mimarisinde en s\u0131k kullan\u0131lan \u00fc\u00e7 ara\u00e7 tipi \u015funlard\u0131r: Extensions Function Calling Data Stores Bu ara\u00e7lar\u0131n hangi senaryoda ne\u2026 <span class=\"read-more\"><a href=\"https:\/\/blog.firatyasar.com\/?p=373\">Read More &raquo;<\/a><\/span><\/p>\n","protected":false},"author":1,"featured_media":348,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":[],"categories":[1],"tags":[149,161,160,162,163,150,133],"_links":{"self":[{"href":"https:\/\/blog.firatyasar.com\/index.php?rest_route=\/wp\/v2\/posts\/373"}],"collection":[{"href":"https:\/\/blog.firatyasar.com\/index.php?rest_route=\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/blog.firatyasar.com\/index.php?rest_route=\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/blog.firatyasar.com\/index.php?rest_route=\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/blog.firatyasar.com\/index.php?rest_route=%2Fwp%2Fv2%2Fcomments&post=373"}],"version-history":[{"count":1,"href":"https:\/\/blog.firatyasar.com\/index.php?rest_route=\/wp\/v2\/posts\/373\/revisions"}],"predecessor-version":[{"id":374,"href":"https:\/\/blog.firatyasar.com\/index.php?rest_route=\/wp\/v2\/posts\/373\/revisions\/374"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/blog.firatyasar.com\/index.php?rest_route=\/wp\/v2\/media\/348"}],"wp:attachment":[{"href":"https:\/\/blog.firatyasar.com\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=373"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/blog.firatyasar.com\/index.php?rest_route=%2Fwp%2Fv2%2Fcategories&post=373"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/blog.firatyasar.com\/index.php?rest_route=%2Fwp%2Fv2%2Ftags&post=373"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}