{"id":43813,"date":"2026-08-03T21:04:03","date_gmt":"2026-08-03T18:04:03","guid":{"rendered":"https:\/\/fatihsoysal.com\/blog\/otonom-yapay-zeka-bir-etiket-degil-spektrum\/"},"modified":"2026-08-03T21:04:03","modified_gmt":"2026-08-03T18:04:03","slug":"otonom-yapay-zeka-bir-etiket-degil-spektrum","status":"publish","type":"post","link":"https:\/\/fatihsoysal.com\/blog\/otonom-yapay-zeka-bir-etiket-degil-spektrum\/","title":{"rendered":"Otonom Yapay Zeka: Bir Etiket De\u011fil Spektrum"},"content":{"rendered":"<h2>Otonom Yapay Zeka: Bir Etiket De\u011fil Spektrum<\/h2>\n<p>Otonom yapay zeka (Agentic AI) basit bir pazarlama etiketi de\u011fil; dinamik karar alma ve eylem yetene\u011fine dayanan geni\u015f bir geli\u015fim spektrumudur.<\/p>\n<p>Geleneksel yapay zeka sistemleri, kullan\u0131c\u0131n\u0131n verdi\u011fi girdilere tek seferlik yan\u0131tlar \u00fcreten pasif ara\u00e7lar olarak hayat\u0131m\u0131za girdi. Ancak g\u00fcn\u00fcm\u00fczde teknoloji d\u00fcnyas\u0131 b\u00fcy\u00fck bir zihniyet de\u011fi\u015fimi ya\u015f\u0131yor. Art\u0131k sadece metin \u00fcreten veya sorular\u0131 yan\u0131tlayan B\u00fcy\u00fck Dil Modelleri (Large Language Models &#8211; LLM) yerine, belirli hedefleri ger\u00e7ekle\u015ftirmek i\u00e7in kendi kendine plan yapabilen, ara\u00e7lar\u0131 kullanabilen ve hatalar\u0131ndan ders \u00e7\u0131kararak ilerleyen sistemleri konu\u015fuyoruz. \u0130\u015fte bu noktada kar\u015f\u0131m\u0131za &#8220;Agentic AI&#8221; (Otonom Yapay Zeka) kavram\u0131 \u00e7\u0131kmaktad\u0131r. Fakat san\u0131lan\u0131n aksine Agentic AI, bir yaz\u0131l\u0131m\u0131n ya sahip oldu\u011fu ya da olmad\u0131\u011f\u0131 ikili bir \u00f6zellik de\u011fildir. Aksine bu yap\u0131, s\u0131f\u0131r otonomiden tam ba\u011f\u0131ms\u0131zl\u0131\u011fa uzanan \u00e7ok katmanl\u0131 bir spektrumu temsil eder.<\/p>\n<p>Bu makalede, otonom yapay zeka sistemlerinin ne anlama geldi\u011fini, yaz\u0131l\u0131m mimarilerindeki evrimini, otonomi seviyelerini ve ger\u00e7ek d\u00fcnyadaki uygulamalar\u0131n\u0131 detayl\u0131 bir \u015fekilde inceleyece\u011fiz. S\u0131f\u0131rdan bir ajan mimarisi kurman\u0131n mant\u0131\u011f\u0131n\u0131 kavrayarak, sistemlerinizi bu spektrumda nas\u0131l do\u011fru konumland\u0131rabilece\u011finizi \u00f6\u011freneceksiniz.<\/p>\n<h2>Agentic AI Nedir ve Neden Bir Spektrum Olarak Ele Al\u0131nmal\u0131d\u0131r?<\/h2>\n<p>Geleneksel sohbet robotlar\u0131 (chatbots), do\u011frudan kullan\u0131c\u0131 y\u00f6nlendirmesine ihtiya\u00e7 duyar. \u00d6rne\u011fin, bir yaz\u0131l\u0131m geli\u015ftirici bir ChatGPT modeline <code>\"Bu Python kodundaki hatay\u0131 ay\u0131kla\"<\/code> talimat\u0131 verdi\u011finde, model sadece \u00f6n\u00fcne koyulan kodu analiz eder ve bir \u00f6neri sunar. Ancak i\u015flem orada biter. Kodu \u00e7al\u0131\u015ft\u0131rmaz, test yapmaz veya projenin di\u011fer dosyalar\u0131yla etkile\u015fime ge\u00e7mez.<\/p>\n<p>Otonom yapay zeka ise \u00e7\u0131kt\u0131 \u00fcretmek yerine <strong>hedefe ula\u015fmaya<\/strong> odaklan\u0131r. Kullan\u0131c\u0131 ajana <code>\"Sistemdeki veri taban\u0131 ba\u011flant\u0131 hatas\u0131n\u0131 bul ve d\u00fczelt\"<\/code> talimat\u0131n\u0131 verdi\u011finde, ajan \u015fu ad\u0131mlar\u0131 ba\u011f\u0131ms\u0131z olarak y\u00fcr\u00fct\u00fcr:<\/p>\n<ul>\n<li><strong>Log Dosyalar\u0131n\u0131 \u0130nceleme:<\/strong> Sunucudaki hata kay\u0131tlar\u0131n\u0131 okur ve sorunun kayna\u011f\u0131n\u0131 tespit eder.<\/li>\n<li><strong>Kod Analizi:<\/strong> \u0130lgili konfig\u00fcrasyon dosyalar\u0131n\u0131 veya veri taban\u0131 s\u00fcr\u00fcc\u00fclerini denetler.<\/li>\n<li><strong>Aksiyon Alma:<\/strong> Hatal\u0131 parametreyi g\u00fcnceller veya eksik paketi sisteme y\u00fckler.<\/li>\n<li><strong>Do\u011frulama:<\/strong> Kodun \u00e7al\u0131\u015f\u0131p \u00e7al\u0131\u015fmad\u0131\u011f\u0131n\u0131 do\u011frulamak i\u00e7in test ortam\u0131nda ilgili komutlar\u0131 ko\u015fturur.<\/li>\n<\/ul>\n<p>Bu durum, yapay zekan\u0131n sadece yan\u0131t veren bir sistemden (generative), hedefe y\u00f6nelik eylem ger\u00e7ekle\u015ftiren bir ajana (agentic) d\u00f6n\u00fc\u015fmesidir. Ancak her ajan ayn\u0131 seviyede otonomiye sahip de\u011fildir. Kimi sistemler sadece basit API (Uygulama Programlama Aray\u00fcz\u00fc) \u00e7a\u011fr\u0131lar\u0131 yapabilirken, kimileri karma\u015f\u0131k kararlar alarak insan m\u00fcdahalesine ihtiya\u00e7 duymadan s\u00fcre\u00e7leri y\u00f6netir. Dolay\u0131s\u0131yla Agentic AI kavram\u0131n\u0131 &#8220;var ya da yok&#8221; \u015feklinde de\u011fil, bir yetkinlik ve yetkilendirme \u00f6l\u00e7e\u011finde de\u011ferlendirmek gerekir.<\/p>\n<h2>Statik LLM&#8217;lerden Otonom Ajanlara Ge\u00e7i\u015f Nas\u0131l Ger\u00e7ekle\u015fir?<\/h2>\n<p>Yaz\u0131l\u0131m mimarilerinin evrimi incelendi\u011finde, basit istem m\u00fchendisli\u011finden (prompt engineering) tam otonom ajansal yap\u0131lara ge\u00e7i\u015fin belirli a\u015famalardan ge\u00e7ti\u011fi g\u00f6r\u00fcl\u00fcr. Bu d\u00f6n\u00fc\u015f\u00fcm, yaz\u0131l\u0131mlar\u0131n \u00e7evreleriyle kurdu\u011fu etkile\u015fimin derinle\u015fmesiyle do\u011frudan ili\u015fkilidir.<\/p>\n<p>S\u00fcrecin ilk a\u015famas\u0131nda salt \u00fcretken modeller yer al\u0131r. Bu modeller yaln\u0131zca e\u011fitildikleri veri k\u00fcmesiyle s\u0131n\u0131rl\u0131 yan\u0131tlarlar verirler. \u0130kinci a\u015famada hayat\u0131m\u0131za Bilgi Art\u0131r\u0131ml\u0131 \u00dcretim (Retrieval-Augmented Generation &#8211; RAG) mimarileri girmi\u015ftir. RAG sayesinde modeller, kurum i\u00e7i dok\u00fcmanlardan veya g\u00fcncel veri tabanlar\u0131ndan bilgi \u00e7ekerek daha do\u011fru yan\u0131tlar \u00fcretmeye ba\u015flam\u0131\u015ft\u0131r. Fakat bu a\u015famada da sistem d\u0131\u015f d\u00fcnyaya etki edemez; sadece bilgiyi okur ve sunar.<\/p>\n<p>\u00dc\u00e7\u00fcnc\u00fc ve en kritik k\u0131r\u0131lma noktas\u0131 ise &#8220;Fonksiyon \u00c7a\u011f\u0131rma&#8221; (Function Calling) ve ara\u00e7 kullan\u0131m\u0131d\u0131r. Yapay zeka modeli, d\u0131\u015f d\u00fcnyadaki sistemlerle ileti\u015fim kurabilece\u011fini fark etti\u011fi an ajansal nitelik kazan\u0131r. Model, bir eposta g\u00f6ndermek, SQL sorgusu \u00e7al\u0131\u015ft\u0131rmak veya bir iklim API&#8217;sinden hava durumu verisi \u00e7ekmek i\u00e7in uygun JSON format\u0131nda parametreler \u00fcretebilir. Nitekim g\u00fcn\u00fcm\u00fczdeki modern ajansal yap\u0131lar, bu yetene\u011fin \u00fczerine in\u015fa edilen bir karar alma d\u00f6ng\u00fcs\u00fc (reasoning loop) ile \u00e7al\u0131\u015f\u0131r.<\/p>\n<h2>Yapay Zeka Otonom Seviyeleri Nelerdir?<\/h2>\n<p>Agentic AI yap\u0131s\u0131n\u0131 do\u011fru kavramak i\u00e7in bu sistemleri 5 farkl\u0131 otonomi seviyesinde s\u0131n\u0131fland\u0131rmak m\u00fcmk\u00fcnd\u00fcr. Her seviye, sistemin karma\u015f\u0131kl\u0131\u011f\u0131n\u0131 ve insana olan ba\u011f\u0131ml\u0131l\u0131\u011f\u0131n\u0131 g\u00f6sterir.<\/p>\n<table>\n<thead>\n<tr>\n<th>Seviye<\/th>\n<th>Otonomi Tan\u0131m\u0131<\/th>\n<th>Karar Alma Mekanizmas\u0131<\/th>\n<th>\u0130nsan Rol\u00fc<\/th>\n<\/tr>\n<\/thead>\n<tbody>\n<tr>\n<td><strong>Seviye 0<\/strong><\/td>\n<td>Otonomi Yok (Salt LLM)<\/td>\n<td>Sadece metin\/kod \u00fcretimi yapar. Eylem alma yetene\u011fi yoktur.<\/td>\n<td>Tam kontrol ve y\u00f6nlendirme sa\u011flar.<\/td>\n<\/tr>\n<tr>\n<td><strong>Seviye 1<\/strong><\/td>\n<td>Y\u00f6nlendirmeli Zincirler (Chains)<\/td>\n<td>\u00d6nceden tan\u0131ml\u0131 ad\u0131mlar\u0131 (workflow) s\u0131rayla izler.<\/td>\n<td>Her ad\u0131m\u0131n kural\u0131n\u0131 koyar ve tetikler.<\/td>\n<\/tr>\n<tr>\n<td><strong>Seviye 2<\/strong><\/td>\n<td>Ara\u00e7 Kullanan Ajanlar<\/td>\n<td>Hangi arac\u0131 ne zaman kullanaca\u011f\u0131na model kendisi karar verir.<\/td>\n<td>Sadece nihai girdiyi verir, s\u00fcreci izler.<\/td>\n<\/tr>\n<tr>\n<td><strong>Seviye 3<\/strong><\/td>\n<td>Otonom Hedef Odakl\u0131 Ajanlar<\/td>\n<td>Hedefe ula\u015fmak i\u00e7in kendi alt g\u00f6revlerini (sub-tasks) olu\u015fturur ve dinamik plan yapar.<\/td>\n<td>Sadece hedefi belirler, kritik kararlar\u0131 onaylar.<\/td>\n<\/tr>\n<tr>\n<td><strong>Seviye 4<\/strong><\/td>\n<td>\u00c7oklu Ajan Orkestrasyonu<\/td>\n<td>Farkl\u0131 uzmanl\u0131k alan\u0131 olan ajanlar kendi aralar\u0131nda ileti\u015fim kurarak i\u015f b\u00f6l\u00fcm\u00fc yapar.<\/td>\n<td>Sadece stratejik hedefleri tan\u0131mlar.<\/td>\n<\/tr>\n<tr>\n<td><strong>Seviye 5<\/strong><\/td>\n<td>Tam Otonom ve Kendi Kendine Geli\u015fen<\/td>\n<td>Sistem kendi kodunu iyile\u015ftirir, hatalar\u0131ndan \u00f6\u011frenir ve kesintisiz \u00e7al\u0131\u015f\u0131r.<\/td>\n<td>Tamamen pasif g\u00f6zlemcidir.<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<p>G\u00fcn\u00fcm\u00fczdeki kurumsal yaz\u0131l\u0131mlar\u0131n b\u00fcy\u00fck \u00e7o\u011funlu\u011fu Seviye 2 ve Seviye 3 aras\u0131nda konumlanmaktad\u0131r. Seviye 4 ve Seviye 5 sistemler ise hen\u00fcz ara\u015ft\u0131rma laboratuvarlar\u0131nda ve ileri d\u00fczey deneysel projelerde kar\u015f\u0131m\u0131za \u00e7\u0131kmaktad\u0131r.<\/p>\n<h2>Otonom Yapay Zeka Mimarisi Nas\u0131l \u00c7al\u0131\u015f\u0131r?<\/h2>\n<p>Ba\u015far\u0131l\u0131 bir Agentic AI mimarisi, d\u00f6rt temel yap\u0131 ta\u015f\u0131n\u0131n bir araya gelmesiyle olu\u015fur: <strong>Alg\u0131lama (Perception)<\/strong>, <strong>Bellek (Memory)<\/strong>, <strong>Planlama (Planning)<\/strong> ve <strong>Eylem (Action)<\/strong>. Bu bile\u015fenler s\u00fcrekli bir d\u00f6ng\u00fc i\u00e7erisinde \u00e7al\u0131\u015farak ajan\u0131n \u00e7evresiyle etkile\u015fimde kalmas\u0131n\u0131 sa\u011flar.<\/p>\n<p>Alg\u0131lama katman\u0131, kullan\u0131c\u0131dan gelen girdileri, sistem loglar\u0131n\u0131 veya API yan\u0131tlar\u0131n\u0131 toplar. Bellek katman\u0131 ise ajan\u0131n ge\u00e7mi\u015fte yapt\u0131\u011f\u0131 eylemleri ve elde etti\u011fi sonu\u00e7lar\u0131 hat\u0131rlar. Burada iki t\u00fcr bellek kullan\u0131l\u0131r: K\u0131sa s\u00fcreli bellek (o anki konu\u015fma veya g\u00f6rev ba\u011flam\u0131) ve uzun s\u00fcreli bellek (vekt\u00f6r veri tabanlar\u0131 \u00fczerinde saklanan ge\u00e7mi\u015f deneyimler). Planlama katman\u0131, ReAct (Reasoning + Acting) veya CoT (Chain of Thought &#8211; D\u00fc\u015f\u00fcnce Zinciri) gibi teknikleri kullanarak karma\u015f\u0131k problemleri k\u00fc\u00e7\u00fck ad\u0131mlara b\u00f6ler. Son olarak eylem katman\u0131, belirlenen plan\u0131 hayata ge\u00e7irmek i\u00e7in Python kodlar\u0131 \u00e7al\u0131\u015ft\u0131r\u0131r, REST API istekleri atar veya veritaban\u0131 i\u015flemlerini ger\u00e7ekle\u015ftirir.<\/p>\n<h2>Kod \u00dczerinden Otonom Ajan Mant\u0131\u011f\u0131 Nas\u0131l Ger\u00e7ekle\u015ftirilir?<\/h2>\n<p>Bir ajan\u0131n karar alma d\u00f6ng\u00fcs\u00fcn\u00fc anlamak i\u00e7in basit bir Python \u00f6rne\u011fi \u00fczerinden ilerleyelim. A\u015fa\u011f\u0131daki kod par\u00e7ac\u0131\u011f\u0131, bir ajan\u0131n elindeki ara\u00e7lar\u0131 (tools) kullanarak kullan\u0131c\u0131 talebini nas\u0131l ad\u0131m ad\u0131m \u00e7\u00f6zd\u00fc\u011f\u00fcn\u00fc temsil eden konseptsel bir ReAct d\u00f6ng\u00fcs\u00fcn\u00fc g\u00f6stermektedir.<\/p>\n<div class=\"code-container\">\n<pre><code>import json\n\n# Ajan\u0131n kullanabilece\u011fi ara\u00e7 tan\u0131mlar\u0131\ndef veritabani_sorgula(sorgu):\n    # Temsili veri taban\u0131 fonksiyonu\n    if \"stok\" in sorgu:\n        return json.dumps({\"urun\": \"Diz\u00fcst\u00fc Bilgisayar\", \"stok\": 15})\n    return json.dumps({\"hata\": \"Veri bulunamad\u0131\"})\n\ndef eposta_gonder(alici, mesaj):\n    # Temsili e-posta g\u00f6nderme fonksiyonu\n    return f\"E-posta ba\u015far\u0131yla {alici} adresine g\u00f6nderildi.\"\n\n# Ajan\u0131n karar d\u00f6ng\u00fcs\u00fc (Reasoning Loop)\nclass OtonomAjan:\n    def __init__(self, araclar):\n        self.araclar = araclar\n        self.hafiza = []\n\n    def calistir(self, hedef):\n        print(f\"Hedef Al\u0131nd\u0131: {hedef}\")\n        self.hafiza.append({\"role\": \"user\", \"content\": hedef})\n        \n        # D\u00f6ng\u00fcsel d\u00fc\u015f\u00fcnme ve eylem ad\u0131m\u0131\n        tamamlandi = False\n        adimsayisi = 0\n        \n        while not tamamlandi and adimsayisi &lt; 3:\n            adimsayisi += 1\n            print(f\"\\n--- Ad\u0131m {adimsayisi} ---\")\n            \n            # 1. D\u00fc\u015f\u00fcnme A\u015famas\u0131 (Reasoning)\n            if adimsayisi == 1:\n                dusunce = \"Stok durumunu kontrol etmek i\u00e7in veritaban\u0131n\u0131 sorgulamal\u0131y\u0131m.\"\n                eylem = \"veritabani_sorgula\"\n                parametre = \"diz\u00fcst\u00fc bilgisayar stok\"\n            elif adimsayisi == 2:\n                dusunce = \"Stok bilgisi al\u0131nd\u0131 (15 adet). Y\u00f6neticime e-posta ile raporlamal\u0131y\u0131m.\"\n                eylem = \"eposta_gonder\"\n                parametre = {\"alici\": \"yonetici@sirket.com\", \"mesaj\": \"Diz\u00fcst\u00fc bilgisayar sto\u011fu: 15\"}\n            \n            print(f\"D\u00fc\u015f\u00fcnce: {dusunce}\")\n            print(f\"Eylem: {eylem}({parametre})\")\n            \n            # 2. Eylem A\u015famas\u0131 (Action)\n            if eylem == \"veritabani_sorgula\":\n                sonuc = self.araclar[\"veritabani_sorgula\"](parametre)\n            elif eylem == \"eposta_gonder\":\n                sonuc = self.araclar[\"eposta_gonder\"](parametre[\"alici\"], parametre[\"mesaj\"])\n                tamamlandi = True\n                \n            print(f\"G\u00f6zlem\/Sonu\u00e7: {sonuc}\")\n            self.hafiza.append({\"step\": adimsayisi, \"thought\": dusunce, \"result\": sonuc})\n\n# Ajan\u0131n kurulmas\u0131 ve \u00e7al\u0131\u015ft\u0131r\u0131lmas\u0131\nmevcut_araclar = {\n    \"veritabani_sorgula\": veritabani_sorgula,\n    \"eposta_gonder\": eposta_gonder\n}\n\najan = OtonomAjan(araclar=mevcut_araclar)\najan.calistir(\"Diz\u00fcst\u00fc bilgisayar stok durumunu kontrol et ve y\u00f6neticiye mail at.\")\n<\/code><\/pre>\n<\/div>\n<p>Yukar\u0131daki basit kod blo\u011funda g\u00f6r\u00fcld\u00fc\u011f\u00fc \u00fczere, ajan sabit bir \u00e7\u0131kt\u0131 \u00fcretmek yerine her ad\u0131mda elde etti\u011fi g\u00f6zlemleri (Observation) bir sonraki karar\u0131 almak i\u00e7in girdi olarak kullanmaktad\u0131r. Nitekim modern k\u00fct\u00fcphaneler (\u00f6rne\u011fin LangChain, CrewAI veya AutoGen) bu mant\u0131\u011f\u0131 \u00e7ok daha geli\u015fmi\u015f bellek, hata tolerans\u0131 ve dinamik LLM \u00e7a\u011fr\u0131lar\u0131 ile otomatize etmektedir.<\/p>\n<h2>T\u00fcrkiye\u2019deki \u015eirketler \u0130\u00e7in Agentic AI Vaka Analizleri Nedir?<\/h2>\n<p>Otonom yapay zeka sistemleri teorik bir kavram olman\u0131n \u00f6tesine ge\u00e7erek T\u00fcrkiye&#8217;deki farkl\u0131 sekt\u00f6rlerde somut de\u011ferler \u00fcretmeye ba\u015flam\u0131\u015ft\u0131r. \u00d6zellikle e-ticaret, finans ve lojistik alanlar\u0131nda Agentic AI d\u00f6n\u00fc\u015f\u00fcm\u00fc h\u0131z kazanmaktad\u0131r.<\/p>\n<h3>Vaka 1: E-Ticaret Lojistik ve \u0130ade Y\u00f6netimi<\/h3>\n<p>T\u00fcrkiye&#8217;nin \u00f6nde gelen bir e-ticaret platformu, m\u00fc\u015fteri hizmetlerindeki iade taleplerini y\u00f6netmek i\u00e7in geleneksel kurall\u0131 botlar yerine Seviye 3 otonom bir ajan mimarisine ge\u00e7i\u015f yapm\u0131\u015ft\u0131r. Kullan\u0131c\u0131 bir \u00fcr\u00fcn iade talebi olu\u015fturdu\u011funda ajan \u015fu s\u00fcre\u00e7leri y\u00fcr\u00fctmektedir:<\/p>\n<ul>\n<li>M\u00fc\u015fterinin ge\u00e7mi\u015f sipari\u015f ge\u00e7mi\u015fini ve sadakat puan\u0131n\u0131 kontrol eder.<\/li>\n<li>Kargo \u015firketinin API&#8217;sine ba\u011flanarak \u00fcr\u00fcn\u00fcn teslimat durumunu do\u011frular.<\/li>\n<li>E\u011fer \u00fcr\u00fcn hasarl\u0131 geldiyse, m\u00fc\u015fterinin y\u00fckledi\u011fi foto\u011fraf\u0131 g\u00f6r\u00fcnt\u00fc i\u015fleme modelleriyle analiz eder.<\/li>\n<li>\u0130ade onay\u0131n\u0131 kendisi vererek banka entegrasyonu \u00fczerinden para iadesi talimat\u0131n\u0131 ba\u015flat\u0131r.<\/li>\n<\/ul>\n<p>Bu uygulama sayesinde insan temsilcilerin m\u00fcdahale etmesi gereken bilet say\u0131s\u0131 %65 oran\u0131nda azalm\u0131\u015f ve s\u00fcre\u00e7 \u00e7\u00f6z\u00fcm s\u00fcresi dakikalardan saniyelere inmi\u015ftir.<\/p>\n<h3>Vaka 2: Bankac\u0131l\u0131kta Doland\u0131r\u0131c\u0131l\u0131k (Fraud) Tespiti ve M\u00fcdahalesi<\/h3>\n<p>Bir T\u00fcrk bankas\u0131, \u015f\u00fcpheli i\u015flem tespiti sonras\u0131nda devreye giren bir otonom g\u00fcvenlik ajan\u0131 konumland\u0131rm\u0131\u015ft\u0131r. Sistem, ola\u011fand\u0131\u015f\u0131 bir transfer alg\u0131lad\u0131\u011f\u0131nda sadece alarm \u00fcretmekle kalmaz; m\u00fc\u015fterinin ge\u00e7mi\u015f transfer kal\u0131plar\u0131n\u0131 inceler, risk skorunu hesaplar ve gerekirse kart\u0131 ge\u00e7ici olarak bloke edip m\u00fc\u015fteriye SMS veya WhatsApp \u00fczerinden do\u011frulama mesaj\u0131 g\u00f6nderir. M\u00fc\u015fteriden gelen yan\u0131t onayland\u0131\u011f\u0131nda bloke i\u015flemini otomatik olarak kald\u0131r\u0131r. Bu yakla\u015f\u0131m, sadece tespitte kalan eski sistemlerin aksine olaya an\u0131nda m\u00fcdahale eden aktif bir koruma sa\u011flar.<\/p>\n<h2>Karma\u015f\u0131k Sistemlerde Agentic AI Entegrasyonu Nas\u0131l Yap\u0131l\u0131r?<\/h2>\n<p>\u015eirketlerin Agentic AI d\u00f6n\u00fc\u015f\u00fcm\u00fcnde en \u00e7ok zorland\u0131\u011f\u0131 konu, ajanlar\u0131n kontrols\u00fcz eylemler ger\u00e7ekle\u015ftirmesini engellemektir. Bir ajan\u0131n yanl\u0131\u015f bir SQL sorgusu ile kritik bir tabloyu silmesi veya m\u00fc\u015fterilere hatal\u0131 indirim tan\u0131mlamas\u0131 telafisi zor zararlara yol a\u00e7abilir. Bu nedenle kurumsal seviyede ajan entegrasyonu yap\u0131l\u0131rken \u015fu ileri d\u00fczey teknikler uygulanmal\u0131d\u0131r:<\/p>\n<ol>\n<li><strong>Korkuluklar (Guardrails) ve Do\u011frulama Katmanlar\u0131:<\/strong> Ajan\u0131n \u00fcretti\u011fi parametreler ve API \u00e7a\u011fr\u0131lar\u0131 \u00e7al\u0131\u015ft\u0131rmadan \u00f6nce bir g\u00fcvenlik katman\u0131ndan ge\u00e7irilmelidir. <code>pydantic<\/code> gibi veri do\u011frulama k\u00fct\u00fcphaneleriyle \u015fema kontrol\u00fc yap\u0131lmal\u0131d\u0131r.<\/li>\n<li><strong>\u0130nsan Denetimli D\u00f6ng\u00fc (Human-in-the-loop &#8211; HITL):<\/strong> Kritik eylemlerde (\u00f6rne\u011fin 10.000 TL \u00fczeri \u00f6demeler veya silme i\u015flemleri) ajan\u0131n insandan onay almas\u0131 zorunlu k\u0131l\u0131nmal\u0131d\u0131r. Ajan s\u00fcreci haz\u0131rlar, insan onaylad\u0131\u011f\u0131 an eylemi tamamlar.<\/li>\n<li><strong>Maliyet ve Token Y\u00f6netimi:<\/strong> Ajanlar sonsuz bir d\u00f6ng\u00fcye (infinite loop) girebilir. Bu nedenle her ajana maksimum ad\u0131m say\u0131s\u0131 (max_iterations) ve harcayabilece\u011fi token limiti koyulmal\u0131d\u0131r.<\/li>\n<li><strong>Belirleyici Olmayan Davran\u0131\u015flar\u0131n Test Edilmesi:<\/strong> Yaz\u0131l\u0131m test s\u00fcre\u00e7lerinde LLM yan\u0131tlar\u0131n\u0131n de\u011fi\u015fkenli\u011fi dikkate al\u0131nmal\u0131d\u0131r. Ajan\u0131n \u00fcretti\u011fi kararlar s\u00fcrekli loglanmal\u0131 ve geriye d\u00f6n\u00fck izlenebilirlik (traceability) sa\u011flanmal\u0131d\u0131r.<\/li>\n<\/ol>\n<h2>Otonom Yapay Zeka Hakk\u0131nda S\u0131k\u00e7a Sorulan Sorular<\/h2>\n<h3>Agentic AI, geleneksel RPA (RPA &#8211; Yapay Zekas\u0131z S\u00fcre\u00e7 Otomasyonu) sistemlerinden nas\u0131l ayr\u0131l\u0131r?<\/h3>\n<p>RPA (Robotic Process Automation) tamamen kural tabanl\u0131d\u0131r ve belirli ad\u0131mlar\u0131 rijit bir \u015fekilde takip eder. Ekranda bir butonun yeri de\u011fi\u015fti\u011finde veya beklenmeyen bir girdi geldi\u011finde RPA patlar. Agentic AI ise belirsizlik kar\u015f\u0131s\u0131nda ak\u0131l y\u00fcr\u00fctebilir, plan\u0131n\u0131 dinamik olarak de\u011fi\u015ftirebilir ve farkl\u0131 ara\u00e7lar\u0131 esnek bir \u015fekilde kullanabilir.<\/p>\n<h3>Sistemlerimizin Seviye 3 veya Seviye 4 Otonomiye Ula\u015fmas\u0131 G\u00fcvenli midir?<\/h3>\n<p>G\u00fcvenlik, ajan\u0131n sahip oldu\u011fu yetki alan\u0131yla (scope) do\u011frudan ili\u015fkilidir. Do\u011fru g\u00fcvenlik bariyerleri (Guardrails) koyuldu\u011fu, hassas i\u015flemlerde &#8220;Human-in-the-loop&#8221; mekanizmas\u0131 \u00e7al\u0131\u015ft\u0131r\u0131ld\u0131\u011f\u0131 ve ajanlar\u0131n yetkileri k\u0131s\u0131tland\u0131\u011f\u0131 s\u00fcrece \u00fcst seviye otonomi g\u00fcvenle kullan\u0131labilir.<\/p>\n<h3>Bir Agentic AI projesi geli\u015ftirmek i\u00e7in hangi k\u00fct\u00fcphaneler tercih edilmelidir?<\/h3>\n<p>Python ekosisteminde en pop\u00fcler a\u00e7\u0131k kaynakl\u0131 \u00e7at\u0131lar (frameworks); LangChain, LangGraph, CrewAI, AutoGen ve LlamaIndex&#8217;tir. Bu k\u00fct\u00fcphaneler ajan\u0131n bellek, karar alma ve \u00e7oklu ajan orkestrasyon s\u00fcre\u00e7lerini standart kal\u0131plar halinde sunar.<\/p>\n<h3>Ajanlar\u0131n sonsuz d\u00f6ng\u00fcye girmesi nas\u0131l engellenir?<\/h3>\n<p>Ajan mimarilerine mutlaka teknik k\u0131s\u0131tlamalar eklenmelidir. Maksimum ad\u0131m say\u0131s\u0131 (\u00f6rne\u011fin en fazla 10 ad\u0131m), zaman a\u015f\u0131m\u0131 (timeout) s\u00fcreleri ve ayn\u0131 eylemin art arda tekrar edilmesini engelleyen kontrol mekanizmalar\u0131 mimariye entegre edilmelidir.<\/p>\n<h2>Sonu\u00e7: Otonomi Bir Var\u0131\u015f Noktas\u0131 De\u011fil Yolculuktur<\/h2>\n<p>Yapay zeka teknolojisi h\u0131zla ilerlerken, \u015firketlerin ve yaz\u0131l\u0131m geli\u015ftiricilerin yapaca\u011f\u0131 en b\u00fcy\u00fck hata, Agentic AI yap\u0131s\u0131n\u0131 kutudan \u00e7\u0131kan haz\u0131r bir \u00fcr\u00fcn olarak g\u00f6rmektir. Otonomi bir spektrumdur. En basit y\u00f6nlendirmeli zincirlerden, kendi kendine karar alan \u00e7oklu ajan orkestrasyonlar\u0131na kadar her a\u015fama i\u015f ihtiya\u00e7lar\u0131na g\u00f6re de\u011ferlendirilmelidir.<\/p>\n<p>Ba\u015far\u0131l\u0131 bir yapay zeka stratejisi, ihtiyaca en uygun otonomi seviyesini se\u00e7mekten ge\u00e7er. Do\u011fru tasarlanm\u0131\u015f mimariler, kat\u0131 g\u00fcvenlik bariyerleri ve etkili bellek y\u00f6netimi ile desteklenen otonom ajanlar, gelece\u011fin yaz\u0131l\u0131m d\u00fcnyas\u0131n\u0131n temel yap\u0131 ta\u015f\u0131 olacakt\u0131r.<\/p>\n<p>#YapayZeka #AgenticAI #YazilimGelistirme #VeriBilimi #Teknoloji<\/p>\n","protected":false},"excerpt":{"rendered":"Otonom yapay zeka (Agentic AI) basit bir pazarlama etiketi de\u011fil; dinamik karar alma ve eylem yetene\u011fine dayanan geni\u015f bir geli\u015fim spektrumudur.","protected":false},"author":1,"featured_media":0,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"csco_page_header_type":"","csco_page_load_nextpost":"","csco_page_subscribe_form":"","csco_page_contact_form":"","footnotes":""},"categories":[1],"tags":[],"class_list":{"0":"post-43813","1":"post","2":"type-post","3":"status-publish","4":"format-standard","6":"category-genel","7":"cs-entry","8":"cs-video-wrap"},"yoast_head":"<!-- This site is optimized with the Yoast SEO Premium plugin v20.5 (Yoast SEO v25.3.1) - https:\/\/yoast.com\/wordpress\/plugins\/seo\/ -->\n<title>Otonom Yapay Zeka: Bir Etiket De\u011fil Spektrum - Kodlar\u0131n Gizemli D\u00fcnyas\u0131<\/title>\n<meta name=\"robots\" content=\"index, follow, max-snippet:-1, max-image-preview:large, max-video-preview:-1\" \/>\n<link rel=\"canonical\" href=\"https:\/\/fatihsoysal.com\/blog\/otonom-yapay-zeka-bir-etiket-degil-spektrum\/\" \/>\n<meta property=\"og:locale\" content=\"tr_TR\" \/>\n<meta property=\"og:type\" content=\"article\" \/>\n<meta property=\"og:title\" content=\"Otonom Yapay Zeka: Bir Etiket De\u011fil Spektrum\" \/>\n<meta property=\"og:description\" content=\"Otonom yapay zeka (Agentic AI) basit bir pazarlama etiketi de\u011fil; dinamik karar alma ve eylem yetene\u011fine dayanan geni\u015f bir geli\u015fim spektrumudur.\" \/>\n<meta property=\"og:url\" content=\"https:\/\/fatihsoysal.com\/blog\/otonom-yapay-zeka-bir-etiket-degil-spektrum\/\" \/>\n<meta property=\"og:site_name\" content=\"Kodlar\u0131n Gizemli D\u00fcnyas\u0131\" \/>\n<meta property=\"article:published_time\" content=\"2026-08-03T18:04:03+00:00\" \/>\n<meta name=\"author\" content=\"Fatih Soysal\" \/>\n<meta name=\"twitter:card\" content=\"summary_large_image\" \/>\n<meta name=\"twitter:label1\" content=\"Yazan:\" \/>\n\t<meta name=\"twitter:data1\" content=\"Fatih Soysal\" \/>\n\t<meta name=\"twitter:label2\" content=\"Tahmini okuma s\u00fcresi\" \/>\n\t<meta name=\"twitter:data2\" content=\"12 dakika\" \/>\n<script type=\"application\/ld+json\" class=\"yoast-schema-graph\">{\"@context\":\"https:\/\/schema.org\",\"@graph\":[{\"@type\":\"Article\",\"@id\":\"https:\/\/fatihsoysal.com\/blog\/otonom-yapay-zeka-bir-etiket-degil-spektrum\/#article\",\"isPartOf\":{\"@id\":\"https:\/\/fatihsoysal.com\/blog\/otonom-yapay-zeka-bir-etiket-degil-spektrum\/\"},\"author\":{\"name\":\"Fatih Soysal\",\"@id\":\"https:\/\/fatihsoysal.com\/blog\/#\/schema\/person\/002a254750921dcfd568a99e48240dd1\"},\"headline\":\"Otonom Yapay Zeka: Bir Etiket De\u011fil Spektrum\",\"datePublished\":\"2026-08-03T18:04:03+00:00\",\"mainEntityOfPage\":{\"@id\":\"https:\/\/fatihsoysal.com\/blog\/otonom-yapay-zeka-bir-etiket-degil-spektrum\/\"},\"wordCount\":2191,\"commentCount\":0,\"publisher\":{\"@id\":\"https:\/\/fatihsoysal.com\/blog\/#\/schema\/person\/002a254750921dcfd568a99e48240dd1\"},\"inLanguage\":\"tr\",\"potentialAction\":[{\"@type\":\"CommentAction\",\"name\":\"Comment\",\"target\":[\"https:\/\/fatihsoysal.com\/blog\/otonom-yapay-zeka-bir-etiket-degil-spektrum\/#respond\"]}],\"copyrightYear\":\"2026\",\"copyrightHolder\":{\"@id\":\"https:\/\/fatihsoysal.com\/blog\/#organization\"}},{\"@type\":\"WebPage\",\"@id\":\"https:\/\/fatihsoysal.com\/blog\/otonom-yapay-zeka-bir-etiket-degil-spektrum\/\",\"url\":\"https:\/\/fatihsoysal.com\/blog\/otonom-yapay-zeka-bir-etiket-degil-spektrum\/\",\"name\":\"Otonom Yapay Zeka: Bir Etiket De\u011fil Spektrum - Kodlar\u0131n Gizemli D\u00fcnyas\u0131\",\"isPartOf\":{\"@id\":\"https:\/\/fatihsoysal.com\/blog\/#website\"},\"datePublished\":\"2026-08-03T18:04:03+00:00\",\"breadcrumb\":{\"@id\":\"https:\/\/fatihsoysal.com\/blog\/otonom-yapay-zeka-bir-etiket-degil-spektrum\/#breadcrumb\"},\"inLanguage\":\"tr\",\"potentialAction\":[{\"@type\":\"ReadAction\",\"target\":[\"https:\/\/fatihsoysal.com\/blog\/otonom-yapay-zeka-bir-etiket-degil-spektrum\/\"]}]},{\"@type\":\"BreadcrumbList\",\"@id\":\"https:\/\/fatihsoysal.com\/blog\/otonom-yapay-zeka-bir-etiket-degil-spektrum\/#breadcrumb\",\"itemListElement\":[{\"@type\":\"ListItem\",\"position\":1,\"name\":\"Anasayfa\",\"item\":\"https:\/\/fatihsoysal.com\/blog\/\"},{\"@type\":\"ListItem\",\"position\":2,\"name\":\"Otonom Yapay Zeka: Bir Etiket De\u011fil Spektrum\"}]},{\"@type\":\"WebSite\",\"@id\":\"https:\/\/fatihsoysal.com\/blog\/#website\",\"url\":\"https:\/\/fatihsoysal.com\/blog\/\",\"name\":\"Fatihsoysal.com\",\"description\":\"Blog - Yaz\u0131l\u0131m D\u00fcnyas\u0131 Tecr\u00fcbelerim\",\"publisher\":{\"@id\":\"https:\/\/fatihsoysal.com\/blog\/#\/schema\/person\/002a254750921dcfd568a99e48240dd1\"},\"potentialAction\":[{\"@type\":\"SearchAction\",\"target\":{\"@type\":\"EntryPoint\",\"urlTemplate\":\"https:\/\/fatihsoysal.com\/blog\/?s={search_term_string}\"},\"query-input\":{\"@type\":\"PropertyValueSpecification\",\"valueRequired\":true,\"valueName\":\"search_term_string\"}}],\"inLanguage\":\"tr\"},{\"@type\":[\"Person\",\"Organization\"],\"@id\":\"https:\/\/fatihsoysal.com\/blog\/#\/schema\/person\/002a254750921dcfd568a99e48240dd1\",\"name\":\"Fatih Soysal\",\"image\":{\"@type\":\"ImageObject\",\"inLanguage\":\"tr\",\"@id\":\"https:\/\/fatihsoysal.com\/blog\/#\/schema\/person\/image\/\",\"url\":\"https:\/\/fatihsoysal.com\/blog\/wp-content\/uploads\/2024\/04\/cropped-replicate-prediction-3kgg1hgjn5rgp0cf0p5tr0jw7w-1.png\",\"contentUrl\":\"https:\/\/fatihsoysal.com\/blog\/wp-content\/uploads\/2024\/04\/cropped-replicate-prediction-3kgg1hgjn5rgp0cf0p5tr0jw7w-1.png\",\"width\":512,\"height\":512,\"caption\":\"Fatih Soysal\"},\"logo\":{\"@id\":\"https:\/\/fatihsoysal.com\/blog\/#\/schema\/person\/image\/\"},\"description\":\"Kullan\u0131m ve kodlama m\u00fckemmeliyetini odak alan uygulamalar olu\u015fturma deneyimine sahip, profesyonel olarak 15+ y\u0131l \u00fczeri deneyime sahip bir yaz\u0131l\u0131m m\u00fchendisi.\",\"url\":\"https:\/\/fatihsoysal.com\/blog\/author\/fatihsoysal\/\"}]}<\/script>\n<!-- \/ Yoast SEO Premium plugin. -->","yoast_head_json":{"title":"Otonom Yapay Zeka: Bir Etiket De\u011fil Spektrum - Kodlar\u0131n Gizemli D\u00fcnyas\u0131","robots":{"index":"index","follow":"follow","max-snippet":"max-snippet:-1","max-image-preview":"max-image-preview:large","max-video-preview":"max-video-preview:-1"},"canonical":"https:\/\/fatihsoysal.com\/blog\/otonom-yapay-zeka-bir-etiket-degil-spektrum\/","og_locale":"tr_TR","og_type":"article","og_title":"Otonom Yapay Zeka: Bir Etiket De\u011fil Spektrum","og_description":"Otonom yapay zeka (Agentic AI) basit bir pazarlama etiketi de\u011fil; dinamik karar alma ve eylem yetene\u011fine dayanan geni\u015f bir geli\u015fim spektrumudur.","og_url":"https:\/\/fatihsoysal.com\/blog\/otonom-yapay-zeka-bir-etiket-degil-spektrum\/","og_site_name":"Kodlar\u0131n Gizemli D\u00fcnyas\u0131","article_published_time":"2026-08-03T18:04:03+00:00","author":"Fatih Soysal","twitter_card":"summary_large_image","twitter_misc":{"Yazan:":"Fatih Soysal","Tahmini okuma s\u00fcresi":"12 dakika"},"schema":{"@context":"https:\/\/schema.org","@graph":[{"@type":"Article","@id":"https:\/\/fatihsoysal.com\/blog\/otonom-yapay-zeka-bir-etiket-degil-spektrum\/#article","isPartOf":{"@id":"https:\/\/fatihsoysal.com\/blog\/otonom-yapay-zeka-bir-etiket-degil-spektrum\/"},"author":{"name":"Fatih Soysal","@id":"https:\/\/fatihsoysal.com\/blog\/#\/schema\/person\/002a254750921dcfd568a99e48240dd1"},"headline":"Otonom Yapay Zeka: Bir Etiket De\u011fil Spektrum","datePublished":"2026-08-03T18:04:03+00:00","mainEntityOfPage":{"@id":"https:\/\/fatihsoysal.com\/blog\/otonom-yapay-zeka-bir-etiket-degil-spektrum\/"},"wordCount":2191,"commentCount":0,"publisher":{"@id":"https:\/\/fatihsoysal.com\/blog\/#\/schema\/person\/002a254750921dcfd568a99e48240dd1"},"inLanguage":"tr","potentialAction":[{"@type":"CommentAction","name":"Comment","target":["https:\/\/fatihsoysal.com\/blog\/otonom-yapay-zeka-bir-etiket-degil-spektrum\/#respond"]}],"copyrightYear":"2026","copyrightHolder":{"@id":"https:\/\/fatihsoysal.com\/blog\/#organization"}},{"@type":"WebPage","@id":"https:\/\/fatihsoysal.com\/blog\/otonom-yapay-zeka-bir-etiket-degil-spektrum\/","url":"https:\/\/fatihsoysal.com\/blog\/otonom-yapay-zeka-bir-etiket-degil-spektrum\/","name":"Otonom Yapay Zeka: Bir Etiket De\u011fil Spektrum - Kodlar\u0131n Gizemli D\u00fcnyas\u0131","isPartOf":{"@id":"https:\/\/fatihsoysal.com\/blog\/#website"},"datePublished":"2026-08-03T18:04:03+00:00","breadcrumb":{"@id":"https:\/\/fatihsoysal.com\/blog\/otonom-yapay-zeka-bir-etiket-degil-spektrum\/#breadcrumb"},"inLanguage":"tr","potentialAction":[{"@type":"ReadAction","target":["https:\/\/fatihsoysal.com\/blog\/otonom-yapay-zeka-bir-etiket-degil-spektrum\/"]}]},{"@type":"BreadcrumbList","@id":"https:\/\/fatihsoysal.com\/blog\/otonom-yapay-zeka-bir-etiket-degil-spektrum\/#breadcrumb","itemListElement":[{"@type":"ListItem","position":1,"name":"Anasayfa","item":"https:\/\/fatihsoysal.com\/blog\/"},{"@type":"ListItem","position":2,"name":"Otonom Yapay Zeka: Bir Etiket De\u011fil Spektrum"}]},{"@type":"WebSite","@id":"https:\/\/fatihsoysal.com\/blog\/#website","url":"https:\/\/fatihsoysal.com\/blog\/","name":"Fatihsoysal.com","description":"Blog - Yaz\u0131l\u0131m D\u00fcnyas\u0131 Tecr\u00fcbelerim","publisher":{"@id":"https:\/\/fatihsoysal.com\/blog\/#\/schema\/person\/002a254750921dcfd568a99e48240dd1"},"potentialAction":[{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https:\/\/fatihsoysal.com\/blog\/?s={search_term_string}"},"query-input":{"@type":"PropertyValueSpecification","valueRequired":true,"valueName":"search_term_string"}}],"inLanguage":"tr"},{"@type":["Person","Organization"],"@id":"https:\/\/fatihsoysal.com\/blog\/#\/schema\/person\/002a254750921dcfd568a99e48240dd1","name":"Fatih Soysal","image":{"@type":"ImageObject","inLanguage":"tr","@id":"https:\/\/fatihsoysal.com\/blog\/#\/schema\/person\/image\/","url":"https:\/\/fatihsoysal.com\/blog\/wp-content\/uploads\/2024\/04\/cropped-replicate-prediction-3kgg1hgjn5rgp0cf0p5tr0jw7w-1.png","contentUrl":"https:\/\/fatihsoysal.com\/blog\/wp-content\/uploads\/2024\/04\/cropped-replicate-prediction-3kgg1hgjn5rgp0cf0p5tr0jw7w-1.png","width":512,"height":512,"caption":"Fatih Soysal"},"logo":{"@id":"https:\/\/fatihsoysal.com\/blog\/#\/schema\/person\/image\/"},"description":"Kullan\u0131m ve kodlama m\u00fckemmeliyetini odak alan uygulamalar olu\u015fturma deneyimine sahip, profesyonel olarak 15+ y\u0131l \u00fczeri deneyime sahip bir yaz\u0131l\u0131m m\u00fchendisi.","url":"https:\/\/fatihsoysal.com\/blog\/author\/fatihsoysal\/"}]}},"yoast_meta":{"yoast_wpseo_title":"","yoast_wpseo_metadesc":"","yoast_wpseo_canonical":""},"amp_enabled":true,"_links":{"self":[{"href":"https:\/\/fatihsoysal.com\/blog\/wp-json\/wp\/v2\/posts\/43813","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/fatihsoysal.com\/blog\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/fatihsoysal.com\/blog\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/fatihsoysal.com\/blog\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/fatihsoysal.com\/blog\/wp-json\/wp\/v2\/comments?post=43813"}],"version-history":[{"count":0,"href":"https:\/\/fatihsoysal.com\/blog\/wp-json\/wp\/v2\/posts\/43813\/revisions"}],"wp:attachment":[{"href":"https:\/\/fatihsoysal.com\/blog\/wp-json\/wp\/v2\/media?parent=43813"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/fatihsoysal.com\/blog\/wp-json\/wp\/v2\/categories?post=43813"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/fatihsoysal.com\/blog\/wp-json\/wp\/v2\/tags?post=43813"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}