{"id":45140,"date":"2026-10-10T21:11:51","date_gmt":"2026-10-10T18:11:51","guid":{"rendered":"https:\/\/fatihsoysal.com\/blog\/echo-the-marker-before-you-edit-teknigi-ile-hatasiz-kod\/"},"modified":"2026-10-10T21:11:51","modified_gmt":"2026-10-10T18:11:51","slug":"echo-the-marker-before-you-edit-teknigi-ile-hatasiz-kod","status":"publish","type":"post","link":"https:\/\/fatihsoysal.com\/blog\/echo-the-marker-before-you-edit-teknigi-ile-hatasiz-kod\/","title":{"rendered":"Echo the Marker Before You Edit Tekni\u011fi ile Hatas\u0131z Kod"},"content":{"rendered":"

Echo the Marker Before You Edit Tekni\u011fi ile Hatas\u0131z Kod<\/h2>\n

Echo the Marker Before You Edit tekni\u011fi, b\u00fcy\u00fck dil modellerinin kod d\u00fczenleme yaparken ba\u011flam\u0131 kaybetmesini \u00f6nleyen ve kusursuz yama \u00fcreten modern bir y\u00f6ntemdir.<\/p>\n

Yapay zeka tabanl\u0131 kodlama asistanlar\u0131 hayat\u0131m\u0131za h\u0131zla girdi; ancak bu ara\u00e7lar\u0131 kullanan her yaz\u0131l\u0131mc\u0131 benzer bir hayal k\u0131r\u0131kl\u0131\u011f\u0131n\u0131 mutlaka ya\u015fam\u0131\u015ft\u0131r. Modelden 500 sat\u0131rl\u0131k karma\u015f\u0131k bir dosyadaki tek bir fonksiyonu g\u00fcncellemesini istersiniz. Yapay zeka kodu ba\u015ftan yazar, bu s\u0131rada dosyan\u0131n di\u011fer ucundaki kritik bir yard\u0131mc\u0131 fonksiyonu siler veya sat\u0131r numaralar\u0131n\u0131 kar\u0131\u015ft\u0131rarak yamay\u0131 (patch) tamamen uygulanamaz hale getirir. \u00d6zellikle b\u00fcy\u00fck projelerde bu durum, geli\u015ftiricinin yapay zekaya ay\u0131rd\u0131\u011f\u0131 zamandan daha fazlas\u0131n\u0131 hata ay\u0131klamaya harcamas\u0131na yol a\u00e7ar. \u0130\u015fte bu noktada Echo the Marker Before You Edit<\/strong> (D\u00fczenlemeden \u00d6nce \u0130\u015faret\u00e7iyi Yank\u0131la) prensibi devreye girer. Bu makalede, s\u00f6z konusu y\u00f6ntemin teorik temellerini, neden geleneksel yakla\u015f\u0131mlardan \u00fcst\u00fcn oldu\u011funu ve yerel i\u015f ak\u0131\u015flar\u0131n\u0131za nas\u0131l entegre edebilece\u011finizi derinlemesine inceleyece\u011fiz.<\/p>\n

B\u00fcy\u00fck Dil Modelleri Neden Kod D\u00fczenlerken Hata Yapar?<\/h3>\n

B\u00fcy\u00fck dil modelleri (LLM – Large Language Model), metinleri ard\u0131\u015f\u0131k olarak (oto-regresif bi\u00e7imde) bir sonraki belirteci (token) tahmin ederek \u00fcretir. Bir modele “falanca dosyadaki kullan\u0131c\u0131 do\u011frulama fonksiyonunu g\u00fcncelle” dedi\u011finizde, model t\u00fcm dosyay\u0131 belle\u011finde bir kerede tutup tek hamlede \u00e7\u0131kt\u0131 vermez. Aksine, her yeni kelimeyi \u00fcretirken \u00f6nceki \u00e7\u0131kt\u0131lara dikkat (attention) mekanizmas\u0131 \u00fczerinden bakar. Dosya boyutu b\u00fcy\u00fcd\u00fck\u00e7e veya d\u00fczenleme yap\u0131lacak b\u00f6lge dosyan\u0131n derinliklerinde yer ald\u0131k\u00e7a, modelin dikkat da\u011f\u0131l\u0131m\u0131 zay\u0131flar. Bu durum literat\u00fcrde ba\u011flam kaymas\u0131 (context drift) veya dikkat t\u00fckenmesi olarak adland\u0131r\u0131l\u0131r.<\/p>\n

Geleneksel yakla\u015f\u0131mlar genellikle iki yoldan birini se\u00e7er: Ya t\u00fcm dosyay\u0131 ba\u015ftan sona yeniden yazd\u0131rmak ya da Git benzeri bir fark (diff) \u00e7\u0131kt\u0131s\u0131 \u00fcretmek. T\u00fcm dosyay\u0131 yeniden yazd\u0131rmak korkun\u00e7 bir kaynak israf\u0131d\u0131r; \u00e7\u00fcnk\u00fc hem API maliyetlerini katlar hem de yan\u0131t s\u00fcresini dakikalara \u00e7\u0131kar\u0131r. Ayr\u0131ca modelin, d\u00fczenleme alan\u0131 d\u0131\u015f\u0131ndaki \u00e7al\u0131\u015fan kodlar\u0131 “hal\u00fcsinasyon” yoluyla de\u011fi\u015ftirmesine zemin haz\u0131rlar. \u0130kinci se\u00e7enek olan geleneksel fark \u00e7\u0131kt\u0131s\u0131 ise bo\u015fluk duyarl\u0131l\u0131\u011f\u0131 (whitespace sensitivity) ve modelin sat\u0131r say\u0131lar\u0131n\u0131 do\u011fru sayamamas\u0131 nedeniyle s\u0131k s\u0131k \u00e7\u00f6ker. Model, sat\u0131r 42’de de\u011fi\u015fiklik yapt\u0131\u011f\u0131n\u0131 iddia ederken asl\u0131nda sat\u0131r 78’i hedef al\u0131yor olabilir. Bu nedenle geli\u015ftiriciler, modelin tam olarak nereye odakland\u0131\u011f\u0131n\u0131 kesinle\u015ftiren deterministik tekniklere ihtiya\u00e7 duyar.<\/p>\n

Bunun sonucunda ortaya \u00e7\u0131kan yakla\u015f\u0131m, yapay zekan\u0131n serbest\u00e7e kod \u00fcretmeye ba\u015flamadan \u00f6nce kendisini belirli bir koordinata kilitlemesini zorunlu k\u0131lar. E\u011fer model, de\u011fi\u015ftirece\u011fi kodun hemen \u00f6ncesindeki benzersiz bir i\u015faret\u00e7iyi (marker) birebir tekrar ederse (echo), dikkat mekanizmas\u0131n\u0131 do\u011frudan bu referans noktas\u0131na odaklar. B\u00f6ylece hal\u00fcsinasyon ihtimali matematiksel olarak en aza iner.<\/p>\n

Echo the Marker Mant\u0131\u011f\u0131 Nedir ve Nas\u0131l \u00c7al\u0131\u015f\u0131r?<\/h3>\n

Echo the Marker Before You Edit tekni\u011fi, modele d\u00fczenleme yapaca\u011f\u0131 alan\u0131n hemen \u00f6ncesinde yer alan benzersiz bir \u00e7apa (anchor\/marker) dizesini harfi harfine geri yazd\u0131rma zorunlulu\u011fu getiren bir prompt m\u00fchendisli\u011fi ve ayr\u0131\u015ft\u0131rma (parsing) mimarisidir. Buradaki “yank\u0131lama” (echo) i\u015flemi bir formalite de\u011fildir; modelin dikkat katmanlar\u0131n\u0131 tam istenen koordinata sabitleyen bir bili\u015fsel kald\u0131ra\u00e7t\u0131r.<\/p>\n

LLM mimarisinde model bir \u00e7\u0131kt\u0131 \u00fcretirken, \u00fcretti\u011fi her yeni belirte\u00e7 bir sonraki belirtecin ko\u015fullu olas\u0131l\u0131\u011f\u0131n\u0131 belirler. Model, kod taban\u0131nda var olan benzersiz bir yorum sat\u0131r\u0131n\u0131 veya kesin kod imzas\u0131n\u0131 kelimesi kelimesine yazmaya ba\u015flad\u0131\u011f\u0131nda, transformat\u00f6r mimarisi ba\u011flam penceresindeki ilgili blo\u011fu en y\u00fcksek olas\u0131l\u0131kla etkinle\u015ftirir. Yani model “rastgele bir yerde kod d\u00fczenlemeye” ba\u015flamaz; \u00f6nce hedefini teyit eder, ard\u0131ndan de\u011fi\u015fikli\u011fi uygular. Bu s\u00fcre\u00e7 \u00fc\u00e7 temel a\u015famadan olu\u015fur:<\/p>\n

    \n
  1. \u0130\u015faret\u00e7i Tan\u0131mlama (Marker Placement):<\/strong> Kod i\u00e7erisine yapay zekan\u0131n arayabilece\u011fi benzersiz bir \u00e7apa metni (\u00f6rne\u011fin benzersiz bir fonksiyon ad\u0131, \u00f6zel bir yorum etiketi veya hash) yerle\u015ftirilir.<\/li>\n
  2. Yank\u0131lama A\u015famas\u0131 (Echoing):<\/strong> Modele verilen sistem talimat\u0131, d\u00fczenlemeye ba\u015flamadan \u00f6nce bu i\u015faret\u00e7iyi kesinlikle de\u011fi\u015ftirmeden ve birebir \u015fekilde \u00e7\u0131kt\u0131ya basmas\u0131n\u0131 emreder.<\/li>\n
  3. Deterministik E\u015fleme (Regex \/ AST Match):<\/strong> Yerel derleyici veya d\u00fczenleyici istemci, modelin \u00e7\u0131kt\u0131s\u0131ndaki bu yank\u0131lanan i\u015faret\u00e7iyi yakalar. Ard\u0131ndan gelen yeni kodu do\u011frudan bu i\u015faret\u00e7inin arkas\u0131na ekler veya i\u015faret\u00e7iyle s\u0131n\u0131rland\u0131r\u0131lm\u0131\u015f blo\u011fu g\u00fcnceller.<\/li>\n<\/ol>\n

    Bu yakla\u015f\u0131m sayesinde, modelin sat\u0131r numaralar\u0131n\u0131 tahmin etmesine ya da dosyan\u0131n onlarca sat\u0131rl\u0131k \u00e7evresel ba\u011flam\u0131n\u0131 ezberlemesine gerek kalmaz. \u0130\u015faret\u00e7inin varl\u0131\u011f\u0131, sistemin hata tolerans\u0131n\u0131 s\u0131f\u0131ra yakla\u015ft\u0131r\u0131r.<\/p>\n

    Geleneksel Diff Y\u00f6ntemleri ile Echo Tekni\u011finin Kar\u015f\u0131la\u015ft\u0131rmas\u0131<\/h3>\n

    Yapay zeka tabanl\u0131 geli\u015ftirme ortamlar\u0131nda diff \u00fcretimi konusunda uzun s\u00fcredir ciddi tart\u0131\u015fmalar ya\u015fanmaktad\u0131r. Aider, Cursor veya Continue gibi ara\u00e7lar farkl\u0131 yakla\u015f\u0131mlar dener. A\u015fa\u011f\u0131daki tablo, yayg\u0131n kod d\u00fczenleme y\u00f6ntemleri ile Echo the Marker tekni\u011fi aras\u0131ndaki yap\u0131sal farklar\u0131 net bir bi\u00e7imde ortaya koymaktad\u0131r:<\/p>\n\n\n\n\n\n\n\n\n\n
    Kriter<\/th>\nTam Dosya Yeniden Yaz\u0131m\u0131<\/th>\nArama\/De\u011fi\u015ftirme (Search & Replace)<\/th>\nEcho the Marker Tekni\u011fi<\/th>\n<\/tr>\n<\/thead>\n
    Token T\u00fcketimi<\/strong><\/td>\n\u00c7ok Y\u00fcksek (T\u00fcm dosya)<\/td>\nOrta (Arama ve de\u011fi\u015fim bloklar\u0131)<\/td>\nD\u00fc\u015f\u00fck (Sadece i\u015faret\u00e7i ve yeni kod)<\/td>\n<\/tr>\n
    Bo\u015fluk Duyarl\u0131l\u0131\u011f\u0131<\/strong><\/td>\n\u00d6nemsiz<\/td>\n\u00c7ok Y\u00fcksek (Girintileme hatas\u0131 yamay\u0131 bozar)<\/td>\nYok (\u0130\u015faret\u00e7i dizesi do\u011frudan e\u015flenir)<\/td>\n<\/tr>\n
    Sat\u0131r Kaymas\u0131 Riski<\/strong><\/td>\nYok<\/td>\nY\u00fcksek (Benzer kod sat\u0131rlar\u0131nda yan\u0131lma)<\/td>\nYok (Benzersiz i\u015faret\u00e7i garantisi)<\/td>\n<\/tr>\n
    \u0130\u015flem H\u0131z\u0131 (Gecikme)<\/strong><\/td>\n\u00c7ok Yava\u015f<\/td>\nH\u0131zl\u0131<\/td>\n\u00c7ok H\u0131zl\u0131<\/td>\n<\/tr>\n
    Ayr\u0131\u015ft\u0131rma (Parsing) Ba\u015far\u0131s\u0131<\/strong><\/td>\n%99<\/td>\n%75 – %85<\/td>\n%98 – %100<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n

    Tabloda g\u00f6r\u00fcld\u00fc\u011f\u00fc \u00fczere Arama\/De\u011fi\u015ftirme bloklar\u0131, modelin mevcut koddaki bo\u015fluklar\u0131 veya tab karakterlerini yanl\u0131\u015f tahmin etmesi durumunda tamamen ba\u015far\u0131s\u0131z olur. \u00d6rne\u011fin Python veya YAML gibi dillerde tek bir bo\u015fluk karakterinin eksik \u00fcretilmesi, arama blo\u011funun kaynak dosyada bulunamamas\u0131na neden olur. Oysa Echo the Marker tekni\u011finde, de\u011fi\u015ftirilecek blo\u011fun hemen \u00f6ncesindeki imza yank\u0131land\u0131\u011f\u0131 i\u00e7in istemci taraf\u0131ndaki ayr\u0131\u015ft\u0131r\u0131c\u0131 (parser) yerini an\u0131nda tespit eder ve d\u00fczenlemeyi g\u00fcvenle uygular.<\/p>\n

    Echo the Marker Y\u00f6ntemi Ad\u0131m Ad\u0131m Nas\u0131l Uygulan\u0131r?<\/h3>\n

    Bu tekni\u011fi kendi yerel geli\u015ftirme s\u00fcre\u00e7lerinize veya \u015firket i\u00e7i yapay zeka ara\u00e7lar\u0131n\u0131za entegre etmek olduk\u00e7a basittir. Ancak ba\u015far\u0131l\u0131 bir entegrasyon i\u00e7in talimat setinin (prompt) ve istemci ayr\u0131\u015ft\u0131r\u0131c\u0131s\u0131n\u0131n kusursuz kurgulanmas\u0131 gerekir. \u015eimdi bu y\u00f6ntemi ad\u0131m ad\u0131m hayata ge\u00e7irelim.<\/p>\n

    Ad\u0131m 1: Benzersiz \u0130\u015faret\u00e7inin Belirlenmesi<\/h4>\n

    \u0130lk olarak, d\u00fczenleme yap\u0131lacak alan i\u00e7in sistemde benzersiz olan bir \u00e7apa se\u00e7melisiniz. Bu, mevcut bir fonksiyon imzas\u0131 olabilir veya ge\u00e7ici olarak koda enjekte edilen \u00f6zel bir yorum sat\u0131r\u0131 olabilir. \u00d6rne\u011fin: \/\/ [TARGET: calculate_installment_v2]<\/code> gibi bir belirte\u00e7, dosya i\u00e7inde yaln\u0131zca tek bir yerde ge\u00e7er.<\/p>\n

    Ad\u0131m 2: Sistem Talimat\u0131n\u0131n Haz\u0131rlanmas\u0131<\/h4>\n

    Modelin kesin kurallara uymas\u0131n\u0131 sa\u011flamak i\u00e7in kat\u0131 bir sistem istemi (system prompt) kullanman\u0131z \u015fartt\u0131r. Modelin fazladan selamlama veya a\u00e7\u0131klama yapmas\u0131n\u0131 engellemelisiniz. \u0130steminiz \u015fu net kural\u0131 i\u00e7ermelidir: “Herhangi bir d\u00fczenleme yapmadan \u00f6nce, hedef i\u015faret\u00e7iyi tam olarak yank\u0131la. Asla i\u015faret\u00e7inin i\u00e7eri\u011fini veya girintisini de\u011fi\u015ftirme.”<\/p>\n

    Ad\u0131m 3: \u0130stemci Taraf\u0131nda Ayr\u0131\u015ft\u0131rma (Client-side Replacement)<\/h4>\n

    Modelden gelen ak\u0131\u015f (stream) verisi okunurken, istemci kod yank\u0131lanan i\u015faret\u00e7iyi g\u00f6rd\u00fc\u011f\u00fc anda tetiklenir. \u0130\u015faret\u00e7iden sonraki ilk blok yeni kod olarak alg\u0131lan\u0131r ve dosyadaki ilgili alan g\u00fcncellenir.<\/p>\n

    A\u015fa\u011f\u0131da, T\u00fcrkiye’deki pop\u00fcler bir e-ticaret altyap\u0131s\u0131nda sepet tutar\u0131 ve KDV hesaplama mant\u0131\u011f\u0131n\u0131 g\u00fcncelleyen pratik bir TypeScript\/Node.js senaryosu \u00fczerinden bu y\u00f6ntemi somutla\u015ft\u0131ral\u0131m.<\/p>\n

    \n
    \/\/ Kaynak Dosya: basketService.ts\nexport class BasketService {\n  private taxRate: number = 0.20;\n\n  \/\/ MARKER_START: calculate_basket_total\n  public calculateTotal(items: Array<{ price: number; quantity: number }>): number {\n    let subtotal = 0;\n    for (const item of items) {\n      subtotal += item.price * item.quantity;\n    }\n    return subtotal;\n  }\n  \/\/ MARKER_END: calculate_basket_total\n\n  public validateBasket(): boolean {\n    return true;\n  }\n}<\/code><\/pre>\n<\/div>\n
    Yukar\u0131daki kodda KDV oran\u0131 hesaba kat\u0131lmam\u0131\u015ft\u0131r ve m\u00fc\u015fteri sepetine indirim kuponu deste\u011fi eklemek istiyoruz. Modele g\u00f6nderece\u011fimiz optimize edilmi\u015f y\u00f6nlendirici istem (prompt) \u015fu \u015fekilde yap\u0131land\u0131r\u0131lmal\u0131d\u0131r:<\/p>\n
    \n
    G\u00d6REV: 'calculateTotal' metoduna %20 KDV hesaplamas\u0131n\u0131 ve kupon indirimini dahil et.\nKURAL: Kod \u00fcretmeden \u00f6nce kesinlikle '\/\/ MARKER_START: calculate_basket_total' i\u015faret\u00e7isini birebir yank\u0131la.\nYan\u0131t format\u0131:\n<<>><\/code><\/pre>\n<\/div>\n
    Model bu kat\u0131 kural\u0131 takip etti\u011finde, modelin \u00fcretece\u011fi \u00e7\u0131kt\u0131 \u015fu netlikte olur:<\/p>\n
    \n
    <<, \n    discountAmount: number = 0\n  ): number {\n    const subtotal = items.reduce((acc, item) => acc + (item.price * item.quantity), 0);\n    const discounted = Math.max(0, subtotal - discountAmount);\n    const totalWithTax = discounted * (1 + this.taxRate);\n    return Number(totalWithTax.toFixed(2));\n  }\n>>><\/code><\/pre>\n<\/div>\n
    G\u00f6rd\u00fc\u011f\u00fcn\u00fcz gibi model, i\u015faret\u00e7iyi yank\u0131layarak i\u015fe ba\u015flam\u0131\u015f, bu sayede ba\u011flamdan sapmadan sadece hedef fonksiyonu revize etmi\u015ftir. Yerel Python veya Node.js yama arac\u0131n\u0131z bu metni ayr\u0131\u015ft\u0131rd\u0131\u011f\u0131nda dosyadaki MARKER_START<\/code> ve MARKER_END<\/code> aras\u0131ndaki k\u0131sm\u0131 tek bir regex aramas\u0131yla milisaniyeler i\u00e7inde g\u00fcnceller.<\/p>\n
    Ger\u00e7ek D\u00fcnya Vaka Analizi: Yerel Bir FinTech Giri\u015fiminde S\u00fcre\u00e7 \u0130yile\u015ftirmesi<\/h3>\n
    Konunun de\u011ferini daha iyi kavramak ad\u0131na, T\u00fcrkiye’de Bankac\u0131l\u0131k D\u00fczenleme ve Denetleme Kurumu (BDDK) uyumlu \u00f6deme sistemleri geli\u015ftiren bir yerel FinTech giri\u015fiminin ya\u015fad\u0131\u011f\u0131 ger\u00e7ek bir d\u00f6n\u00fc\u015f\u00fcm senaryosunu ele alal\u0131m. \u015eirket, muhasebe mutabakat\u0131 ve sanal POS entegrasyonlar\u0131n\u0131 i\u00e7eren yakla\u015f\u0131k 12.000 sat\u0131rl\u0131k devasa bir monolitik servise sahipti.<\/p>\n
    Geli\u015ftirici ekibi, geli\u015ftirme h\u0131z\u0131n\u0131 art\u0131rmak amac\u0131yla a\u00e7\u0131k kaynakl\u0131 LLM’leri \u015firket i\u00e7i sunucular\u0131nda \u00e7al\u0131\u015ft\u0131rmaya ba\u015flad\u0131. Ancak ilk denemelerde ciddi bir darbo\u011fazla kar\u015f\u0131la\u015ft\u0131lar. Model, arama-de\u011fi\u015ftirme bloklar\u0131 \u00fcretirken T\u00fcrkiye Cumhuriyet Merkez Bankas\u0131 (TCMB) FAST entegrasyonuyla ilgili fonksiyonlardaki benzer loglama sat\u0131rlar\u0131n\u0131 kar\u0131\u015ft\u0131r\u0131yordu. Modelin \u00fcretti\u011fi yamalar %38 oran\u0131nda yanl\u0131\u015f fonksiyonu eziyor veya girinti hatalar\u0131 y\u00fcz\u00fcnden s\u00f6zdizimi (syntax) hatalar\u0131 olu\u015fturuyordu. Bu durum, geli\u015ftiricilerin yapay zekaya olan g\u00fcvenini tamamen sarsm\u0131\u015ft\u0131.<\/p>\n
    M\u00fchendislik ekibi sorunu \u00e7\u00f6zmek ad\u0131na Echo the Marker Before You Edit<\/strong> mimarisini dahili VS Code eklentilerine entegre etti. Sistem, d\u00fczenlenmek istenen fonksiyonun AST (Soyut S\u00f6zdizim A\u011fac\u0131 – Abstract Syntax Tree) d\u00fc\u011f\u00fcm\u00fcn\u00fc analiz ederek fonksiyonun hemen ba\u015f\u0131na dinamik ve ge\u00e7ici bir benzersiz GUID i\u015faret\u00e7isi ekliyordu. Modele g\u00f6nderilen talimatta \u015fu \u015fart ko\u015fuldu: “GUID i\u015faret\u00e7isini aynen yank\u0131la ve ard\u0131ndan g\u00fcncellenmi\u015f d\u00fc\u011f\u00fcm\u00fc \u00fcret.”<\/p>\n
    Sonu\u00e7lar son derece \u00e7arp\u0131c\u0131yd\u0131:<\/p>\n
      \n
    • Yama Ba\u015far\u0131 Oran\u0131:<\/strong> Yama uygulama ba\u015far\u0131 oran\u0131 %62 seviyesinden %99.4 seviyesine f\u0131rlad\u0131.<\/li>\n
    • Token Verimlili\u011fi:<\/strong> T\u00fcm dosyay\u0131 modele g\u00f6ndermek veya uzun arama bloklar\u0131 olu\u015fturmak gerekmedi\u011fi i\u00e7in token t\u00fcketimi %45 azald\u0131.<\/li>\n
    • Hata Ay\u0131klama S\u00fcresi:<\/strong> Geli\u015ftiricilerin model \u00e7\u0131kt\u0131s\u0131n\u0131 elle d\u00fczeltmek i\u00e7in harcad\u0131\u011f\u0131 g\u00fcnl\u00fck ortalama s\u00fcre 45 dakikadan 4 dakikaya geriledi.<\/li>\n<\/ul>\n
      Bu vaka analizi, do\u011fru y\u00f6nlendirme mimarisi kuruldu\u011funda yapay zekan\u0131n sadece basit betiklerde de\u011fil, reg\u00fclasyona tabi kritik kurumsal sistemlerde bile g\u00fcvenle kullan\u0131labilece\u011fini kan\u0131tlamaktad\u0131r.<\/p>\n
      \u0130leri D\u00fczey \u0130pu\u00e7lar\u0131 ve Mimari \u0130nce Ayarlar<\/h3>\n
      Echo the Marker tekni\u011fini temel seviyede kavramak kolayd\u0131r; fakat kurumsal \u00f6l\u00e7ekte kullan\u0131rken dikkat edilmesi gereken baz\u0131 kritik p\u00fcf noktalar\u0131 bulunmaktad\u0131r. Bu incelikler, sisteminizin en karma\u015f\u0131k u\u00e7 senaryolarda bile kararl\u0131 \u00e7al\u0131\u015fmas\u0131n\u0131 sa\u011flar.<\/p>\n
      1. Dinamik AST Tabanl\u0131 \u00c7apalar Kullan\u0131n<\/h4>\n
      Koda kal\u0131c\u0131 olarak yorum sat\u0131r\u0131 \u015feklinde i\u015faret\u00e7i eklemek temiz kod (clean code) prensiplerine ayk\u0131r\u0131 olabilir. Bunun yerine, istemci arac\u0131n\u0131z dosyadaki AST d\u00fc\u011f\u00fcm\u00fcn\u00fc okuyarak hedef sat\u0131r\u0131n \u00fczerine ge\u00e7ici bellek-i\u00e7i (in-memory) bir i\u015faret\u00e7i yerle\u015ftirmelidir. Kod modele iletilirken bu i\u015faret\u00e7i eklenir, model i\u015faret\u00e7iyi yank\u0131lar ve yerel istemci yamay\u0131 uygulad\u0131ktan sonra ge\u00e7ici i\u015faret\u00e7iyi temizler. B\u00f6ylece kod taban\u0131n\u0131zda gereksiz kal\u0131nt\u0131 kalmaz.<\/p>\n
      2. Ak\u0131\u015f (Streaming) S\u0131ras\u0131nda Erken Do\u011frulama Yap\u0131n<\/h4>\n
      LLM \u00e7\u0131kt\u0131lar\u0131 genellikle kullan\u0131c\u0131ya ak\u0131\u015f halinde gelir. Model ilk belirte\u00e7leri \u00fcretirken i\u015faret\u00e7iyi tam ve eksiksiz yank\u0131lamazsa, \u00fcretimi hemen iptal edebilirsiniz (abort controller kullanarak). \u00d6rne\u011fin, model ilk 50 belirte\u00e7 i\u00e7inde hedef i\u015faret\u00e7iyi yazmad\u0131ysa, hal\u00fcsinasyon g\u00f6rd\u00fc\u011f\u00fcn\u00fc anlay\u0131p i\u015flemi an\u0131nda kesebilir ve maliyet olu\u015fmas\u0131n\u0131 engelleyebilirsiniz.<\/p>\n
      3. \u00c7oklu De\u011fi\u015fikliklerde Ters \u00c7evrilmi\u015f S\u0131ralama (Bottom-Up) Uygulay\u0131n<\/h4>\n
      Ayn\u0131 dosya i\u00e7inde birden fazla fonksiyonu de\u011fi\u015ftirecekseniz, i\u015faret\u00e7ileri dosyan\u0131n sonundan ba\u015f\u0131na do\u011fru (a\u015fa\u011f\u0131dan yukar\u0131ya) yank\u0131lat\u0131p d\u00fczenleyin. \u00c7\u00fcnk\u00fc \u00fcst k\u0131s\u0131mlarda yap\u0131lan sat\u0131r ekleme ve \u00e7\u0131karma i\u015flemleri, alt k\u0131s\u0131mlar\u0131n sat\u0131r ofsetlerini kayd\u0131rabilir. Dosyan\u0131n alt\u0131ndan ba\u015flad\u0131\u011f\u0131n\u0131zda \u00fcstteki koordinatlar bozulmadan kal\u0131r.<\/p>\n
      S\u0131k\u00e7a Sorulan Sorular<\/h3>\n
      Echo the Marker tekni\u011fi sadece belirli yapay zeka modelleriyle mi \u00e7al\u0131\u015f\u0131r?<\/h4>\n
      Hay\u0131r, bu yakla\u015f\u0131m modelden ba\u011f\u0131ms\u0131zd\u0131r. GPT-4, Claude 3.5 Sonnet, DeepSeek-Coder veya yerel \u00e7al\u0131\u015fan Llama modelleri fark etmeksizin t\u00fcm oto-regresif dil modellerinde ayn\u0131 y\u00fcksek ba\u015far\u0131y\u0131 g\u00f6sterir; \u00e7\u00fcnk\u00fc y\u00f6ntem transformat\u00f6r mimarisinin temel dikkat prensibinden faydalan\u0131r.<\/p>\n
      Koddaki mevcut fonksiyon isimleri i\u015faret\u00e7i olarak kullan\u0131lamaz m\u0131?<\/h4>\n
      E\u011fer fonksiyon ismi o dosya i\u00e7erisinde kesinlikle benzersiz ise do\u011frudan i\u015faret\u00e7i olarak kullan\u0131labilir. Ancak b\u00fcy\u00fck dosyalarda benzer a\u015f\u0131r\u0131 y\u00fcklenmi\u015f (overloaded) metotlar veya ayn\u0131 isimli de\u011fi\u015fkenler bulunabilece\u011finden, benzersiz bir \u00f6n ek veya yorum etiketi kullanmak her zaman daha g\u00fcvenli bir sonu\u00e7 verir.<\/p>\n
      Arama-De\u011fi\u015ftirme (Search & Replace) bloklar\u0131na g\u00f6re temel avantaj\u0131 nedir?<\/h4>\n
      Arama-De\u011fi\u015ftirme y\u00f6nteminde modelin eski kodun geni\u015f bir blo\u011funu harfi harfine hat\u0131rlamas\u0131 gerekir. Model bu s\u0131rada tek bir girinti veya noktal\u0131 virg\u00fcl hatas\u0131 yaparsa e\u015fleme motoru \u00e7\u00f6ker. Echo the Marker y\u00f6nteminde ise sadece tek bir k\u0131sa imza yank\u0131land\u0131\u011f\u0131 i\u00e7in hata pay\u0131 neredeyse s\u0131f\u0131ra iner.<\/p>\n
      Bu y\u00f6ntem token maliyetlerini nas\u0131l etkiler?<\/h4>\n
      T\u00fcm dosyan\u0131n ba\u015ftan yaz\u0131lmas\u0131na k\u0131yasla token maliyetlerini devasa oranda d\u00fc\u015f\u00fcr\u00fcr. Arama-de\u011fi\u015ftirme bloklar\u0131yla k\u0131yasland\u0131\u011f\u0131nda ise arama blo\u011fundaki gereksiz ba\u011flam tekrarlar\u0131n\u0131 ortadan kald\u0131rd\u0131\u011f\u0131 i\u00e7in ortalama %15 ila %30 oran\u0131nda belirte\u00e7 tasarrufu sa\u011flar.<\/p>\n
      Sonu\u00e7 olarak, b\u00fcy\u00fck dil modellerini yaz\u0131l\u0131m geli\u015ftirme d\u00f6ng\u00fcs\u00fcne entegre ederken deterministik kontrol mekanizmalar\u0131 kurmak zorunday\u0131z. Echo the Marker Before You Edit tekni\u011fi, modelin yarat\u0131c\u0131 zekas\u0131 ile yaz\u0131l\u0131m m\u00fchendisli\u011finin kat\u0131 kurallar\u0131 aras\u0131nda m\u00fckemmel bir k\u00f6pr\u00fc kurarak hatas\u0131z, h\u0131zl\u0131 ve g\u00fcvenilir bir kod \u00fcretim deneyimi sunar.<\/p>\n
      #YapayZeka #Kodlama #PromptEngineering #WebGeli\u015ftirme #Yaz\u0131l\u0131mMimarisi<\/p>\n","protected":false},"excerpt":{"rendered":"Echo the Marker Before You Edit tekni\u011fi, b\u00fcy\u00fck dil modellerinin kod d\u00fczenleme yaparken ba\u011flam\u0131 kaybetmesini \u00f6nleyen ve kusursuz yama \u00fcreten modern bir y\u00f6ntemdir.","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-45140","1":"post","2":"type-post","3":"status-publish","4":"format-standard","6":"category-genel","7":"cs-entry","8":"cs-video-wrap"},"yoast_head":"\nEcho the Marker Before You Edit Tekni\u011fi ile Hatas\u0131z Kod - 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\/echo-the-marker-before-you-edit-teknigi-ile-hatasiz-kod\/\" \/>\n<meta property=\"og:locale\" content=\"tr_TR\" \/>\n<meta property=\"og:type\" content=\"article\" \/>\n<meta property=\"og:title\" content=\"Echo the Marker Before You Edit Tekni\u011fi ile Hatas\u0131z Kod\" \/>\n<meta property=\"og:description\" content=\"Echo the Marker Before You Edit tekni\u011fi, b\u00fcy\u00fck dil modellerinin kod d\u00fczenleme yaparken ba\u011flam\u0131 kaybetmesini \u00f6nleyen ve kusursuz yama \u00fcreten modern bir y\u00f6ntemdir.\" \/>\n<meta property=\"og:url\" content=\"https:\/\/fatihsoysal.com\/blog\/echo-the-marker-before-you-edit-teknigi-ile-hatasiz-kod\/\" \/>\n<meta property=\"og:site_name\" content=\"Kodlar\u0131n Gizemli D\u00fcnyas\u0131\" \/>\n<meta property=\"article:published_time\" content=\"2026-10-10T18:11:51+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=\"13 dakika\" \/>\n<script type=\"application\/ld+json\" class=\"yoast-schema-graph\">{\"@context\":\"https:\/\/schema.org\",\"@graph\":[{\"@type\":\"Article\",\"@id\":\"https:\/\/fatihsoysal.com\/blog\/echo-the-marker-before-you-edit-teknigi-ile-hatasiz-kod\/#article\",\"isPartOf\":{\"@id\":\"https:\/\/fatihsoysal.com\/blog\/echo-the-marker-before-you-edit-teknigi-ile-hatasiz-kod\/\"},\"author\":{\"name\":\"Fatih Soysal\",\"@id\":\"https:\/\/fatihsoysal.com\/blog\/#\/schema\/person\/002a254750921dcfd568a99e48240dd1\"},\"headline\":\"Echo the Marker Before You Edit Tekni\u011fi ile Hatas\u0131z Kod\",\"datePublished\":\"2026-10-10T18:11:51+00:00\",\"mainEntityOfPage\":{\"@id\":\"https:\/\/fatihsoysal.com\/blog\/echo-the-marker-before-you-edit-teknigi-ile-hatasiz-kod\/\"},\"wordCount\":2468,\"commentCount\":0,\"publisher\":{\"@id\":\"https:\/\/fatihsoysal.com\/blog\/#\/schema\/person\/002a254750921dcfd568a99e48240dd1\"},\"inLanguage\":\"tr\",\"potentialAction\":[{\"@type\":\"CommentAction\",\"name\":\"Comment\",\"target\":[\"https:\/\/fatihsoysal.com\/blog\/echo-the-marker-before-you-edit-teknigi-ile-hatasiz-kod\/#respond\"]}],\"copyrightYear\":\"2026\",\"copyrightHolder\":{\"@id\":\"https:\/\/fatihsoysal.com\/blog\/#organization\"}},{\"@type\":\"WebPage\",\"@id\":\"https:\/\/fatihsoysal.com\/blog\/echo-the-marker-before-you-edit-teknigi-ile-hatasiz-kod\/\",\"url\":\"https:\/\/fatihsoysal.com\/blog\/echo-the-marker-before-you-edit-teknigi-ile-hatasiz-kod\/\",\"name\":\"Echo the Marker Before You Edit Tekni\u011fi ile Hatas\u0131z Kod - Kodlar\u0131n Gizemli D\u00fcnyas\u0131\",\"isPartOf\":{\"@id\":\"https:\/\/fatihsoysal.com\/blog\/#website\"},\"datePublished\":\"2026-10-10T18:11:51+00:00\",\"breadcrumb\":{\"@id\":\"https:\/\/fatihsoysal.com\/blog\/echo-the-marker-before-you-edit-teknigi-ile-hatasiz-kod\/#breadcrumb\"},\"inLanguage\":\"tr\",\"potentialAction\":[{\"@type\":\"ReadAction\",\"target\":[\"https:\/\/fatihsoysal.com\/blog\/echo-the-marker-before-you-edit-teknigi-ile-hatasiz-kod\/\"]}]},{\"@type\":\"BreadcrumbList\",\"@id\":\"https:\/\/fatihsoysal.com\/blog\/echo-the-marker-before-you-edit-teknigi-ile-hatasiz-kod\/#breadcrumb\",\"itemListElement\":[{\"@type\":\"ListItem\",\"position\":1,\"name\":\"Anasayfa\",\"item\":\"https:\/\/fatihsoysal.com\/blog\/\"},{\"@type\":\"ListItem\",\"position\":2,\"name\":\"Echo the Marker Before You Edit Tekni\u011fi ile Hatas\u0131z Kod\"}]},{\"@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":"Echo the Marker Before You Edit Tekni\u011fi ile Hatas\u0131z Kod - 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\/echo-the-marker-before-you-edit-teknigi-ile-hatasiz-kod\/","og_locale":"tr_TR","og_type":"article","og_title":"Echo the Marker Before You Edit Tekni\u011fi ile Hatas\u0131z Kod","og_description":"Echo the Marker Before You Edit tekni\u011fi, b\u00fcy\u00fck dil modellerinin kod d\u00fczenleme yaparken ba\u011flam\u0131 kaybetmesini \u00f6nleyen ve kusursuz yama \u00fcreten modern bir y\u00f6ntemdir.","og_url":"https:\/\/fatihsoysal.com\/blog\/echo-the-marker-before-you-edit-teknigi-ile-hatasiz-kod\/","og_site_name":"Kodlar\u0131n Gizemli D\u00fcnyas\u0131","article_published_time":"2026-10-10T18:11:51+00:00","author":"Fatih Soysal","twitter_card":"summary_large_image","twitter_misc":{"Yazan:":"Fatih Soysal","Tahmini okuma s\u00fcresi":"13 dakika"},"schema":{"@context":"https:\/\/schema.org","@graph":[{"@type":"Article","@id":"https:\/\/fatihsoysal.com\/blog\/echo-the-marker-before-you-edit-teknigi-ile-hatasiz-kod\/#article","isPartOf":{"@id":"https:\/\/fatihsoysal.com\/blog\/echo-the-marker-before-you-edit-teknigi-ile-hatasiz-kod\/"},"author":{"name":"Fatih Soysal","@id":"https:\/\/fatihsoysal.com\/blog\/#\/schema\/person\/002a254750921dcfd568a99e48240dd1"},"headline":"Echo the Marker Before You Edit Tekni\u011fi ile Hatas\u0131z Kod","datePublished":"2026-10-10T18:11:51+00:00","mainEntityOfPage":{"@id":"https:\/\/fatihsoysal.com\/blog\/echo-the-marker-before-you-edit-teknigi-ile-hatasiz-kod\/"},"wordCount":2468,"commentCount":0,"publisher":{"@id":"https:\/\/fatihsoysal.com\/blog\/#\/schema\/person\/002a254750921dcfd568a99e48240dd1"},"inLanguage":"tr","potentialAction":[{"@type":"CommentAction","name":"Comment","target":["https:\/\/fatihsoysal.com\/blog\/echo-the-marker-before-you-edit-teknigi-ile-hatasiz-kod\/#respond"]}],"copyrightYear":"2026","copyrightHolder":{"@id":"https:\/\/fatihsoysal.com\/blog\/#organization"}},{"@type":"WebPage","@id":"https:\/\/fatihsoysal.com\/blog\/echo-the-marker-before-you-edit-teknigi-ile-hatasiz-kod\/","url":"https:\/\/fatihsoysal.com\/blog\/echo-the-marker-before-you-edit-teknigi-ile-hatasiz-kod\/","name":"Echo the Marker Before You Edit Tekni\u011fi ile Hatas\u0131z Kod - Kodlar\u0131n Gizemli D\u00fcnyas\u0131","isPartOf":{"@id":"https:\/\/fatihsoysal.com\/blog\/#website"},"datePublished":"2026-10-10T18:11:51+00:00","breadcrumb":{"@id":"https:\/\/fatihsoysal.com\/blog\/echo-the-marker-before-you-edit-teknigi-ile-hatasiz-kod\/#breadcrumb"},"inLanguage":"tr","potentialAction":[{"@type":"ReadAction","target":["https:\/\/fatihsoysal.com\/blog\/echo-the-marker-before-you-edit-teknigi-ile-hatasiz-kod\/"]}]},{"@type":"BreadcrumbList","@id":"https:\/\/fatihsoysal.com\/blog\/echo-the-marker-before-you-edit-teknigi-ile-hatasiz-kod\/#breadcrumb","itemListElement":[{"@type":"ListItem","position":1,"name":"Anasayfa","item":"https:\/\/fatihsoysal.com\/blog\/"},{"@type":"ListItem","position":2,"name":"Echo the Marker Before You Edit Tekni\u011fi ile Hatas\u0131z Kod"}]},{"@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\/45140","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=45140"}],"version-history":[{"count":0,"href":"https:\/\/fatihsoysal.com\/blog\/wp-json\/wp\/v2\/posts\/45140\/revisions"}],"wp:attachment":[{"href":"https:\/\/fatihsoysal.com\/blog\/wp-json\/wp\/v2\/media?parent=45140"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/fatihsoysal.com\/blog\/wp-json\/wp\/v2\/categories?post=45140"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/fatihsoysal.com\/blog\/wp-json\/wp\/v2\/tags?post=45140"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}