{"id":11400,"date":"2026-10-04T00:43:59","date_gmt":"2026-10-04T00:43:59","guid":{"rendered":"https:\/\/bitep.net\/blog\/2026\/10\/04\/infoworldcom-2\/"},"modified":"2026-10-04T00:43:59","modified_gmt":"2026-10-04T00:43:59","slug":"infoworldcom-2","status":"publish","type":"post","link":"https:\/\/bitep.net\/blog\/2026\/10\/04\/infoworldcom-2\/","title":{"rendered":"infoworld.com"},"content":{"rendered":"<h2>Avtonom S\u00fcni \u0130ntellekt \u0130\u015f Ax\u0131nlar\u0131: Passiv Cavablardan M\u00fcst\u0259qil \u0130craya<\/h2>\n<p>Generativ s\u00fcni intellekt sisteml\u0259ri sad\u0259 sual-cavab (prompt-response) paradiqmas\u0131ndan \u00e7\u0131xaraq m\u00fcr\u0259kk\u0259b, \u00e7oxm\u0259rh\u0259l\u0259li v\u0259 m\u00fcst\u0259qil q\u0259rarverm\u0259 qabiliyy\u0259tin\u0259 malik &#8220;avtonom i\u015f ax\u0131nlar\u0131&#8221; (autonomous workflows) m\u0259rh\u0259l\u0259sin\u0259 ke\u00e7id edir. M\u00fcasir b\u00f6y\u00fck dil modell\u0259ri (LLM) art\u0131q sad\u0259c\u0259 m\u0259tn generasiya ed\u0259n m\u00fch\u0259rrikl\u0259r deyil; onlar geni\u015f t\u0259limatlar\u0131 q\u0259bul ed\u0259n, onlar\u0131 alt tap\u015f\u0131r\u0131qlara b\u00f6l\u0259n, m\u00fcxt\u0259lif proqram m\u00fchitl\u0259rind\u0259 icra ed\u0259n v\u0259 n\u0259tic\u0259ni insan m\u00fcdaxil\u0259si olmadan yoxlayan r\u0259q\u0259msal agentl\u0259r kimi f\u0259aliyy\u0259t g\u00f6st\u0259rirl\u0259r.<\/p>\n<p>Bu arxitektur d\u0259yi\u015fiklik ma\u015f\u0131n \u00f6yr\u0259nm\u0259si m\u00fch\u0259ndisliyind\u0259 &#8220;Agentic AI&#8221; adlanan yeni yana\u015fman\u0131n \u0259sas\u0131n\u0131 t\u0259\u015fkil edir. \u018fg\u0259r \u0259n\u0259n\u0259vi modell\u0259r h\u0259r add\u0131mda insan t\u0259r\u0259find\u0259n istiqam\u0259tl\u0259ndirilm\u0259y\u0259 ehtiyac duyurdusa, m\u00fcasir modell\u0259r qeyri-m\u00fc\u0259yy\u0259n \u015f\u0259raitd\u0259 strateji planlama apara bilir v\u0259 g\u00f6zl\u0259nilm\u0259z x\u0259talarla qar\u015f\u0131la\u015fd\u0131qda \u00f6z i\u015fini korreksiya ed\u0259 bilir.<\/p>\n<h2>Texniki Arxitektura: M\u0259qs\u0259dd\u0259n N\u0259tic\u0259y\u0259 Do\u011fru<\/h2>\n<p>Avtonom i\u015f ax\u0131nlar\u0131n\u0131n u\u011furla \u00e7al\u0131\u015fmas\u0131 bir ne\u00e7\u0259 kritik komponentin v\u0259hd\u0259tin\u0259 \u0259saslan\u0131r:<\/p>\n<ul>\n<li><strong>M\u0259qs\u0259din Dekompozisiyas\u0131 (Goal Decomposition):<\/strong> Model q\u0259bul etdiyi qlobal h\u0259d\u0259fi ki\u00e7ik, icra oluna bil\u0259n alt tap\u015f\u0131r\u0131qlara (sub-tasks) par\u00e7alay\u0131r. Bu prosesd\u0259 <em>Chain-of-Thought (CoT)<\/em> v\u0259 <em>Tree-of-Thoughts (ToT)<\/em> kimi metodologiyalardan istifad\u0259 edilir.<\/li>\n<li><strong>Al\u0259tl\u0259rl\u0259 \u0130nteqrasiya (Tool Use &#038; Function Calling):<\/strong> Agentl\u0259r API-l\u0259r, veril\u0259nl\u0259r bazalar\u0131, terminal m\u00fchitl\u0259ri v\u0259 brauzerl\u0259rl\u0259 birba\u015fa qar\u015f\u0131l\u0131ql\u0131 \u0259laq\u0259d\u0259 olur. Model m\u00fcvafiq funksiyan\u0131 se\u00e7ir, parametrl\u0259ri generasiya edir v\u0259 k\u0259nar m\u00fchitd\u0259n g\u0259l\u0259n cavab\u0131 n\u00f6vb\u0259ti add\u0131m\u0131n kontekstin\u0259 daxil edir.<\/li>\n<li><strong>\u00d6z\u00fcn\u00fc Yoxlama v\u0259 D\u00fcz\u0259li\u015f (Self-Reflection \/ Critic Loop):<\/strong> \u0130\u015f ba\u015fa \u00e7atd\u0131qdan sonra agent \u00f6z n\u0259tic\u0259sini ilkin meyarlarla m\u00fcqayis\u0259 edir. M\u0259s\u0259l\u0259n, generasiya etdiyi kod vahid testl\u0259rd\u0259n (unit tests) ke\u00e7m\u0259zs\u0259, agent x\u0259ta loqlar\u0131n\u0131 oxuyaraq kodu yenid\u0259n redakt\u0259 edir.<\/li>\n<\/ul>\n<h3>\u018fn\u0259n\u0259vi v\u0259 Avtonom \u0130\u015f Ax\u0131nlar\u0131n\u0131n M\u00fcqayis\u0259si<\/h3>\n<table>\n<thead>\n<tr>\n<th>X\u00fcsusiyy\u0259t<\/th>\n<th>Klassik LLM \u0130\u015f Ax\u0131n\u0131<\/th>\n<th>Avtonom Agent \u0130\u015f Ax\u0131n\u0131<\/th>\n<\/tr>\n<\/thead>\n<tbody>\n<tr>\n<td><strong>\u0130dar\u0259etm\u0259<\/strong><\/td>\n<td>\u0130nsan t\u0259r\u0259find\u0259n add\u0131m-add\u0131m y\u00f6nl\u0259ndirm\u0259<\/td>\n<td>Geni\u015f m\u0259qs\u0259d verilir, icra model\u0259 h\u0259val\u0259 olunur<\/td>\n<\/tr>\n<tr>\n<td><strong>X\u0259talar\u0131n idar\u0259si<\/strong><\/td>\n<td>\u0130stifad\u0259\u00e7i x\u0259tan\u0131 bildirir v\u0259 yeni prompt yaz\u0131r<\/td>\n<td>Model m\u00fchitd\u0259n g\u0259l\u0259n x\u0259tan\u0131 oxuyur v\u0259 \u00f6z\u00fc d\u00fcz\u0259ldir<\/td>\n<\/tr>\n<tr>\n<td><strong>M\u00fchitl\u0259 \u0259laq\u0259<\/strong><\/td>\n<td>T\u0259crid olunmu\u015f m\u0259tn interfeysi<\/td>\n<td>CLI, API, veril\u0259nl\u0259r bazalar\u0131 v\u0259 \u015f\u0259b\u0259k\u0259 al\u0259tl\u0259ri<\/td>\n<\/tr>\n<tr>\n<td><strong>D\u00f6vr say\u0131<\/strong><\/td>\n<td>T\u0259k add\u0131ml\u0131 (Single-turn)<\/td>\n<td>\u00c7ox add\u0131ml\u0131 dinamik d\u00f6vr\u0259l\u0259r (Multi-turn loop)<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<h2>Praktiki T\u0259tbiq: Sad\u0259 Avtonom Agentin Yarad\u0131lmas\u0131<\/h2>\n<p>A\u015fa\u011f\u0131dak\u0131 n\u00fcmun\u0259d\u0259 Python m\u00fchitind\u0259 \u00f6z\u00fcn\u00fc yoxlayan v\u0259 xarici m\u00fchitd\u0259 (terminal) komandalar\u0131 icra ed\u0259n sad\u0259 bir avtonom agent arxitekturas\u0131n\u0131n qurulmas\u0131 add\u0131m-add\u0131m izah edilir.<\/p>\n<h3>Add\u0131m 1: T\u0259l\u0259b Olunan Kitabxanalar\u0131n Qura\u015fd\u0131r\u0131lmas\u0131<\/h3>\n<pre><code class=\"language-bash\">pip install langchain langchain-openai langchain-community<\/code><\/pre>\n<h3>Add\u0131m 2: Al\u0259tl\u0259rin v\u0259 D\u00fc\u015f\u00fcnc\u0259 D\u00f6vr\u0259sinin T\u0259yin Edilm\u0259si<\/h3>\n<p>Agent sistem \u0259mrl\u0259rini icra etm\u0259k \u00fc\u00e7\u00fcn Python <code>subprocess<\/code> modulundan istifad\u0259 ed\u0259n al\u0259tl\u0259 t\u0259chiz edilir v\u0259 n\u0259tic\u0259ni yoxlayan d\u00f6vr\u0259y\u0259 sal\u0131n\u0131r.<\/p>\n<pre><code class=\"language-python\">import os\nimport subprocess\nfrom langchain_openai import ChatOpenAI\nfrom langchain.agents import initialize_agent, Tool, AgentType\n\n# LLM-in inisializasiyas\u0131\nllm = ChatOpenAI(model=\"gpt-4o\", temperature=0)\n\ndef terminal_executor(command: str) -> str:\n    \"\"\"Terminal \u0259mrini icra edir v\u0259 \u00e7\u0131x\u0131\u015f\u0131 qaytar\u0131r.\"\"\"\n    try:\n        result = subprocess.run(\n            command, shell=True, check=True, \n            stdout=subprocess.PIPE, stderr=subprocess.PIPE, text=True\n        )\n        return result.stdout if result.stdout else \"\u018fmr u\u011furla icra olundu (\u00c7\u0131x\u0131\u015f yoxdur).\"\n    except subprocess.CalledProcessError as e:\n        return f\"X\u0259ta ba\u015f verdi: {e.stderr}\"\n\n# Agent \u00fc\u00e7\u00fcn al\u0259tl\u0259r siyah\u0131s\u0131\ntools = [\n    Tool(\n        name=\"Terminal\",\n        func=terminal_executor,\n        description=\"Bax\u0131lan m\u00fchitd\u0259 terminal\/bash \u0259mrl\u0259rini icra etm\u0259k \u00fc\u00e7\u00fcn istifad\u0259 olunur.\"\n    )\n]\n\n# Agentin qurulmas\u0131\nagent = initialize_agent(\n    tools=tools,\n    llm=llm,\n    agent=AgentType.STRUCTURED_CHAT_ZERO_SHOT_REACT_DESCRIPTION,\n    verbose=True\n)\n\n# Geni\u015f h\u0259d\u0259fin agent\u0259 t\u0259qdim edilm\u0259si\ngoal = \"\"\"\nLayih\u0259 qovlu\u011funda 'src' adl\u0131 direktoriyan\u0131n olub-olmad\u0131\u011f\u0131n\u0131 yoxla. \nYoxdursa, onu yarat, daxilind\u0259 'app.py' fayl\u0131 a\u00e7 v\u0259 i\u00e7in\u0259 sad\u0259 'Hello World' yazan kod \u0259lav\u0259 et. \nSonda fayl\u0131n m\u00f6vcudlu\u011funu v\u0259 m\u0259zmununu oxuyaraq i\u015fini t\u0259sdiql\u0259.\n\"\"\"\n\nresponse = agent.run(goal)\nprint(response)\n<\/code><\/pre>\n<p>Bu n\u00fcmun\u0259d\u0259 agent ona veril\u0259n h\u0259d\u0259fi t\u0259k bir sor\u011fu il\u0259 yerin\u0259 yetirmir. O, \u0259vv\u0259lc\u0259 direktoriyan\u0131 yoxlamaq \u00fc\u00e7\u00fcn terminal \u0259mrini icra edir, n\u0259tic\u0259ni oxuyur, ard\u0131nca fayl\u0131 yarad\u0131r v\u0259 \u0259n son m\u0259rh\u0259l\u0259d\u0259 fayl\u0131n m\u0259zmununu oxuyaraq \u00f6z i\u015fini yoxlay\u0131r (Self-Check).<\/p>\n<h2>S\u0259naye T\u0259sirl\u0259ri: Proqram T\u0259minat\u0131 M\u00fch\u0259ndisliyi v\u0259 T\u0259hl\u00fck\u0259sizlik<\/h2>\n<p>Avtonom i\u015f ax\u0131nlar\u0131n\u0131n inki\u015faf\u0131 DevOps, kibert\u0259hl\u00fck\u0259sizlik v\u0259 proqram t\u0259minat\u0131n\u0131n keyfiyy\u0259t t\u0259minat\u0131 (QA) sah\u0259l\u0259rin\u0259 d\u0259rind\u0259n t\u0259sir g\u00f6st\u0259rir:<\/p>\n<ul>\n<li><strong>Avtonom Kod T\u0259miri (Self-Healing Code):<\/strong> CI\/CD boru k\u0259m\u0259rl\u0259rind\u0259 x\u0259ta ver\u0259n testl\u0259ri t\u0259hlil ed\u0259n agentl\u0259r avtomatik olaraq Pull Request (PR) haz\u0131rlaya, as\u0131l\u0131l\u0131qlar\u0131 yenil\u0259y\u0259 v\u0259 z\u0259iflikl\u0259ri aradan qald\u0131ra bil\u0259r.<\/li>\n<li><strong>SRE v\u0259 \u0130nfrastruktur \u0130dar\u0259etm\u0259si:<\/strong> Serverl\u0259rd\u0259 ba\u015f ver\u0259n anomaliyalar zaman\u0131 agent loqlar\u0131 oxuyaraq ilkin diaqnostikan\u0131 apara v\u0259 z\u0259rur\u0259t olduqda resurslar\u0131 yenid\u0259n ba\u015fladaraq insidenti h\u0259ll ed\u0259 bil\u0259r.<\/li>\n<li><strong>T\u0259hl\u00fck\u0259sizlik Riskleri (Sandboxing Z\u0259rur\u0259ti):<\/strong> Avtonom agentl\u0259r\u0259 sistem m\u00fchitl\u0259rin\u0259 birba\u015fa \u00e7\u0131x\u0131\u015f icaz\u0259sinin verilm\u0259si t\u0259hl\u00fck\u0259sizlik riskl\u0259rini art\u0131r\u0131r. Modelin q\u0259r\u0259zli v\u0259 ya s\u0259hv q\u0259rarlar ver\u0259r\u0259k fayllar\u0131 silm\u0259sinin v\u0259 ya sistem parametrl\u0259rini pozmas\u0131n\u0131n qar\u015f\u0131s\u0131n\u0131 almaq \u00fc\u00e7\u00fcn bu agentl\u0259rin qapal\u0131 <em>Docker<\/em> konteynerl\u0259rind\u0259 v\u0259 ya <em>WebAssembly (WASM)<\/em> m\u00fchitl\u0259rind\u0259 (&#8220;sandbox&#8221;) i\u015fl\u0259dilm\u0259si m\u0259cburidir.<\/li>\n<\/ul>\n<h2>BITEP Redaksiyas\u0131n\u0131n Yekun R\u0259yi<\/h2>\n<p>S\u00fcni intellektin passiv k\u00f6m\u0259k\u00e7id\u0259n avtonom agent\u0259 \u00e7evrilm\u0259si texnoloji transformasiyan\u0131n qa\u00e7\u0131lmaz m\u0259rh\u0259l\u0259sidir. Proqram t\u0259minat\u0131 m\u00fch\u0259ndisl\u0259ri art\u0131q sad\u0259c\u0259 kod yazmaqla deyil, agentl\u0259rin i\u015fini t\u0259nziml\u0259y\u0259n arxitekturalar v\u0259 t\u0259hl\u00fck\u0259sizlik \u00e7\u0259r\u00e7iv\u0259l\u0259ri qurmaqla m\u0259\u015f\u011ful olmal\u0131d\u0131rlar. \u018fn b\u00f6y\u00fck \u00e7a\u011f\u0131r\u0131\u015f bu sisteml\u0259rin deterministik (d\u0259qiq proqnozla\u015fd\u0131r\u0131la bil\u0259n) i\u015fl\u0259m\u0259sini t\u0259min etm\u0259k v\u0259 onlara veril\u0259n s\u0259lahiyy\u0259t \u00e7\u0259r\u00e7iv\u0259l\u0259rini minimum z\u0259rur\u0259t prinsipi (Principle of Least Privilege) il\u0259 m\u0259hdudla\u015fd\u0131rmaqd\u0131r.<\/p>\n","protected":false},"excerpt":{"rendered":"<p>Avtonom S\u00fcni \u0130ntellekt \u0130\u015f Ax\u0131nlar\u0131: Passiv Cavablardan M\u00fcst\u0259qil \u0130craya Generativ s\u00fcni intellekt sisteml\u0259ri sad\u0259 sual-cavab (prompt-response) paradiqmas\u0131ndan \u00e7\u0131xaraq m\u00fcr\u0259kk\u0259b, \u00e7oxm\u0259rh\u0259l\u0259li v\u0259 m\u00fcst\u0259qil q\u0259rarverm\u0259 qabiliyy\u0259tin\u0259 malik &#8220;avtonom i\u015f ax\u0131nlar\u0131&#8221; (autonomous workflows) m\u0259rh\u0259l\u0259sin\u0259 ke\u00e7id edir. M\u00fcasir b\u00f6y\u00fck dil modell\u0259ri (LLM) art\u0131q sad\u0259c\u0259 m\u0259tn generasiya ed\u0259n m\u00fch\u0259rrikl\u0259r deyil; onlar geni\u015f t\u0259limatlar\u0131 q\u0259bul ed\u0259n, onlar\u0131 alt tap\u015f\u0131r\u0131qlara b\u00f6l\u0259n, m\u00fcxt\u0259lif [&hellip;]<\/p>\n","protected":false},"author":1,"featured_media":11399,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"om_disable_all_campaigns":false,"_monsterinsights_skip_tracking":false,"_jetpack_memberships_contains_paid_content":false,"footnotes":""},"categories":[1],"tags":[],"class_list":["post-11400","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-uncategorized"],"aioseo_notices":[],"jetpack_featured_media_url":"https:\/\/bitep.net\/blog\/wp-content\/uploads\/2026\/10\/bitep-ai-1791074634555-20.jpg","jetpack_sharing_enabled":true,"_links":{"self":[{"href":"https:\/\/bitep.net\/blog\/wp-json\/wp\/v2\/posts\/11400","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/bitep.net\/blog\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/bitep.net\/blog\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/bitep.net\/blog\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/bitep.net\/blog\/wp-json\/wp\/v2\/comments?post=11400"}],"version-history":[{"count":0,"href":"https:\/\/bitep.net\/blog\/wp-json\/wp\/v2\/posts\/11400\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/bitep.net\/blog\/wp-json\/wp\/v2\/media\/11399"}],"wp:attachment":[{"href":"https:\/\/bitep.net\/blog\/wp-json\/wp\/v2\/media?parent=11400"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/bitep.net\/blog\/wp-json\/wp\/v2\/categories?post=11400"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/bitep.net\/blog\/wp-json\/wp\/v2\/tags?post=11400"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}