{"id":5540,"date":"2025-04-30T04:46:16","date_gmt":"2025-04-29T20:46:16","guid":{"rendered":"https:\/\/infernews.com\/?page_id=5540"},"modified":"2026-09-14T09:34:31","modified_gmt":"2026-09-14T01:34:31","slug":"agentic","status":"publish","type":"page","link":"https:\/\/infernews.com\/blog\/agentic\/","title":{"rendered":"Agentic AI"},"content":{"rendered":"\n<figure class=\"wp-block-image alignright size-full is-resized\"><img data-dominant-color=\"d4d9dd\" data-has-transparency=\"false\" loading=\"lazy\" decoding=\"async\" width=\"1024\" height=\"572\" src=\"https:\/\/infernews.com\/blog\/wp-content\/uploads\/2026\/08\/image-6.png\" alt=\"\" class=\"wp-image-11645 not-transparent\" style=\"--dominant-color: #d4d9dd; aspect-ratio:1.7902285398135933;width:365px;height:auto\" srcset=\"https:\/\/infernews.com\/blog\/wp-content\/uploads\/2026\/08\/image-6.png 1024w, https:\/\/infernews.com\/blog\/wp-content\/uploads\/2026\/08\/image-6-300x168.png 300w, https:\/\/infernews.com\/blog\/wp-content\/uploads\/2026\/08\/image-6-768x429.png 768w\" sizes=\"auto, (max-width: 1024px) 100vw, 1024px\" \/><\/figure>\n\n\n\n<article class=\"ai-agent-guide\">\n<h1>AI Agent \u5be6\u6230\u6307\u5357\uff1a\u5f9e LLM \u5230 Agent<\/h1>\n<p><strong>\u5c0e\u8b80\uff1a<\/strong>\u4e0d\u8981\u53ea\u505c\u7559\u5728\u63d0\u793a\u8a5e\u5de5\u7a0b\uff08Prompt Engineering\uff09\uff01AI Agent\uff08\u4eba\u5de5\u667a\u6167\u4ee3\u7406\uff09\u662f\u5177\u5099<strong>\u300c\u76ee\u6a19\u62c6\u89e3\u3001\u74b0\u5883\u611f\u77e5\u3001\u5de5\u5177\u8abf\u7528\u3001\u8a18\u61b6\u6aa2\u7d22\u8207\u81ea\u4e3b\u6c7a\u7b56\u300d<\/strong>\u7684\u667a\u6167\u7cfb\u7d71\u3002<\/p>\n<hr \/>\n<section>\n<h2>\u4e00\u3001\u4ec0\u9ebc\u662f AI Agent\uff1f\u5b83\u8207\u50b3\u7d71 Chatbot \u6709\u4f55\u672c\u8cea\u5dee\u7570\uff1f<\/h2>\n<p>\u50b3\u7d71\u7684\u5927\u8a9e\u8a00\u6a21\u578b\uff08LLM\uff09\u5c31\u50cf\u4e00\u500b\u300c\u96e2\u7dda\u7684\u5927\u8166\u300d\uff0c\u53ea\u80fd\u88ab\u52d5\u6839\u64da\u8f38\u5165\u7684 Prompt \u7522\u51fa\u6587\u5b57\u3002\u800c <strong>AI Agent<\/strong> \u662f\u4ee5 LLM \u70ba\u6c7a\u7b56\u6838\u5fc3\uff0c\u70ba\u5176\u914d\u5099\u4e86<strong>\u8a18\u61b6\uff08Memory\uff09<\/strong>\u3001<strong>\u611f\u5b98\u8207\u5de5\u5177\uff08Tools\/APIs\uff09<\/strong>\u4ee5\u53ca<strong>\u898f\u5283\u80fd\u529b\uff08Planning\uff09<\/strong>\u7684\u81ea\u4e3b\u5be6\u9ad4\u3002<\/p>\n<table style=\"width: 100%;\" border=\"1\" width=\"100%\" cellspacing=\"0\" cellpadding=\"8\">\n<thead>\n<tr>\n<th style=\"width: 15.5878%;\">\u8a55\u4f30\u7dad\u5ea6<\/th>\n<th style=\"width: 29.7226%;\">\u50b3\u7d71 LLM \/ Chatbot<\/th>\n<th style=\"width: 54.4254%;\">\u81ea\u4e3b AI Agent<\/th>\n<\/tr>\n<\/thead>\n<tbody>\n<tr>\n<td style=\"width: 15.5878%;\"><strong>\u4e92\u52d5\u6a21\u5f0f<\/strong><\/td>\n<td style=\"width: 29.7226%;\">\u55ae\u6b21\u8f38\u5165\u8f38\u51fa\uff08One-shot Request &amp; Response\uff09<\/td>\n<td style=\"width: 54.4254%;\">\u6301\u7e8c\u904b\u884c\u7684\u81ea\u4e3b\u601d\u8003\u8ff4\u5708\uff08ReAct Loop\uff09\u76f4\u5230\u76ee\u6a19\u9054\u6210<\/td>\n<\/tr>\n<tr>\n<td style=\"width: 15.5878%;\"><strong>\u5916\u90e8\u884c\u52d5\u529b<\/strong><\/td>\n<td style=\"width: 29.7226%;\">\u7121\uff08\u50c5\u80fd\u8f38\u51fa\u6587\u5b57\u5efa\u8b70\uff09<\/td>\n<td style=\"width: 54.4254%;\">\u5177\u5099\u5de5\u5177\u8abf\u7528\u80fd\u529b\uff08\u53ef\u547c\u53eb API\u3001\u8b80\u5beb\u8cc7\u6599\u5eab\u3001\u57f7\u884c Python \u4ee3\u78bc\uff09<\/td>\n<\/tr>\n<tr>\n<td style=\"width: 15.5878%;\"><strong>\u932f\u8aa4\u8655\u7406<\/strong><\/td>\n<td style=\"width: 29.7226%;\">\u9700\u7531\u4eba\u985e\u624b\u52d5\u6307\u51fa\u932f\u8aa4\u4e26\u91cd\u65b0\u767c\u554f<\/td>\n<td style=\"width: 54.4254%;\">\u81ea\u52d5\u6355\u6349\u74b0\u5883\u932f\u8aa4\u53cd\u994b\uff08Observation\uff09\uff0c\u81ea\u6211\u53cd\u601d\u4e26\u4fee\u6b63\uff08Self-Correction\uff09<\/td>\n<\/tr>\n<tr>\n<td style=\"width: 15.5878%;\"><strong>\u9069\u7528\u5834\u666f<\/strong><\/td>\n<td style=\"width: 29.7226%;\">\u6587\u7ae0\u7ffb\u8b6f\u3001\u6587\u5b57\u6458\u8981\u3001\u7c21\u55ae\u554f\u7b54<\/td>\n<td style=\"width: 54.4254%;\">\u7aef\u5230\u7aef\u8edf\u9ad4\u958b\u767c\u3001\u81ea\u4e3b\u5e02\u5834\u8abf\u7814\u3001\u591a\u6b65\u9a5f\u81ea\u52d5\u5316\u4efb\u52d9<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<\/section>\n<hr \/>\n<section>\n<h2>\u4e8c\u3001AI Agent \u7684\u56db\u5927\u6838\u5fc3\u67b6\u69cb\u9aa8\u67b6<\/h2>\n<p>\u73fe\u4ee3 AI Agent \u7cfb\u7d71\u901a\u5e38\u7531\u4ee5\u4e0b\u56db\u500b\u6838\u5fc3\u6a21\u7d44\u69cb\u6210\uff1a<\/p>\n<ol>\n<li>\n<h3>1. \u5927\u8166\u8207\u898f\u5283\u6a21\u7d44 (Planning &amp; Reasoning)<\/h3>\n<p>\u8ca0\u8cac\u5c07\u4f7f\u7528\u8005\u6a21\u7cca\u7684\u9ad8\u968e\u76ee\u6a19\u62c6\u89e3\u70ba\u7d30\u9805\u5b50\u6b65\u9a5f\uff08Sub-goal Decomposition\uff09\u3002\u5e38\u7528\u65b9\u6cd5\u5305\u542b\uff1a<\/p>\n<ul>\n<li><strong>Chain of Thought (CoT)\uff1a<\/strong> \u9010\u6b65\u63a8\u5c0e\u908f\u8f2f\u3002<\/li>\n<li><strong>Plan-and-Solve\uff1a<\/strong> \u5148\u751f\u6210\u6574\u9ad4\u57f7\u884c\u8a08\u756b\uff0c\u518d\u9010\u9805\u59d4\u6d3e\u7d66\u5de5\u5177\u6216\u5b50 Agent\u3002<\/li>\n<li><strong>Reflexion \/ Self-Critique\uff1a<\/strong> \u4efb\u52d9\u5931\u6557\u6642\u81ea\u52d5\u8a55\u4f30\u539f\u56e0\u4e26\u91cd\u8a66\u3002<\/li>\n<\/ul>\n<\/li>\n<li>\n<h3>2. \u8a18\u61b6\u7cfb\u7d71 (Memory Systems)<\/h3>\n<p>\u7a81\u7834\u4e0a\u4e0b\u6587\u8996\u7a97\uff08Context Window\uff09\u9650\u5236\uff0c\u5305\u542b\uff1a<\/p>\n<ul>\n<li><strong>\u77ed\u671f\u5de5\u4f5c\u8a18\u61b6 (Short-term Memory)\uff1a<\/strong> \u7576\u524d\u5c0d\u8a71\u8f2a\u6b21\u8207\u5373\u6642 Context\u3002<\/li>\n<li><strong>\u9577\u671f\u8a18\u61b6 (Long-term Memory)\uff1a<\/strong> \u5229\u7528\u5411\u91cf\u8cc7\u6599\u5eab\uff08Vector DB \/ RAG\uff09\u6216\u77e5\u8b58\u5716\u8b5c\u6aa2\u7d22\u6b77\u53f2\u7d93\u9a57\u8207\u7279\u5b9a\u9818\u57df\u77e5\u8b58\u3002<\/li>\n<\/ul>\n<\/li>\n<li>\n<h3>3. \u5de5\u5177\u8207\u884c\u52d5 (Tools &amp; Actions)<\/h3>\n<p>\u8ce6\u4e88 Agent \u8207\u5916\u754c\u4e92\u52d5\u7684\u80fd\u529b\uff0c\u4f8b\u5982\uff1a<\/p>\n<ul>\n<li><strong>\u641c\u5c0b\u8207\u8cc7\u6599\u6aa2\u7d22\uff1a<\/strong> Google Search, DuckDuckGo API, \u5167\u90e8 API\u3002<\/li>\n<li><strong>\u4ee3\u78bc\u57f7\u884c\u74b0\u5883\uff1a<\/strong> \u9694\u96e2\u7684 Python \u6c99\u76d2\uff08\u5982 Docker \u6216 E2B\uff09\uff0c\u53ef\u9032\u884c\u6578\u5b78\u8a08\u7b97\u8207\u6578\u64da\u8655\u7406\u3002<\/li>\n<li><strong>\u7cfb\u7d71\u64cd\u4f5c\uff1a<\/strong> \u8b80\u5beb\u6a94\u6848\u3001\u767c\u9001 Slack \u8a0a\u606f\u6216\u5bc4\u767c\u96fb\u5b50\u90f5\u4ef6\u3002<\/li>\n<\/ul>\n<\/li>\n<li>\n<h3>4. \u591a\u4ee3\u7406\u5354\u540c\u6a5f\u5236 (Multi-Agent Collaboration)<\/h3>\n<p>\u7576\u4efb\u52d9\u6975\u5ea6\u8907\u96dc\u6642\uff0c\u55ae\u4e00 Agent \u5bb9\u6613\u7522\u751f\u5e7b\u89ba\u3002\u591a Agent \u7cfb\u7d71\u900f\u904e\u89d2\u8272\u5206\u5de5\uff08\u5982\uff1a\u7522\u54c1\u7d93\u7406\u3001\u67b6\u69cb\u5e2b\u3001\u5de5\u7a0b\u5e2b\u3001\u4ee3\u78bc\u5be9\u67e5\u54e1\uff09\u76f8\u4e92\u8faf\u8ad6\u3001\u6aa2\u67e5\u4e26\u5b8c\u6210\u4ea4\u4ed8\u3002<\/p>\n<\/li>\n<\/ol>\n<\/section>\n<hr \/>\n<section>\n<h2>\u4e09\u3001\u6df1\u5165\u7406\u89e3 ReAct \u601d\u8003\u8ff4\u5708 (Reasoning + Acting)<\/h2>\n<p>ReAct \u662f AI Agent \u6700\u7d93\u5178\u7684\u6c7a\u7b56\u67b6\u69cb\uff0c\u5176\u6838\u5fc3\u6a21\u5f0f\u70ba\uff1a<strong>Thought\uff08\u601d\u8003\uff09 \u27a1\ufe0f Action\uff08\u884c\u52d5\/\u547c\u53eb\u5de5\u5177\uff09 \u27a1\ufe0f Observation\uff08\u89c0\u5bdf\u74b0\u5883\u53cd\u994b\uff09<\/strong>\uff0c\u4e0d\u65b7\u5faa\u74b0\u76f4\u81f3\u8f38\u51fa Final Answer\u3002<\/p>\n<blockquote>\n<p><strong>\u5be6\u4f8b\u6f14\u793a\uff1a\u667a\u6167\u91d1\u878d\u6295\u8cc7\u8abf\u7814\u4efb\u52d9<\/strong><\/p>\n<p><strong>\u4f7f\u7528\u8005\u76ee\u6a19\uff1a<\/strong><em>\u300c\u5206\u6790\u53f0\u7a4d\u96fb\u8fd1\u671f\u8ca1\u5831\u91cd\u9ede\uff0c\u4e26\u8a08\u7b97\u82e5\u4ee5\u73fe\u50f9\u8cb7\u9032 15 \u80a1\u9700\u8981\u591a\u5c11\u65b0\u53f0\u5e63\uff1f\u300d<\/em><\/p>\n<ul>\n<li><strong>Step 1 [Thought]\uff1a<\/strong> \u4f7f\u7528\u8005\u9700\u8981\u5169\u9805\u8cc7\u8a0a\uff1a1. \u53f0\u7a4d\u96fb\u6700\u65b0\u8ca1\u5831\u91cd\u9ede\uff0c2. \u5373\u6642\u80a1\u50f9\u4e26\u4e58\u4ee5 15\u3002\u6211\u9700\u8981\u5148\u641c\u5c0b\u6700\u65b0\u7684\u6cd5\u8aaa\u6703\u91cd\u9ede\u8207\u73fe\u50f9\u3002<\/li>\n<li><strong>Step 2 [Action]\uff1a<\/strong> \u8abf\u7528\u5de5\u5177 <code>web_search(\"TSMC latest earnings report highlights &amp; stock price TWD\")<\/code><\/li>\n<li><strong>Step 3 [Observation]\uff1a<\/strong> \u641c\u5c0b\u7d50\u679c\u56de\u50b3\uff1a\u53f0\u7a4d\u96fb\u6cd5\u8aaa\u6703\u986f\u793a\u5148\u9032\u88fd\u7a0b\u4f54\u6bd4\u903e 65%\uff0cAI \u6676\u7247\u9700\u6c42\u5f37\u52c1\uff1b\u76ee\u524d\u53f0\u7a4d\u96fb\u76e4\u4e2d\u53c3\u8003\u50f9\u70ba 1,020 \u5143\u65b0\u53f0\u5e63\u3002<\/li>\n<li><strong>Step 4 [Thought]\uff1a<\/strong> \u5df2\u53d6\u5f97\u80a1\u50f9\uff081,020 \u5143\uff09\uff0c\u73fe\u5728\u9700\u8981\u7cbe\u78ba\u8a08\u7b97 1,020 * 15 \u7684\u6578\u503c\u3002<\/li>\n<li><strong>Step 5 [Action]\uff1a<\/strong> \u8abf\u7528\u5de5\u5177 <code>calculator(\"1020 * 15\")<\/code><\/li>\n<li><strong>Step 6 [Observation]\uff1a<\/strong> \u8a08\u7b97\u6a5f\u8fd4\u56de\uff1a<code>15300<\/code>\u3002<\/li>\n<li><strong>Step 7 [Final Answer]\uff1a<\/strong> \u5f59\u6574\u91cd\u9ede\u8207\u8a66\u7b97\u7d50\u679c\uff0c\u5411\u4f7f\u7528\u8005\u7522\u51fa\u7d50\u69cb\u5316\u7814\u5831\u3002<\/li>\n<\/ul>\n<\/blockquote>\n<\/section>\n<hr \/>\n<section>\n<h2>\u56db\u3001AI Agent \u4e94\u500b\u968e\u6bb5\u5b8c\u6574\u5b78\u7fd2\u8def\u5f91<\/h2>\n<details open=\"\">\n<summary><strong>\u968e\u6bb5\u4e00\uff1a\u57fa\u790e\u5730\u57fa (Foundation &amp; Prompting) \u2014 \u7d04 1 ~ 2 \u9031<\/strong><\/summary>\n<ul>\n<li>\u7cbe\u901a Python \u7570\u6b65\u7de8\u7a0b (AsyncIO) \u8207\u4e3b\u6d41 LLM API (OpenAI, Anthropic, Gemini)\u3002<\/li>\n<li>\u638c\u63e1\u7d50\u69cb\u5316\u8f38\u51fa (Structured Output \/ Pydantic \/ JSON Schema Validation)\u3002<\/li>\n<li>\u7406\u89e3 Function Calling \/ Tool Calling \u7684\u5e95\u5c64 JSON \u683c\u5f0f\u4ea4\u4e92\u3002<\/li>\n<\/ul>\n<\/details><br \/><details open=\"\">\n<summary><strong>\u968e\u6bb5\u4e8c\uff1a\u6838\u5fc3\u5fc3\u667a\u6a21\u578b (ReAct &amp; Memory) \u2014 \u7d04 2 ~ 3 \u9031<\/strong><\/summary>\n<ul>\n<li>\u624b\u5beb\u539f\u751f ReAct \u8ff4\u5708\uff0c\u4e0d\u4f9d\u8cf4\u4efb\u4f55\u5916\u90e8\u91cd\u578b\u6846\u67b6\u3002<\/li>\n<li>\u7406\u89e3\u9577\u77ed\u671f\u8a18\u61b6\u8a2d\u8a08\uff1aConversation Buffer\u3001\u5411\u91cf\u6aa2\u7d22 (RAG) \u8207\u60c5\u7bc0\u8a18\u61b6 (Episodic Memory)\u3002<\/li>\n<li>\u5be6\u4f5c\u81ea\u6211\u53cd\u601d (Reflexion) \u8207\u91cd\u8a66\u6a5f\u5236\uff0c\u9632\u6b62\u6b7b\u5faa\u74b0\u3002<\/li>\n<\/ul>\n<\/details><br \/><details open=\"\">\n<summary><strong>\u968e\u6bb5\u4e09\uff1a\u73fe\u4ee3 Agent \u6846\u67b6\u5be6\u6230 (Modern Frameworks) \u2014 \u7d04 3 ~ 4 \u9031<\/strong><\/summary>\n<ul>\n<li><strong>LangGraph\uff1a<\/strong> \u638c\u63e1\u72c0\u614b\u5716 (StateGraph)\u3001\u5faa\u74b0\u7bc0\u9ede\u8207\u689d\u4ef6\u5206\u652f (Conditional Edges)\u3002<\/li>\n<li><strong>CrewAI\uff1a<\/strong> \u638c\u63e1\u89d2\u8272\u626e\u6f14\u578b (Role-playing) \u591a Agent \u5718\u968a\u5354\u540c\u3002<\/li>\n<li><strong>AutoGen \/ LlamaIndex Workflows\uff1a<\/strong> \u638c\u63e1\u5c0d\u8a71\u9a45\u52d5\u8207\u4e8b\u4ef6\u9a45\u52d5\u7684 Agent \u8a2d\u8a08\u3002<\/li>\n<\/ul>\n<\/details><br \/><details open=\"\">\n<summary><strong>\u968e\u6bb5\u56db\uff1a\u591a Agent \u7cfb\u7d71\u8207\u7de8\u6392 (Multi-Agent Systems) \u2014 \u7d04 3 ~ 4 \u9031<\/strong><\/summary>\n<ul>\n<li>\u4e3b\u7ba1-\u54e1\u5de5\u67b6\u69cb (Supervisor-Worker Pattern)\u3002<\/li>\n<li>\u540c\u5115\u8faf\u8ad6\u8207\u5171\u8b58\u6a5f\u5236 (Debate &amp; Consensus)\u3002<\/li>\n<li>\u4eba\u6a5f\u5354\u540c (Human-in-the-loop) \u5be9\u6279\u8207\u78ba\u8a8d\u6d41\u7a0b\u3002<\/li>\n<\/ul>\n<\/details><br \/><details open=\"\">\n<summary><strong>\u968e\u6bb5\u4e94\uff1a\u8a55\u4f30\u3001\u5b89\u5168\u9632\u8b77\u8207\u751f\u7522\u843d\u5730 (Production &amp; Eval) \u2014 \u7d04 2 ~ 3 \u9031<\/strong><\/summary>\n<ul>\n<li>Agent \u8a55\u6e2c\u57fa\u6e96\u8207\u6307\u6a19 (Task Success Rate, SWE-bench, Ragas)\u3002<\/li>\n<li>\u5168\u93c8\u8def\u8ffd\u8e64\u8207\u53ef\u89c0\u6e2c\u6027 (LangSmith, Langfuse)\u3002<\/li>\n<li>\u5b89\u5168\u9632\u79a6\uff1aPrompt Injection \u9632\u8b77\u8207\u5b89\u5168\u6c99\u76d2\u57f7\u884c (Docker \/ E2B Sandbox)\u3002<\/li>\n<\/ul>\n<\/details><\/section>\n<hr \/>\n<section>\n<h2><img data-dominant-color=\"ccd8e1\" data-has-transparency=\"false\" style=\"--dominant-color: #ccd8e1;\" loading=\"lazy\" decoding=\"async\" class=\"alignnone  wp-image-12256 not-transparent\" src=\"https:\/\/infernews.com\/blog\/wp-content\/uploads\/2026\/09\/image-9-300x84.png\" alt=\"\" width=\"754\" height=\"211\" srcset=\"https:\/\/infernews.com\/blog\/wp-content\/uploads\/2026\/09\/image-9-300x84.png 300w, https:\/\/infernews.com\/blog\/wp-content\/uploads\/2026\/09\/image-9-768x216.png 768w, https:\/\/infernews.com\/blog\/wp-content\/uploads\/2026\/09\/image-9.png 1024w\" sizes=\"auto, (max-width: 754px) 100vw, 754px\" \/><\/h2>\n<h2>\u4e94\u3001\u4e3b\u6d41\u6846\u67b6\u7a0b\u5f0f\u78bc\u7bc4\u4f8b<\/h2>\n<h3>1. LangGraph\uff08\u72c0\u614b\u5716\u67b6\u69cb &#8211; \u751f\u7522\u74b0\u5883\u9996\u9078\uff09<\/h3>\n<div class=\"codecopy-container\">\n                <div class=\"codecopy-header\">\n                    \n                    <button class=\"codecopy-btn\" data-target=\"codecopy-6ac1925b96c28\" title=\"Copy code to clipboard\">\n                <svg width=\"16\" height=\"16\" viewBox=\"0 0 24 24\" fill=\"none\" xmlns=\"http:\/\/www.w3.org\/2000\/svg\">\n                    <path d=\"M16 1H4C2.9 1 2 1.9 2 3V17H4V3H16V1ZM19 5H8C6.9 5 6 5.9 6 7V21C6 22.1 6.9 23 8 23H19C20.1 23 21 22.1 21 21V7C21 5.9 20.1 5 19 5ZM19 21H8V7H19V21Z\" fill=\"currentColor\"\/>\n                <\/svg>\n                <span class=\"codecopy-text\">Copy<\/span>\n            <\/button>\n                <\/div>\n                <pre class=\"line-numbers\" id=\"codecopy-6ac1925b96c28\"><code class=\"language-text\">import operator\nfrom typing import Annotated, TypedDict, List\nfrom langchain_core.messages import BaseMessage, HumanMessage\nfrom langchain_openai import ChatOpenAI\nfrom langgraph.graph import StateGraph, END\nfrom langgraph.prebuilt import ToolNode\n\n# 1. \u5b9a\u7fa9 Agent \u72c0\u614b (State)\nclass AgentState(TypedDict):\n    messages: Annotated[List[BaseMessage], operator.add]\n\n# 2. \u5b9a\u7fa9\u6a21\u578b\u8207\u5de5\u5177\nmodel = ChatOpenAI(model=\"gpt-4o\", temperature=0)\ntools = [my_search_tool, my_calculator_tool]\nmodel_with_tools = model.bind_tools(tools)\n\n# 3. \u5b9a\u7fa9\u7bc0\u9ede\u908f\u8f2f\ndef call_agent(state: AgentState):\n    response = model_with_tools.invoke(state[\"messages\"])\n    return {\"messages\": [response]}\n\ndef should_continue(state: AgentState):\n    last_message = state[\"messages\"][-1]\n    if last_message.tool_calls:\n        return \"tools\"\n    return END\n\n# 4. \u69cb\u5efa\u72c0\u614b\u5716\nworkflow = StateGraph(AgentState)\nworkflow.add_node(\"agent\", call_agent)\nworkflow.add_node(\"tools\", ToolNode(tools))\n\nworkflow.set_entry_point(\"agent\")\nworkflow.add_conditional_edges(\"agent\", should_continue)\nworkflow.add_edge(\"tools\", \"agent\")\n\napp = workflow.compile()\nresult = app.invoke({\"messages\": [HumanMessage(content=\"\u5206\u6790\u9019\u6a94\u80a1\u7968\u7684\u672c\u76ca\u6bd4\")]})<\/code><\/pre>\n            <\/div>\n<h3>2. CrewAI\uff08\u89d2\u8272\u626e\u6f14\u5718\u968a\u5354\u540c &#8211; \u5feb\u901f\u4e0a\u624b\uff09<\/h3>\n<div class=\"codecopy-container\">\n                <div class=\"codecopy-header\">\n                    \n                    <button class=\"codecopy-btn\" data-target=\"codecopy-6ac1925b96cbd\" title=\"Copy code to clipboard\">\n                <svg width=\"16\" height=\"16\" viewBox=\"0 0 24 24\" fill=\"none\" xmlns=\"http:\/\/www.w3.org\/2000\/svg\">\n                    <path d=\"M16 1H4C2.9 1 2 1.9 2 3V17H4V3H16V1ZM19 5H8C6.9 5 6 5.9 6 7V21C6 22.1 6.9 23 8 23H19C20.1 23 21 22.1 21 21V7C21 5.9 20.1 5 19 5ZM19 21H8V7H19V21Z\" fill=\"currentColor\"\/>\n                <\/svg>\n                <span class=\"codecopy-text\">Copy<\/span>\n            <\/button>\n                <\/div>\n                <pre class=\"line-numbers\" id=\"codecopy-6ac1925b96cbd\"><code class=\"language-text\">from crewai import Agent, Task, Crew, Process\nfrom langchain_community.tools import DuckDuckGoSearchRun\n\nsearch_tool = DuckDuckGoSearchRun()\n\n# \u5b9a\u7fa9\u5c08\u696d\u89d2\u8272\nresearcher = Agent(\n    role='\u8cc7\u6df1\u7522\u696d\u7814\u7a76\u54e1',\n    goal='\u6df1\u5165\u63a2\u52d8\u7279\u5b9a\u9818\u57df\u7684\u6700\u65b0\u6280\u8853\u7a81\u7834\u8207\u6578\u64da',\n    backstory='\u64c1\u670910\u5e74\u79d1\u6280\u7522\u696d\u5206\u6790\u7d93\u9a57\uff0c\u5584\u65bc\u5728\u6d77\u91cf\u8cc7\u8a0a\u4e2d\u53bb\u856a\u5b58\u83c1\u3002',\n    tools=[search_tool],\n    verbose=True\n)\n\nwriter = Agent(\n    role='\u9996\u5e2d\u79d1\u6280\u5c08\u6b04\u4f5c\u5bb6',\n    goal='\u5c07\u8abf\u7814\u6578\u64da\u6574\u7406\u6210\u7d50\u69cb\u6e05\u6670\u3001\u5f15\u4eba\u5165\u52dd\u7684\u767d\u76ae\u66f8',\n    backstory='\u77e5\u540d\u79d1\u6280\u5a92\u9ad4\u5c08\u6b04\u4e3b\u7b46\uff0c\u6587\u7b46\u7cbe\u7149\u6709\u529b\u3002',\n    verbose=True\n)\n\n# \u5b9a\u7fa9\u4efb\u52d9\ntask1 = Task(\n    description='\u8abf\u7814 2026 \u5e74\u6700\u65b0 AI Agent \u67b6\u69cb\u8da8\u52e2',\n    expected_output='\u689d\u5217 5 \u5927\u95dc\u9375\u6280\u8853\u7a81\u7834',\n    agent=researcher\n)\ntask2 = Task(\n    description='\u6839\u64da\u8abf\u7814\u7522\u51fa Markdown \u683c\u5f0f\u7684\u7e3d\u7d50\u5831\u544a',\n    expected_output='\u5b8c\u6574\u7684 Markdown \u7814\u5831',\n    agent=writer\n)\n\n# \u7d44\u6210\u5718\u968a\u4e26\u57f7\u884c\ntech_crew = Crew(\n    agents=[researcher, writer],\n    tasks=[task1, task2],\n    process=Process.sequential\n)\n\nresult = tech_crew.kickoff()<\/code><\/pre>\n            <\/div>\n<h3>3. \u539f\u751f Function Calling\uff08\u7406\u89e3\u5e95\u5c64\u539f\u7406\uff09<\/h3>\n<div class=\"codecopy-container\">\n                <div class=\"codecopy-header\">\n                    \n                    <button class=\"codecopy-btn\" data-target=\"codecopy-6ac1925b96cc3\" title=\"Copy code to clipboard\">\n                <svg width=\"16\" height=\"16\" viewBox=\"0 0 24 24\" fill=\"none\" xmlns=\"http:\/\/www.w3.org\/2000\/svg\">\n                    <path d=\"M16 1H4C2.9 1 2 1.9 2 3V17H4V3H16V1ZM19 5H8C6.9 5 6 5.9 6 7V21C6 22.1 6.9 23 8 23H19C20.1 23 21 22.1 21 21V7C21 5.9 20.1 5 19 5ZM19 21H8V7H19V21Z\" fill=\"currentColor\"\/>\n                <\/svg>\n                <span class=\"codecopy-text\">Copy<\/span>\n            <\/button>\n                <\/div>\n                <pre class=\"line-numbers\" id=\"codecopy-6ac1925b96cc3\"><code class=\"language-text\">from openai import OpenAI\nimport json\n\nclient = OpenAI()\n\ntools = [{\n    \"type\": \"function\",\n    \"function\": {\n        \"name\": \"get_stock_price\",\n        \"description\": \"\u53d6\u5f97\u6307\u5b9a\u80a1\u7968\u4ee3\u78bc\u7684\u5373\u6642\u80a1\u50f9\",\n        \"parameters\": {\n            \"type\": \"object\",\n            \"properties\": {\n                \"ticker\": {\"type\": \"string\", \"description\": \"\u80a1\u7968\u4ee3\u78bc\uff0c\u4f8b\u5982 2330.TW\"}\n            },\n            \"required\": [\"ticker\"]\n        }\n    }\n}]\n\nmessages = [{\"role\": \"user\", \"content\": \"\u8acb\u554f\u53f0\u7a4d\u96fb\u73fe\u50f9\u591a\u5c11\uff1f\"}]\nresponse = client.chat.completions.create(\n    model=\"gpt-4o\",\n    messages=messages,\n    tools=tools\n)\n\ntool_call = response.choices[0].message.tool_calls\nif tool_call:\n    fn_name = tool_call[0].function.name\n    fn_args = json.loads(tool_call[0].function.arguments)\n    print(f\"\u8abf\u7528\u5de5\u5177\u540d\u7a31: {fn_name}, \u53c3\u6578: {fn_args}\")<\/code><\/pre>\n            <\/div>\n<\/section>\n<hr \/>\n<section>\n<h2>\u516d\u3001\u7531\u6dfa\u5165\u6df1\u7684 3 \u500b\u5be6\u6230\u5c08\u6848\u63a8\u85a6<\/h2>\n<h3>\u5c08\u6848 1 (\u5165\u9580)\uff1a\u500b\u4eba Notion \u884c\u4e8b\u66c6\u8207\u5f85\u8fa6\u667a\u6167\u79d8\u66f8<\/h3>\n<ul>\n<li><strong>\u5c08\u6848\u76ee\u6a19\uff1a<\/strong> \u4f7f\u7528\u8005\u900f\u904e\u81ea\u7136\u8a9e\u8a00\u8f38\u5165\u300c\u4e0b\u9031\u4e09\u4e0b\u5348\u5169\u9ede\u548c Alex \u958b\u6280\u8853\u5be9\u67e5\u6703\u300d\uff0cAgent \u81ea\u52d5\u6aa2\u67e5\u884c\u4e8b\u66c6\u885d\u7a81\u3001\u547c\u53eb Notion API \u65b0\u589e\u884c\u7a0b\u4e26\u767c\u9001\u78ba\u8a8d\u90f5\u4ef6\u3002<\/li>\n<li><strong>\u6838\u5fc3\u935b\u934a\uff1a<\/strong> Tool Calling\u3001Pydantic \u8cc7\u6599\u9a57\u8b49\u3001\u5916\u90e8 API \u4e32\u63a5\u3002<\/li>\n<li><strong>\u5efa\u8b70\u8017\u6642\uff1a<\/strong> \u7d04 3 \u5929\u3002<\/li>\n<\/ul>\n<h3>\u5c08\u6848 2 (\u9032\u968e)\uff1a\u81ea\u4e3b\u5168\u7db2\u7522\u696d\u8abf\u7814\u8207 PPT \u751f\u6210 Agent<\/h3>\n<ul>\n<li><strong>\u5c08\u6848\u76ee\u6a19\uff1a<\/strong> \u8f38\u5165\u7279\u5b9a\u7814\u7a76\u4e3b\u984c\uff0cAgent \u81ea\u4e3b\u767c\u8d77\u591a\u8f2a\u641c\u5c0b\u3001\u95b1\u8b80 20+ \u7bc7\u82f1\u6587\u5831\u544a\u8207\u8ad6\u6587\uff0c\u8403\u53d6\u6838\u5fc3\u5716\u8868\u8207\u6578\u64da\uff0c\u4e26\u81ea\u52d5\u5408\u6210 Markdown \u7814\u5831\u8207 PowerPoint \u7c21\u5831\u3002<\/li>\n<li><strong>\u6838\u5fc3\u935b\u934a\uff1a<\/strong> Multi-query RAG\u3001Plan-and-Solve \u9577\u4efb\u52d9\u62c6\u89e3\u3001\u591a\u6587\u4ef6\u7e3d\u7d50\u3002<\/li>\n<li><strong>\u5efa\u8b70\u8017\u6642\uff1a<\/strong> \u7d04 1 \u9031\u3002<\/li>\n<\/ul>\n<h3>\u5c08\u6848 3 (\u5c08\u5bb6)\uff1aSWE-Bench \u8edf\u9ad4\u81ea\u4e3b\u4fee\u5fa9\u5de5\u7a0b\u5e2b<\/h3>\n<ul>\n<li><strong>\u5c08\u6848\u76ee\u6a19\uff1a<\/strong> \u5728\u5b89\u5168\u7684 Docker \u6c99\u76d2\u4e2d\uff0cAgent \u81ea\u52d5\u62c9\u53d6 GitHub Issue\u3001\u641c\u5c0b\u95dc\u806f\u7a0b\u5f0f\u78bc\u5eab\u3001\u5b9a\u4f4d Bug\u3001\u64b0\u5beb\u4fee\u5fa9\u7a0b\u5f0f\u78bc\u4e26\u81ea\u52d5\u57f7\u884c\u55ae\u5143\u6e2c\u8a66\uff0c\u76f4\u5230\u5168\u90e8\u6e2c\u8a66\u901a\u904e\uff08Green\uff09\u3002<\/li>\n<li><strong>\u6838\u5fc3\u935b\u934a\uff1a<\/strong> Docker \/ E2B \u6c99\u76d2\u74b0\u5883\u3001\u6a94\u6848\u7cfb\u7d71\u64cd\u63a7\u3001Reflexion \u53cd\u601d\u6a5f\u5236\u3002<\/li>\n<li><strong>\u5efa\u8b70\u8017\u6642\uff1a<\/strong> \u7d04 2 ~ 3 \u9031\u3002<\/li>\n<\/ul>\n<\/section>\n<hr \/>\n<section>\n<h2>\u4e03\u3001\u5e38\u898b\u554f\u984c\u8207\u89c0\u5ff5\u81ea\u6211\u6aa2\u6e2c (FAQ)<\/h2>\n<details>\n<summary><strong>Q1: \u70ba\u4ec0\u9ebc Agent \u5728\u9577\u4efb\u52d9\u4e2d\u5bb9\u6613\u9677\u5165\u300c\u6b7b\u5faa\u74b0\u300d\uff1f\u5982\u4f55\u89e3\u6c7a\uff1f<\/strong><\/summary>\n<p>\u6b7b\u5faa\u74b0\u901a\u5e38\u767c\u751f\u5728\u5de5\u5177\u8fd4\u56de\u932f\u8aa4\u8a0a\u606f\u6642\uff0c\u6a21\u578b\u6301\u7e8c\u7d66\u51fa\u76f8\u540c\u7684\u91cd\u8a66\u53c3\u6578\u3002\u89e3\u6c7a\u65b9\u6848\u5305\u542b\uff1a<\/p>\n<ul>\n<li>\u5728\u72c0\u614b\u6a5f\u4e2d\u8a2d\u7f6e <code>max_iterations<\/code> \u786c\u6027\u8a08\u6578\u4e0a\u9650\u3002<\/li>\n<li>\u5f15\u5165 <strong>Reflexion \u6a5f\u5236<\/strong>\uff1a\u7576\u5169\u6b21\u5617\u8a66\u5931\u6557\u5f8c\uff0c\u5f37\u5236\u5207\u63db\u81f3\u53cd\u601d\u7bc0\u9ede\uff0c\u5206\u6790\u5931\u6557\u539f\u56e0\u518d\u91cd\u65b0\u898f\u5283\u3002<\/li>\n<li>\u63d0\u4f9b\u6e05\u6670\u7684\u5de5\u5177\u932f\u8aa4\u8a0a\u606f\uff08Error Messages\uff09\uff0c\u544a\u77e5\u6a21\u578b\u5177\u9ad4\u7f3a\u5c11\u54ea\u9805\u53c3\u6578\u3002<\/li>\n<\/ul>\n<\/details><br \/><details>\n<summary><strong>Q2: \u4ec0\u9ebc\u6642\u5019\u8a72\u7528 LangGraph\uff0c\u4ec0\u9ebc\u6642\u5019\u8a72\u7528 CrewAI\uff1f<\/strong><\/summary>\n<ul>\n<li><strong>\u9078\u64c7 CrewAI\uff1a<\/strong> \u9069\u5408\u64ec\u4eba\u5316\u3001\u89d2\u8272\u660e\u78ba\u3001\u4ee5\u5c0d\u8a71\u5354\u4f5c\u70ba\u4e3b\u7684\u5834\u666f\uff08\u4f8b\u5982\uff1a\u8abf\u7814\u5718\u968a\u3001\u5167\u5bb9\u5275\u4f5c\u5718\u968a\u3001\u8faf\u8ad6\u7cfb\u7d71\uff09\uff0c\u4e0a\u624b\u901f\u5ea6\u6700\u5feb\u3002<\/li>\n<li><strong>\u9078\u64c7 LangGraph\uff1a<\/strong> \u9069\u5408\u9ad8\u63a7\u5236\u5ea6\u3001\u5177\u5099\u8907\u96dc\u689d\u4ef6\u5206\u652f\u3001\u72c0\u614b\u6301\u4e45\u5316\uff08Persistence\uff09\u8207\u9700\u8981\u4eba\u5de5\u4ecb\u5165\uff08Human-in-the-loop\uff09\u7684\u4f01\u696d\u7d1a\u751f\u7522\u74b0\u5883\u3002<\/li>\n<\/ul>\n<\/details><br \/><details>\n<summary><strong>Q3: \u5982\u4f55\u78ba\u4fdd Agent \u547c\u53eb\u5de5\u5177\u6642\u7684\u5b89\u5168\u6027\uff1f<\/strong><\/summary>\n<ul>\n<li>\u6c38\u9060\u4f7f\u7528\u7368\u7acb\u7684\u6c99\u76d2\u74b0\u5883\uff08\u5982 Docker \u5bb9\u5668\u3001E2B Sandbox\uff09\u57f7\u884c\u4efb\u610f\u4ee3\u78bc\u3002<\/li>\n<li>\u5c0d\u654f\u611f\u64cd\u4f5c\uff08\u5982\uff1a\u767c\u9001\u90f5\u4ef6\u3001\u522a\u9664\u8cc7\u6599\u5eab\u3001\u4ed8\u6b3e\uff09\u8a2d\u7f6e <strong>Human-in-the-loop<\/strong> \u78ba\u8a8d\u95dc\u5361\u3002<\/li>\n<li>\u5c0d\u8f38\u5165\u8207\u8f38\u51fa\u9032\u884c\u9632\u8b77\u6b04\uff08Guardrails\uff09\u6aa2\u67e5\uff0c\u9632\u7bc4 Prompt Injection \u8207\u8d8a\u7344\u653b\u64ca\u3002<\/li>\n<\/ul>\n<\/details><\/section>\n<\/article>\n\n\n\n<hr class=\"wp-block-separator has-alpha-channel-opacity\"\/>\n\n\n\n<p class=\"wp-block-paragraph\">\u7531\u804a\u5929\u6a21\u578b\u9032\u4e00\u6b65\u5efa\u7acb\u80fd\u5920\u898f\u5283\u3001\u4f7f\u7528\u5de5\u5177\u53ca\u6301\u7e8c\u57f7\u884c\u4efb\u52d9\u7684 AI Agent\uff0c\u55ae\u9760\u63d0\u793a\u8a5e\u5f80\u5f80\u4e0d\u8db3\u3002<\/p>\n\n\n<div class=\"vlp-link-container vlp-layout-spotlight-clone block-editor-block-list__block wp-block wp-block-visual-link-preview-link\"><a href=\"https:\/\/openclaw.ai\/\" class=\"vlp-link\" title=\"OpenClaw \u2014 Personal AI Assistant\" rel=\"nofollow\" target=\"_blank\"><\/a><div class=\"vlp-layout-zone-main\"><span class=\"vlp-block-0 vlp-link-title\">OpenClaw \u2014 Personal AI Assistant<\/span><div class=\"vlp-block-1 vlp-link-summary\">OpenClaw \u2014 The AI that actually does things. Your personal assistant on any platform.<\/div><div class=\"vlp-block-2 vlp-link-image\"><img decoding=\"async\" src=\"https:\/\/openclaw.ai\/og-image.png\" style=\"max-width: 1024px; max-height: 1024px\" \/><\/div><\/div><\/div>\n\n<div class=\"vlp-link-container vlp-layout-spotlight-clone block-editor-block-list__block wp-block wp-block-visual-link-preview-link\"><a href=\"https:\/\/hermes-agent.nousresearch.com\/\" class=\"vlp-link\" title=\"Hermes Agent \u2014 The Agent That Grows With You\" rel=\"nofollow\" target=\"_blank\"><\/a><div class=\"vlp-layout-zone-main\"><span class=\"vlp-block-0 vlp-link-title\">Hermes Agent \u2014 The Agent That Grows With You<\/span><div class=\"vlp-block-1 vlp-link-summary\">An open-source agent that grows with you. Install it, give it your messaging accounts, and it becomes a persistent personal agent.<\/div><div class=\"vlp-block-2 vlp-link-image\"><img decoding=\"async\" src=\"https:\/\/hermes-agent.nousresearch.com\/img\/hermes-og-image.png\" style=\"max-width: 1024px; max-height: 1024px\" \/><\/div><\/div><\/div>\n\n\n<p class=\"wp-block-paragraph\"><a href=\"https:\/\/infernews.com\/adk\/\">Google ADK \u6280\u8853\u6982\u8ff0<\/a><\/p>\n\n\n\r\n<!--  Starting Simple Link Directory Plugin Output -->\r\n<div class=\"qcld-main-container-style-simple\"><div class=\"qcopd-list-wrapper\" >\r\n\t<div id=\"opd-list-holder\" class=\"qc-grid qcopd-list-holder\">\r\n\t\t\r\n\t\t<!-- Individual List Item -->\r\n\t\t<div id=\"list-item-1-3608\" class=\"qc-grid-item qcopd-list-column opd-column-1 simple opd-list-id-3608\">\r\n\t\t\t<div class=\"qcopd-single-list\">\r\n\t\t\t\t\r\n\t\t\t\t<h2>\r\n\t\t\t\t\tA.I. Agent\t\t\t\t<\/h2>\r\n\t\t\t\t<ul>\r\n\t\t\t\t\t\r\n\t\t\t\t\t<li id=\"item-3608-1\" style=\"\">\r\n\t\t\t\t\t\t\t\t\t\t\t\t<!-- List Anchor -->\r\n\t\t\t\t\t\t<a   href=\"https:\/\/infernews.com\/agent\/\"\r\n\t\t\t\t\t\t\ttarget=\"_blank\">\r\n\r\n\t\t\t\t\t\t\t<!-- Image, If Present -->\r\n\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t<span class=\"list-img\">\r\n\t\t\t\t\t\t\t\t\t<img decoding=\"async\" src=\"https:\/\/infernews.com\/blog\/wp-content\/plugins\/simple-link-directory\/assets\/images\/list-image-placeholder.png\" alt=\"\">\r\n\t\t\t\t\t\t\t\t<\/span>\r\n\t\t\t\t\t\t\t\r\n\t\t\t\t\t\t\t<!-- Link Text -->\r\n\t\t\t\t\t\t\tAI MCP, A2A, MACP \u61c9\u7528\u767c\u5c55\u524d\u666f\r\n\t\t\t\t\t\t<\/a>\r\n\t\t\t\t\t\t\r\n\t\t\t\t\t\t\t<!-- upvote section -->\r\n\t\t\t\t\t\t\t<div class=\"upvote-section\">\r\n\t\t\t\t\t\t\t\t<span data-post-id=\"3608\" data-item-title=\"AI MCP, A2A, MACP \u61c9\u7528\u767c\u5c55\u524d\u666f\" data-item-link=\"https:\/\/infernews.com\/agent\/\" class=\"upvote-btn upvote-on\">\r\n\t\t\t\t\t\t\t\t\t<i class=\"fa fa-thumbs-up\"><\/i>\r\n\t\t\t\t\t\t\t\t<\/span>\r\n\t\t\t\t\t\t\t\t<span class=\"upvote-count\">\r\n\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t<\/span>\r\n\t\t\t\t\t\t\t<\/div>\r\n\t\t\t\t\t\t\t<!-- \/upvote section -->\r\n\r\n\t\t\t\t\t\t\t\t\t\t\t\t\r\n\t\t\t\t\t<\/li>\r\n\t\t\t\t\t\r\n\t\t\t\t\t<li id=\"item-3608-2\" style=\"\">\r\n\t\t\t\t\t\t\t\t\t\t\t\t<!-- List Anchor -->\r\n\t\t\t\t\t\t<a   href=\"https:\/\/youtu.be\/eHEHE2fpnWQ?si=WV7m6w9DF8jS1b4J\"\r\n\t\t\t\t\t\t\ttarget=\"_blank\">\r\n\r\n\t\t\t\t\t\t\t<!-- Image, If Present -->\r\n\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t<span class=\"list-img\">\r\n\t\t\t\t\t\t\t\t\t<img decoding=\"async\" src=\"https:\/\/infernews.com\/blog\/wp-content\/plugins\/simple-link-directory\/assets\/images\/list-image-placeholder.png\" alt=\"\">\r\n\t\t\t\t\t\t\t\t<\/span>\r\n\t\t\t\t\t\t\t\r\n\t\t\t\t\t\t\t<!-- Link Text -->\r\n\t\t\t\t\t\t\tWhat Are AI Agents Really About?\r\n\t\t\t\t\t\t<\/a>\r\n\t\t\t\t\t\t\r\n\t\t\t\t\t\t\t<!-- upvote section -->\r\n\t\t\t\t\t\t\t<div class=\"upvote-section\">\r\n\t\t\t\t\t\t\t\t<span data-post-id=\"3608\" data-item-title=\"What Are AI Agents Really About?\" data-item-link=\"https:\/\/youtu.be\/eHEHE2fpnWQ?si=WV7m6w9DF8jS1b4J\" class=\"upvote-btn upvote-on\">\r\n\t\t\t\t\t\t\t\t\t<i class=\"fa fa-thumbs-up\"><\/i>\r\n\t\t\t\t\t\t\t\t<\/span>\r\n\t\t\t\t\t\t\t\t<span class=\"upvote-count\">\r\n\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t<\/span>\r\n\t\t\t\t\t\t\t<\/div>\r\n\t\t\t\t\t\t\t<!-- \/upvote section -->\r\n\r\n\t\t\t\t\t\t\t\t\t\t\t\t\r\n\t\t\t\t\t<\/li>\r\n\t\t\t\t\t\t\t\t\t<\/ul>\r\n\r\n\t\t\t<\/div>\r\n\r\n\t\t<\/div>\r\n\t\t<!-- \/Individual List Item -->\r\n\r\n\t\t\r\n\t\t<div class=\"sld-clearfix\"><\/div>\r\n\t\t\t<\/div>\r\n\t\t<div class=\"sld-clearfix\"><\/div>\r\n\t<\/div><\/div>\n\n\n\n<h2 class=\"wp-block-heading\">\u81ea\u6211\u9032\u5316\u4ee3\u7406\u8abf\u67e5\uff1a\u9081\u5411\u8d85\u7d1a\u4eba\u5de5\u667a\u6167\u4e4b\u8def<\/h2>\n\n\n\n<p class=\"wp-block-paragraph\">\u4e00\u500b <a href=\"https:\/\/github.com\/CharlesQ9\/Self-Evolving-Agents\">GitHub \u5009\u5eab<\/a>\uff0c\u7528\u65bc\u8a18\u9304\u8207\u63a8\u7406\u76f8\u95dc\u7684\u8ad6\u6587<\/p>\n\n\n\n<figure class=\"wp-block-image size-full\"><a href=\"https:\/\/github.com\/CharlesQ9\/Self-Evolving-Agents\" target=\"_blank\" rel=\" noreferrer noopener\"><img data-dominant-color=\"dee0e1\" data-has-transparency=\"false\" style=\"--dominant-color: #dee0e1;\" loading=\"lazy\" decoding=\"async\" width=\"1024\" height=\"207\" src=\"\/blog\/wp-content\/uploads\/2025\/07\/develop.jpeg\" alt=\"\" class=\"wp-image-6009 not-transparent\" srcset=\"https:\/\/infernews.com\/blog\/wp-content\/uploads\/2025\/07\/develop.jpeg 1024w, https:\/\/infernews.com\/blog\/wp-content\/uploads\/2025\/07\/develop-300x61.jpeg 300w, https:\/\/infernews.com\/blog\/wp-content\/uploads\/2025\/07\/develop-768x155.jpeg 768w\" sizes=\"auto, (max-width: 1024px) 100vw, 1024px\" \/><\/a><\/figure>\n\n\n\n<hr class=\"wp-block-separator has-alpha-channel-opacity\"\/>\n\n\n<figure class=\"wp-block-embed-youtube wp-block-embed is-type-video is-provider-youtube wp-embed-aspect-16-9 wp-has-aspect-ratio\"><div class=\"lyte-wrapper\" title=\"\u539f\u6765\u5199\u4e00\u4e2a AI Agent \u8fd9\u4e48\u7b80\u5355\" style=\"width:853px;max-width:100%;margin:5px auto;\"><div class=\"lyMe\" id=\"WYL_OAE1EWP8dIQ\" itemprop=\"video\" itemscope itemtype=\"https:\/\/schema.org\/VideoObject\"><div><meta itemprop=\"thumbnailUrl\" content=\"https:\/\/infernews.com\/blog\/wp-content\/plugins\/wp-youtube-lyte\/lyteCache.php?origThumbUrl=https%3A%2F%2Fi.ytimg.com%2Fvi%2FOAE1EWP8dIQ%2Fhqdefault.jpg\" \/><meta itemprop=\"embedURL\" content=\"https:\/\/www.youtube.com\/embed\/OAE1EWP8dIQ\" \/><meta itemprop=\"duration\" content=\"PT9M9S\" \/><meta itemprop=\"uploadDate\" content=\"2025-05-08T09:00:39Z\" \/><\/div><meta itemprop=\"accessibilityFeature\" content=\"captions\" \/><div id=\"lyte_OAE1EWP8dIQ\" data-src=\"https:\/\/infernews.com\/blog\/wp-content\/plugins\/wp-youtube-lyte\/lyteCache.php?origThumbUrl=https%3A%2F%2Fi.ytimg.com%2Fvi%2FOAE1EWP8dIQ%2Fhqdefault.jpg\" class=\"pL\"><div class=\"tC\"><div class=\"tT\" itemprop=\"name\">\u539f\u6765\u5199\u4e00\u4e2a AI Agent \u8fd9\u4e48\u7b80\u5355<\/div><\/div><button tabindex=\"0\" class=\"play\"><\/button><div class=\"ctrl\"><div class=\"Lctrl\"><\/div><div class=\"Rctrl\"><\/div><\/div><\/div><noscript><a href=\"https:\/\/youtu.be\/OAE1EWP8dIQ\" rel=\"nofollow\"><img loading=\"lazy\" decoding=\"async\" src=\"https:\/\/infernews.com\/blog\/wp-content\/plugins\/wp-youtube-lyte\/lyteCache.php?origThumbUrl=https%3A%2F%2Fi.ytimg.com%2Fvi%2FOAE1EWP8dIQ%2F0.jpg\" alt=\"\u539f\u6765\u5199\u4e00\u4e2a AI Agent \u8fd9\u4e48\u7b80\u5355\" width=\"853\" height=\"460\" \/><br \/>Watch this video on YouTube<\/a><\/noscript><meta itemprop=\"description\" content=\"10\u5206\u949f\u8bb2\u6e05\u695a Prompt, Agent, MCP \u662f\u4ec0\u4e48: https:\/\/www.youtube.com\/watch?v=1UQ8uz4kuIc https:\/\/ai.pydantic.dev\/models\/ Demo Code: https:\/\/gist.github.com\/cradiator\/3be43011060abb6869f4b41f0a9ce165 \u65f6\u95f4\u8f74 00:00:00 \u4ecb\u7ecd 00:00:47 Tools 00:02:25 Agent 00:06:32 History\"><\/div><\/div><div class=\"lL\" style=\"max-width:100%;width:853px;margin:5px auto;\"><\/div><figcaption><\/figcaption><\/figure>\n\n<figure class=\"wp-block-embed-youtube wp-block-embed is-type-video is-provider-youtube wp-embed-aspect-16-9 wp-has-aspect-ratio\"><div class=\"lyte-wrapper\" title=\"Google A2A + MCP = \u5206\u5e03\u5f0fAgents\u7f51\u7edc\" style=\"width:853px;max-width:100%;margin:5px auto;\"><div class=\"lyMe\" id=\"WYL_B1Nkc_FyoSs\" itemprop=\"video\" itemscope itemtype=\"https:\/\/schema.org\/VideoObject\"><div><meta itemprop=\"thumbnailUrl\" content=\"https:\/\/infernews.com\/blog\/wp-content\/plugins\/wp-youtube-lyte\/lyteCache.php?origThumbUrl=https%3A%2F%2Fi.ytimg.com%2Fvi%2FB1Nkc_FyoSs%2Fhqdefault.jpg\" \/><meta itemprop=\"embedURL\" content=\"https:\/\/www.youtube.com\/embed\/B1Nkc_FyoSs\" \/><meta itemprop=\"duration\" content=\"PT10M36S\" \/><meta itemprop=\"uploadDate\" content=\"2025-04-12T00:13:34Z\" \/><\/div><meta itemprop=\"accessibilityFeature\" content=\"captions\" \/><div id=\"lyte_B1Nkc_FyoSs\" data-src=\"https:\/\/infernews.com\/blog\/wp-content\/plugins\/wp-youtube-lyte\/lyteCache.php?origThumbUrl=https%3A%2F%2Fi.ytimg.com%2Fvi%2FB1Nkc_FyoSs%2Fhqdefault.jpg\" class=\"pL\"><div class=\"tC\"><div class=\"tT\" itemprop=\"name\">Google A2A + MCP = \u5206\u5e03\u5f0fAgents\u7f51\u7edc<\/div><\/div><button tabindex=\"0\" class=\"play\"><\/button><div class=\"ctrl\"><div class=\"Lctrl\"><\/div><div class=\"Rctrl\"><\/div><\/div><\/div><noscript><a href=\"https:\/\/youtu.be\/B1Nkc_FyoSs\" rel=\"nofollow\"><img loading=\"lazy\" decoding=\"async\" src=\"https:\/\/infernews.com\/blog\/wp-content\/plugins\/wp-youtube-lyte\/lyteCache.php?origThumbUrl=https%3A%2F%2Fi.ytimg.com%2Fvi%2FB1Nkc_FyoSs%2F0.jpg\" alt=\"Google A2A + MCP = \u5206\u5e03\u5f0fAgents\u7f51\u7edc\" width=\"853\" height=\"460\" \/><br \/>Watch this video on YouTube<\/a><\/noscript><meta itemprop=\"description\" content=\"#google #a2a #mcp #agents \u63a8\u8350\u4e00\u4e2a\u76ee\u524d\u5168\u7f51\u4ef7\u683c\u6700\u5b9e\u60e0\u7684\u5408\u79df\u5e73\u53f0\uff0cChatGPT\uff0cMidJourney\uff0c\u5948\u98de\uff0c\u8fea\u58eb\u5c3c\uff0c\u82f9\u679cTV\u7b49\u70ed\u95e8\u8f6f\u4ef6\u5e94\u6709\u5c3d\u6709 - https:\/\/dub.sh\/unibus \uff0c\u9996\u53559\u6298\u4f18\u60e0 - \u4f18\u60e0\u7801 01Coder - \u52a0\u5165\u6211\u7684skool\u793e\u533a\uff0c\u83b7\u53d6\u66f4\u591aAI\u76f8\u5173\u7684\u5185\u5bb9\u548c\u8d44\u6e90\uff1ahttps:\/\/www.skool.com\/01coder-5817\/about?ref=8959ed0625534f2f926216d137bdeb38 \u3010MCP\u6781\u7b80\u5165\u95e8\u3011\u8bfe\u7a0b\u8d44\u6e90 \u6587\u5b57\u7248\u53d1\u5e03\u5728\u5c0f\u6728\u5934\u7684AI\u661f\u7403\uff1ahttps:\/\/t.zsxq.com\/W5Oj7 \u89c6\u9891\u7248\u53d1\u5e03\u5728B\u7ad9\u8bfe\u5802\u3010MCP\u6781\u7b80\u5165\u95e8\u3011- https:\/\/www.bilibili.com\/cheese\/play\/ss635342154 - \u5173\u6ce8\u6211\u7684Twitter: https:\/\/twitter.com\/verysmallwoods - \u5173\u6ce8\u6211\u7684Bilibili: https:\/\/space.bilibili.com\/615957867\/ - \u5982\u679c\u60a8\u6709\u4efb\u4f55\u95ee\u9898\u6216\u8005\u5efa\u8bae\uff0c\u6b22\u8fce\u901a\u8fc7\u90ae\u4ef6\u8054\u7cfb\u6211\uff1a0001coder@gmail.com Google\u5b98\u65b9\u53d1\u5e03A2A\u534f\u8bae\uff0c\u5b9e\u73b0Agents\u4e92\u8054\u3002\u672c\u671f\u89c6\u9891\u6211\u4f1a\u5206\u4eab\uff1a 1. \u4ec0\u4e48\u662fA2A\uff1f 2. A2A\u4e0eMCP\u7684\u5173\u7cfb\u662f\u4ec0\u4e48\uff1f 3. A2A\u6f14\u793a Google A2A https:\/\/github.com\/google\/A2A \u5982\u679c\u60a8\u559c\u6b22\u8fd9\u4e2a\u89c6\u9891\u5e76\u5e0c\u671b\u8bf7\u6211\u559d\u4e00\u676f\u5496\u5561\uff0c\u6b22\u8fce\u524d\u5f80\u6211\u7684BuyMeACoffee\u9875\u9762 \ud83d\udc49 https:\/\/www.buymeacoffee.com\/01coder \u3010A2A\u3011\u64ad\u653e\u5217\u8868 https:\/\/www.youtube.com\/playlist?list=PL2fGiugrNoog_5JdirPb41f1oJ_GKuJrg \u7ae0\u8282 00:00 Google\u53d1\u5e03A2A\u534f\u8bae 00:59 \u4ec0\u4e48\u662fA2A\uff1f 05:55 A2A\u4e0eMCP\u7684\u5173\u7cfb\u662f\u4ec0\u4e48\uff1f 07:36 A2A\u6f14\u793a 10:14 \u603b\u7ed3\"><\/div><\/div><div class=\"lL\" style=\"max-width:100%;width:853px;margin:5px auto;\"><\/div><figcaption><\/figcaption><\/figure>\n\n<figure class=\"wp-block-embed-youtube wp-block-embed is-type-video is-provider-youtube wp-embed-aspect-16-9 wp-has-aspect-ratio\"><div class=\"lyte-wrapper\" title=\"Google&rsquo;s A2A Protocol (agent to agent)\" style=\"width:853px;max-width:100%;margin:5px auto;\"><div class=\"lyMe\" id=\"WYL_avcBlmqUOzM\" itemprop=\"video\" itemscope itemtype=\"https:\/\/schema.org\/VideoObject\"><div><meta itemprop=\"thumbnailUrl\" content=\"https:\/\/infernews.com\/blog\/wp-content\/plugins\/wp-youtube-lyte\/lyteCache.php?origThumbUrl=https%3A%2F%2Fi.ytimg.com%2Fvi%2FavcBlmqUOzM%2Fhqdefault.jpg\" \/><meta itemprop=\"embedURL\" content=\"https:\/\/www.youtube.com\/embed\/avcBlmqUOzM\" \/><meta itemprop=\"duration\" content=\"PT28M30S\" \/><meta itemprop=\"uploadDate\" content=\"2025-04-21T23:13:44Z\" \/><\/div><div id=\"lyte_avcBlmqUOzM\" data-src=\"https:\/\/infernews.com\/blog\/wp-content\/plugins\/wp-youtube-lyte\/lyteCache.php?origThumbUrl=https%3A%2F%2Fi.ytimg.com%2Fvi%2FavcBlmqUOzM%2Fhqdefault.jpg\" class=\"pL\"><div class=\"tC\"><div class=\"tT\" itemprop=\"name\">Google\u2019s A2A Protocol (agent to agent)<\/div><\/div><button tabindex=\"0\" class=\"play\"><\/button><div class=\"ctrl\"><div class=\"Lctrl\"><\/div><div class=\"Rctrl\"><\/div><\/div><\/div><noscript><a href=\"https:\/\/youtu.be\/avcBlmqUOzM\" rel=\"nofollow\"><img loading=\"lazy\" decoding=\"async\" src=\"https:\/\/infernews.com\/blog\/wp-content\/plugins\/wp-youtube-lyte\/lyteCache.php?origThumbUrl=https%3A%2F%2Fi.ytimg.com%2Fvi%2FavcBlmqUOzM%2F0.jpg\" alt=\"Google&rsquo;s A2A Protocol (agent to agent)\" width=\"853\" height=\"460\" \/><br \/>Watch this video on YouTube<\/a><\/noscript><meta itemprop=\"description\" content=\"Google&#039;s A2A Protocol, standardizes agent 2 agent communication in the same way that MCP standardizes remote tool calling. In this video chris gives an overview of the A2A protocol using his a2a-cli tool and shows you how to build your own agent using adk and host with it a2a. A2A - CLI ---------- Repo: https:\/\/github.com\/chrishayuk\/a2a-cli A2A - Server ---------- Repo: https:\/\/github.com\/chrishayuk\/a2a-server A2A - Agent ---------- Repo: https:\/\/github.com\/chrishayuk\/a2a-agent-record 00:00 - Introduction 02:15 - A2A Design Principles 03:30 - A2A CLI 08:29 - A2A Protocol Overview 10:15 - Agent Capabilities 12:23 - Agent Tasks 13:43 - Artifacts 18:13 - Create an agent with A2A-Server\"><\/div><\/div><div class=\"lL\" style=\"max-width:100%;width:853px;margin:5px auto;\"><\/div><figcaption><\/figcaption><\/figure>","protected":false},"excerpt":{"rendered":"<p>AI Agent \u5be6\u6230\u6307\u5357\uff1a\u5f9e LLM \u5230 Agent \u5c0e\u8b80\uff1a\u4e0d\u8981\u53ea\u505c\u7559\u5728\u63d0\u793a\u8a5e\u5de5\u7a0b\uff08Prompt Engineering\uff09\uff01AI Agent\uff08\u4eba\u5de5\u667a\u6167\u4ee3\u7406\uff09\u662f\u5177\u5099\u300c\u76ee\u6a19\u62c6\u89e3\u3001\u74b0\u5883\u611f\u77e5\u3001\u5de5\u5177\u8abf\u7528\u3001\u8a18\u61b6\u6aa2\u7d22\u8207\u81ea\u4e3b\u6c7a\u7b56\u300d\u7684\u667a\u6167\u7cfb\u7d71\u3002 \u4e00\u3001\u4ec0\u9ebc\u662f AI Agent\uff1f\u5b83\u8207\u50b3\u7d71 Chatbot \u6709\u4f55\u672c\u8cea\u5dee\u7570\uff1f \u50b3\u7d71\u7684\u5927\u8a9e\u8a00\u6a21\u578b\uff08LLM\uff09\u5c31\u50cf\u4e00\u500b\u300c\u96e2\u7dda\u7684\u5927\u8166\u300d\uff0c\u53ea\u80fd\u88ab\u52d5\u6839\u64da\u8f38\u5165\u7684 Prompt \u7522\u51fa\u6587\u5b57\u3002\u800c AI Agent \u662f\u4ee5 LLM \u70ba\u6c7a\u7b56\u6838\u5fc3\uff0c\u70ba\u5176\u914d\u5099\u4e86\u8a18\u61b6\uff08Memory\uff09\u3001\u611f\u5b98\u8207\u5de5\u5177\uff08Tools\/APIs\uff09\u4ee5\u53ca\u898f\u5283\u80fd\u529b\uff08Planning\uff09\u7684\u81ea\u4e3b\u5be6\u9ad4\u3002 \u8a55\u4f30\u7dad\u5ea6 \u50b3\u7d71 LLM \/ Chatbot \u81ea\u4e3b AI Agent \u4e92\u52d5\u6a21\u5f0f \u55ae\u6b21\u8f38\u5165\u8f38\u51fa\uff08One-shot Request &amp; Response\uff09 \u6301\u7e8c\u904b\u884c\u7684\u81ea\u4e3b\u601d\u8003\u8ff4\u5708\uff08ReAct Loop\uff09\u76f4\u5230\u76ee\u6a19\u9054\u6210 \u5916\u90e8\u884c\u52d5\u529b \u7121\uff08\u50c5\u80fd\u8f38\u51fa\u6587\u5b57\u5efa\u8b70\uff09 \u5177\u5099\u5de5\u5177\u8abf\u7528\u80fd\u529b\uff08\u53ef\u547c\u53eb API\u3001\u8b80\u5beb\u8cc7\u6599\u5eab\u3001\u57f7\u884c Python \u4ee3\u78bc\uff09 \u932f\u8aa4\u8655\u7406 \u9700\u7531\u4eba\u985e\u624b\u52d5\u6307\u51fa\u932f\u8aa4\u4e26\u91cd\u65b0\u767c\u554f \u81ea\u52d5\u6355\u6349\u74b0\u5883\u932f\u8aa4\u53cd\u994b\uff08Observation\uff09\uff0c\u81ea\u6211\u53cd\u601d\u4e26\u4fee\u6b63\uff08Self-Correction\uff09 \u9069\u7528\u5834\u666f \u6587\u7ae0\u7ffb\u8b6f\u3001\u6587\u5b57\u6458\u8981\u3001\u7c21\u55ae\u554f\u7b54 \u7aef\u5230\u7aef\u8edf\u9ad4\u958b\u767c\u3001\u81ea\u4e3b\u5e02\u5834\u8abf\u7814\u3001\u591a\u6b65\u9a5f\u81ea\u52d5\u5316\u4efb\u52d9 \u4e8c\u3001AI Agent \u7684\u56db\u5927\u6838\u5fc3\u67b6\u69cb\u9aa8\u67b6 \u73fe\u4ee3 AI Agent [&hellip;]<\/p>\n","protected":false},"author":1,"featured_media":0,"parent":0,"menu_order":0,"comment_status":"closed","ping_status":"closed","template":"","meta":{"wpai_generated_summary":"","footnotes":""},"class_list":["post-5540","page","type-page","status-publish","hentry"],"_links":{"self":[{"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/pages\/5540","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/pages"}],"about":[{"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/types\/page"}],"author":[{"embeddable":true,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/comments?post=5540"}],"version-history":[{"count":9,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/pages\/5540\/revisions"}],"predecessor-version":[{"id":12259,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/pages\/5540\/revisions\/12259"}],"wp:attachment":[{"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/media?parent=5540"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}