[{"data":1,"prerenderedAt":77},["ShallowReactive",2],{"lab-/labs/nova":3},{"id":4,"title":5,"author":6,"body":7,"category":61,"date":62,"description":63,"extension":64,"featured":65,"home_position":59,"image":66,"meta":67,"navigation":65,"order":68,"path":69,"seo":70,"status":71,"stem":72,"tags":73,"__hash__":76},"content/labs/Nova.md","Nova","sibuchen",{"type":8,"value":9,"toc":58},"minimark",[10,17,22,36,47],[11,12,13],"blockquote",{},[14,15,16],"p",{},"Nova （诺瓦）在拉丁语意为“新的”，天文学中指“新星”。每一次旅行都是新的开始，发现新的风景~",[18,19,21],"h1",{"id":20},"前置知识智能体的构成与运行原理","前置知识：智能体的构成与运行原理",[14,23,24,25,30,31],{},"智能体并非一次性完成任务，而是通过一个持续的循环与环境进行交互，这个核心机制被称为",[26,27,29],"span",{"style":28},"color:red","智能体循环 (Agent Loop)","：\n",[32,33],"img",{"alt":34,"src":35},"","/assets/Nova/%E6%99%BA%E8%83%BD%E4%BD%93%E4%B8%8E%E7%8E%AF%E5%A2%83%E4%BA%A4%E4%BA%92%E7%9A%84%E5%9F%BA%E6%9C%AC%E5%BE%AA%E7%8E%AF.png",[14,37,38,39,42,43,46],{},"在工程实践中，为了让 LLM 能够有效驱动这个循环，我们需要一套明确的",[26,40,41],{"style":28},"交互协议 (Interaction Protocol)","来规范其与环境之间的信息交换。\n在许多现代智能体框架中，这一协议体现在对智能体每一次输出的结构化定义上。智能体的输出不再是单一的自然语言回复，而是一段遵循特定格式的文本，其中明确地展示了其内部的推理过程与最终决策。\n由于该循环结构通常由 Thought、Action、Observation 三个部分构成，因此被称为",[26,44,45],{"style":28},"Thought-Action-Observation交互范式","。通过它，LLM智能体得以将内部的语言推理能力，与外部环境的真实信息和工具操作能力有效地结合起来。",[18,48,50,51],{"id":49},"源码位置github","源码位置：",[52,53,57],"a",{"href":54,"rel":55},"https://github.com/sibuchen/Nova--TravelAgent",[56],"nofollow","Github",{"title":34,"searchDepth":59,"depth":59,"links":60},2,[],"agent","2026-03-20","Nova 是一个由作者设计的简单Agent Demo，集中演示了基于 Thought-Action-Observation 循环的智能体所具备的四项基本能力：任务分解、工具调用、上下文理解和结果合成。通过这种循环的不断迭代，Nova 得以将一个模糊的用户意图，转化为一系列具体、可执行的步骤，并最终达成目标。","md",true,null,{},3,"/labs/nova",{"title":5,"description":63},"ARCHIVED","labs/Nova",[74,75,6],"AIAgent","ReAct","rMZtyutqD2YFPZZkfSXZuFxCwNcPIDVEMCC6ipKjQS8",1774960320851]