惯性聚合 高效追踪和阅读你感兴趣的博客、新闻、科技资讯
阅读原文 在惯性聚合中打开

推荐订阅源

K
Kaspersky official blog
Threat Intelligence Blog | Flashpoint
Threat Intelligence Blog | Flashpoint
AI
AI
SecWiki News
SecWiki News
宝玉的分享
宝玉的分享
Scott Helme
Scott Helme
D
Darknet – Hacking Tools, Hacker News & Cyber Security
Exploit-DB.com RSS Feed
Exploit-DB.com RSS Feed
Engineering at Meta
Engineering at Meta
博客园 - 叶小钗
The GitHub Blog
The GitHub Blog
Microsoft Azure Blog
Microsoft Azure Blog
N
News and Events Feed by Topic
Cloudbric
Cloudbric
B
Blog
Cisco Talos Blog
Cisco Talos Blog
V
Vulnerabilities – Threatpost
N
News and Events Feed by Topic
V
Visual Studio Blog
A
Arctic Wolf
奇客Solidot–传递最新科技情报
奇客Solidot–传递最新科技情报
U
Unit 42
S
Security @ Cisco Blogs
博客园 - 聂微东
T
Threat Research - Cisco Blogs
OSCHINA 社区最新新闻
OSCHINA 社区最新新闻
Apple Machine Learning Research
Apple Machine Learning Research
Y
Y Combinator Blog
G
GRAHAM CLULEY
L
LINUX DO - 热门话题
量子位
NISL@THU
NISL@THU
Webroot Blog
Webroot Blog
让小产品的独立变现更简单 - ezindie.com
让小产品的独立变现更简单 - ezindie.com
T
Troy Hunt's Blog
Application and Cybersecurity Blog
Application and Cybersecurity Blog
T
Tenable Blog
月光博客
月光博客
S
Security Affairs
K
KPMG report finds enterprise disconnect between AI and its ROI | CIO
The Hacker News
The Hacker News
Spread Privacy
Spread Privacy
D
Docker
www.infosecurity-magazine.com
www.infosecurity-magazine.com
雷峰网
雷峰网
博客园 - 司徒正美
T
The Exploit Database - CXSecurity.com
Hugging Face - Blog
Hugging Face - Blog
Help Net Security
Help Net Security
D
DataBreaches.Net

DEV Community

Authentication Security Deep Dive: From Brute Force to Salted Hashing (With Java Examples) Why AI Systems Don’t Fail — They Drift Spilling beans for how i learn for exam😁"Reinforcement Learning Cheat Sheet" I Replaced Chrome with Safari for AI Browser Automation. Here's What Broke (and What Finally Worked) How Python Borrows Other People's Work The $40 Architecture: Processing 1 Billion API Requests with 99.99% Uptime Vibe Coding: A Workflow Guide (From Zero to SaaS) Most webhook security guides protect the wrong side. The scary part is delivery. Headless CMS for TanStack Start: Build a Blog with Cosmic EU Age Verification App "Hacked in 2 Minutes" — What Actually Happened Comfy Cloud’s delete function does not actually remove files Running AI Models on GPU Cloud Servers: A Beginner Guide Event-driven media intelligence with AWS Step Functions and Bedrock I scored 500 AI prompts across 8 quality dimensions — here's what broke How to Call Google Gemini API from Next.js (Free Tier, No Backend Needed) The Portal Protocol: Reclaiming Human Connection in the Age of AI How to Fix Your Team's Scattered Knowledge Problem With a Self-Hosted Forum Intro to tc Cloud Functors: A Graph-First Mental Model for the Modern Cloud Designing Multi-Tenant Backends With Both Ownership and Team Access I Built a Neumorphic CSS Library with 77+ Components — Here's What I Learned PostgreSQL Performance Optimization: Why Connection Pooling Is Critical at Scale Cómo construí un SaaS multi-rubro para gestionar expensas en Argentina con FastAPI + Vue 3 🚀 I Built an Ethical Hacking Scanner Tool – Open Source Project I Replaced /usage and /context in Claude Code With a Single Statusline A Pythonic Way to Handle Emails (IMAP/SMTP) with Auto-Discovery and AI-Ready Design I Collected 8.9 Million Polymarket Price Points — Here's What I Found About How Markets Really Move EcoTrack AI — Carbon Footprint Tracker & Dashboard Everyone's Using AI. No One Agrees How. 5 self-hosted ebook managers worth trying in 2026 Building Your First AI Agent with LangChain: From Chatbot to Autonomous Assistant Common SOC 2 Failures (Real World) Stop Vibe-Checking Your AI App: A Practical Guide to Evals How to Use SonarQube and SonarScanner Locally to Level Up Your Code Quality Your Next To-Do App Is Dead — I Replaced Mine with an OpenClaw AI Sign a Nostr event in 60 lines of Python using coincurve — no nostr-sdk, no nbxplorer, no rust toolchain ITGC Audit Explained Like You’re in Big 4 Patch Tuesday abril 2026: Microsoft parcha 163 vulnerabilidades y un zero-day en SharePoint Stop scraping everything: a better way to track competitor price changes Listing on MCPize + the Official MCP Registry while routing payments OUTSIDE the marketplace — how I kept 100% of my x402 revenue Building an AI-Powered Risk Intelligence System Using Serverless Architecture Why We Ripped Function Overloading Out of Our AI Toolchain Testing AI-Generated Code: How to Actually Know If It Works SaaS Churn Is Killing Your Business. Here Is What to Do About It (Without a Support Team) The Speed of AI Is No Longer Linear - And Self-Improving Models Are Why How to Implement RBAC for MCP Tools: A Practical Guide for Engineering Teams From Standard Quote to Persuasive Proposal: AI Automation for Arborists I built a CLI that scaffolds complete multi-tenant SaaS apps Axios CVE-2025–62718: The Silent SSRF Bug That Could Be Hiding in Your Node.js App Right Now The dashboard that ended our friendship Data Pipelines Explained Simply (and How to Build Them with Python) The Hidden Cost of AI Systems Nobody Talks About. undefined vs undeclared, and how typeof behaves Switching from file-based jobs to NATS/Kafka in Rust without changing code io_uring Adventures: Rust Servers That Love Syscalls Why Agentic AI is Killing the Traditional Database The POUR principles of web accessibility for developers and designers Quantum Neural Network 3D — A Deep Dive into Interactive WebGL Visualization How To Install Caveman In Codex On macOS And Windows Automation Pipeline Reliability: Why Your Workflow Breaks When Nobody Is Watching I Built an 'Open World' AI Coding Agent — It Works From ANY Folder From Freelancing to Product: A Tech Service Company's SaaS Transformation China's AI Giants: Adding Tencent Hunyuan & ByteDance Doubao to AI University (74 Providers) On the Vibe Coders and Their Lies clerk: Auto-Summarize Your Claude Code Sessions AI Weekly — 2026/04/10–04/17 | The Model Lockdown Is Here, but the Toolchain Is the Real Battleground AI 週報 — 2026/04/10–2026/04/17 模型封鎖潮來了,但工具鏈才是真戰場 Maybe this is how Open-Source apps are born... 🚀 Fine-Tune LLMs with LoRA and QLoRA: 2026 Guide tRPC v11 + Next.js App Router: End-to-End Type Safety Without the Boilerplate ShadCN UI in 2026: Why I Stopped Installing Component Libraries and Started Owning My Components SaaS Billing in React Server Components: Stripe + Supabase Without a Single `useEffect` Join our DEV Weekend Challenge — $1,000 in Prizes Across TEN winners! Submissions Due April 20 at 6:59 AM UTC. Implementing FSRS Spaced Repetition in Flutter + Supabase — Adding Memory Science to an AI Learning App "I Texted My Localhost From the Train — Claude Code Fixed the Bug Before I Got Home" I Built a Sales Prep AI and It Went Deeper Than Expected Design to Code #2: One JSON, Eleven Outputs Solving the 100M-Row Problem: A Summary Table Pattern for High-Volume Push Notification Logs Flutter Web With Wasm: What Actually Changes For Developers I Built 50 Royalty-Free Soundtracks for My Side Project in a Weekend Using AI Music Generation The Vibe Coding Security Checklist: 7 Things to Check Before You Ship Stop Letting Googlebot Guess Fix Your React App's SEO Right Desconstruindo o Streaming do LinkedIn: Como Criar um Engine de Extração de Vídeo de Alta Performance com HLS e FFmpeg (EDA Part-1) EDA (Exploratory Data Analysis) Explained With Real Life — Why Looking at Your Data Is the Most Important Step in Machine Learning Brand Relationship Management at Scale: Our 4-Touch Outreach System for 200+ Brands Why String.fromEnvironment() Might Return an Empty String in Dart JGuardrails 1.0.0 — Hardening Java LLM Apps Against Jailbreaks, Toxicity, and Prompt Injection Plan and Schedule a Full Week of Threads Content From One Claude Conversation Coding Cat Oran Ep3, Five Tables Changed Everything Updated: BFF Pattern I'm done watching freelancers get buried by 200 proposals. So I'm building the alternative. This is my first post BFS Algorithm in Java Step by Step Tutorial with Examples Tracking LLM Pricing Monthly: An Open Dataset for 22 AI Models How We Measure Content ROI on a Comparison Site: Revenue Attribution Without Perfect Data Introducing Nova AI Ops: The AI-Native Operating System for SRE Teams I built a free desktop video downloader for Windows — Grabbit How Talkie OCR Helps Vision-Impaired & Dyslexic Users Read the World Around Them VRCFaceTracking安装和iPhone面捕配置教程,有bug Even CrowdStrike Can't See Your Agents The Automation Gold Rush: What n8n Workflows and Claude Are Opening Up for Developers Right Now
超越沙箱:为 AI Agent 构建持久化运行时
cognitalk · 2026-06-25 · via DEV Community

cognitalk

http://www.youtube.com/watch?v=WFyFCFC15dE
这篇演讲的题目是 《Beyond Sandboxes: Architecting Durable Runtimes for AI Agents》(超越沙箱:为 AI Agent 构建持久化运行时),演讲者是 Orkes 的联合创始人兼 CTO Virein Baraiya

演讲的核心观点是:在生产环境中运行 AI Agent,仅仅依靠“沙箱(Sandbox)”来实现隔离和保护是不够的,更需要一个“持久化运行时(Durable Runtime)”来保证 Agent 的状态可持久化、崩溃后可恢复、过程可审计。

以下是演讲的核心内容总结:


一、 核心痛点:为什么传统的“沙箱”和“内存循环”不够用?

传统的 Agent 开发通常把 Agent 当作一个在内存循环(LLM in the loop)中运行的实体:LLM 观察状态 $\rightarrow$ 决定调用工具 $\rightarrow$ 更新状态 $\rightarrow$ 继续循环 08:45。但这种模式在生产环境中存在巨大缺陷:

  • 无法应对长周期任务: 如果一个 Agent 需要运行几天、几周,甚至包含“人工介入(Human-in-the-loop)”环节(例如等待审批或用户提交文档),把整个进程挂在微型虚拟机或沙箱里会极大浪费 CPU 和内存资源 05:54
  • 缺乏崩溃恢复能力(Crash Recovery): 传统的内存循环在进程奔溃(如网络超时、系统断电)后,内存中的状态和上下文就会全部丢失,无法精准从断点恢复 07:38, 09:10
  • 多 Agent 协同复杂: 在沙箱中运行的 Agent 彼此之间通信时,需要处理复杂的 IP 寻址、重试逻辑等胶水代码 06:26

二、 核心架构思想:将“推理”与“执行”分离

演讲者提出,应当把 Agent 的关注点进行分离 04:32

  • 沙箱用于保护“执行(Actions)”: 专门用来运行由 Agent 生成的、可能存在安全风险或 Bug 的工具代码。
  • 持久化框架用于编排“推理(Reasoning)”: LLM 只负责提出计划(例如:“我建议调用天气工具,参数是波士顿”),而不直接执行它。真正的工具调用由底层的运行时(Runtime)系统接管 05:03
  • Agent 本质上是动态构建的“Sagas(长事务工作流)”: 传统工作流图是静态预定义的,而 Agent 是由 LLM 随着运行动态构建图(Late-bound Sagas),但它们同样需要底层的持久化基座来记录每一步 11:04

三、 解决方案:Conductor 与 Agent Span

为了落地这一架构,演讲者介绍了两个开源项目:

  1. Netflix Conductor(微服务工作流引擎): 这是演讲者团队在 Netflix 期间开发的(现由 Orkes 维护)。它作为一个持久化基座,通过数据库(如 Postgres、Redis)将 Agent 运行的每一步(LLM 调用、工具调用、状态流转、人工输入)都写入账本(Ledger) 12:28, 15:03
  2. 按需挂起: 遇到人工介入或长周期等待时,工作流会暂停,释放全部 CPU 和内存,直到下一个事件触发唤醒,即使隔了 6 个月也能继续 18:03, 19:11

  3. Agent Span(Agent 运行时): 构建在 Conductor 之上的 Agent 专用运行时。它像一个编译器,可以把目前市面上主流的 11 种 Agent SDK(如 LangGraph、OpenAI Agents、Vercel AI SDK 等)定义的 Agent,在不修改业务代码的前提下,直接转换为 Conductor 的持久化工作流运行 14:09, 23:34

四、 带来的核心价值

  • 确定性的护栏(Guardrails): 护栏由框架层硬性控制,而不是让 LLM 决定是否调用,防止 LLM 因幻觉绕过安全检查(例如误删数据库) 10:05
  • 全量审计与回放(Audit & Replay): 即使在几个月后,合规部门想要审计某个 Agent 为什么做出了某项决策,开发者都可以查看完整的执行记录,甚至可以通过 Mock 掉 LLM 的方式来“回放”整个执行过程 19:34, 24:17
  • 更高效的测试与 Eval: 可以改变某一步 LLM 的输出,来观察后续工具链和业务逻辑的走向,让 Agent 测试更具确定性 20:13

五、 落地建议与技术护城河

演讲者在最后给出了几点实践建议 21:56

  1. 工具的幂等性: 由于系统采用至少一次交付(At-least-once)的队列,必须注意处理非幂等工具(如转账)的重试逻辑。
  2. 上下文腐化(Context Rot): 随着 Agent 循环次数增多,账本和上下文会极速膨胀。建议将复杂的 Agent 拆分为多个子 Agent(Sub-agents),让每个子 Agent 只维护各自相关的局部上下文 22:58
  3. 企业的真正护城河(The Moat): 基础模型(如 OpenAI、Anthropic)会不断迭代甚至把一些外部过程内化,但它们永远拿不到企业自身的业务上下文(Context)。通过持久化运行时捕获到的、企业独有的业务执行流程和决策上下文,才是企业在 AI 时代真正的壁垒 24:45



通俗易懂的解释下Conductor 与 Agent Span

我们可以用一个通俗的类比来理解 ConductorAgent Span 的关系。

把大模型(Agent)想象成一个“极其聪明但记性极差、随时可能失忆的侦探”。他办案时需要频繁打电话问专家(调用 LLM)、翻查档案(调用工具),甚至需要等领导审批(人工介入)。

这时候,演讲里提到的这两个开源项目就扮演了不同的角色:


1. Netflix Conductor:不知疲倦的“书记官”兼“时间暂停器”

传统的做法是让侦探自己一边查案一边在脑子里记进度。如果中间侦探突然“晕倒”(程序崩溃、网络断开),案子就得从头查起。

  • 它的核心功能(账本机制): Conductor 就像是给侦探配了一个形影不离的“书记官”。侦探每说一句话、每打一个电话、每翻一页档案,书记官都会死死地记在小本本(数据库)上。就算侦探突然晕倒,换个新侦探来,翻开小本本就能立刻接着上一步往下查,绝不浪费时间。
  • 按需挂起(时间暂停): 办案过程中,侦探需要等领导审批一个文件,而领导去度假了。如果让侦探一直在办公室干等着,公司得天天给他发高额薪水(白白占用服务器 CPU 和内存)。 有了 Conductor,书记官会直接让侦探“原地消失”(释放所有计算资源),把案子封存起来。等 6 个月后领导度假回来一签字(触发唤醒事件),书记官啪的一下把案子重新打开,侦探瞬间“复活”,不偏不差地从等签字的那一秒继续往下干。这期间没花公司一分钱的资源费。

2. Agent Span:高级的“同声传译员”(翻译官)

侦探(Agent)有很多流派,有的说英语(LangGraph 框架写出来的),有的说法语(OpenAI SDK 写出来的),他们办案的套路各不相同。而书记官 Conductor 是个老古板,它只听得懂极其严谨的、工厂流水线式的“标准公文”(Deterministic Workflow)。

  • 它的核心功能(编译器/运行时): Agent Span 就是一个“超级同声传译”。 你不需要逼着侦探去学怎么跟老古板书记官沟通(不用修改你现有的 Agent 业务代码)。你用市面上任何主流框架(比如 LangGraph 或是 OpenAI 官方 SDK)写好你的 Agent,Agent Span 会在底层自动把侦探嘴里那些天马行空的推理和想法,翻译成老古板书记官能听懂的“标准公文”并交给它去记录和执行。

总结一下

  • Conductor 解决了 Agent 在生产环境中的“生存问题”——通过记笔记和高阶省电模式,保证 Agent 任务死不掉、开销低、随时可查。
  • Agent Span 解决了 Agent 的“适配问题”——让你用市面上最流行的姿势去写 Agent,然后无缝享受 Conductor 带来的超强稳定性,中间不需要掉一层皮去重构代码。