[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"consumer-news-discovery-%5B%22%22%2Cnull%5D":3},{"categories":4,"categoryError":29,"result":30},[5,9,13,17,21,25],{"id":6,"name":7,"articleCount":8},109,"人工智能",34,{"id":10,"name":11,"articleCount":12},110,"软件开发",22,{"id":14,"name":15,"articleCount":16},111,"综合技术",14,{"id":18,"name":19,"articleCount":20},113,"数据与架构",7,{"id":22,"name":23,"articleCount":24},112,"云原生与运维",5,{"id":26,"name":27,"articleCount":28},114,"硬件与物联网",4,"",{"page":31,"items":235},{"records":32,"total":231,"current":232,"size":233,"pages":234},[33,49,58,67,77,87,96,106,116,125,135,145,154,164,174,184,194,204,213,222],{"id":34,"kind":35,"targetType":36,"targetId":37,"subtype":35,"typeLabel":38,"title":39,"subtitle":40,"summary":41,"coverUrl":42,"badgeText":29,"href":43,"sourceName":40,"meta":44,"metrics":46,"tags":47,"resolved":48},"NEWS_ARTICLE:927","news","NEWS_ARTICLE",927,"资讯","HelloCrab - 短视频开源爬虫，仅供学习参考","博客园","HelloCrab 基于 Avalonia、Playwright、AI与 FFmpeg 的跨平台桌面采集器，支持9大平台，以及 Android、iOS、Browser 远程控制端。 Made By ChatGPT & Vincent with ❤ 平台 是否接入 哔哩哔哩 ✅ 抖音 ✅ 快手","https:\u002F\u002Fwww.cnblogs.com\u002Fhupo376787\u002Fp\u002FScreenshot\u002FWindows.jpg","\u002Fnews\u002F927",[45,11],"2026",{},[11],true,{"id":50,"kind":35,"targetType":36,"targetId":51,"subtype":35,"typeLabel":38,"title":52,"subtitle":40,"summary":53,"coverUrl":29,"badgeText":29,"href":54,"sourceName":40,"meta":55,"metrics":56,"tags":57,"resolved":48},"NEWS_ARTICLE:929",929,".NET 异常处理的\"暗门\"：代码里写满 catch，你依然能抓住它——从一个 AI Agent 运行时的源码说起","一个健壮的系统，必然到处都是有意的 catch；异常被消化不等于问题不存在。 观测与韧性，是一个硬币的两面——降级逻辑保证系统不崩，FirstChance 保证你能看见它为什么降级。","\u002Fnews\u002F929",[45,11],{},[11],{"id":59,"kind":35,"targetType":36,"targetId":60,"subtype":35,"typeLabel":38,"title":61,"subtitle":40,"summary":62,"coverUrl":29,"badgeText":29,"href":63,"sourceName":40,"meta":64,"metrics":65,"tags":66,"resolved":48},"NEWS_ARTICLE:928",928,"架构师化繁为简，执行者化简为繁","新手改三天，你改三行——反而是你显得更不重要。因为化繁为简做得越纯熟，产出看起来越小。这篇聊聊两种能力的辩证关系，以及为什么'看不见'的那部分工作，恰恰是最难的部分。","\u002Fnews\u002F928",[45],{},[],{"id":68,"kind":35,"targetType":36,"targetId":69,"subtype":35,"typeLabel":38,"title":70,"subtitle":40,"summary":71,"coverUrl":72,"badgeText":29,"href":73,"sourceName":40,"meta":74,"metrics":75,"tags":76,"resolved":48},"NEWS_ARTICLE:930",930,"基于 vLLM+Nginx 构建负载均衡推理集群","企业内部私有环境部署大模型推理集群时，很容易遇到流量调度混乱、节点负载失衡、会话上下文丢失、接口缺少鉴权防护等一系列问题，单 vLLM 推理节点难以支撑并发请求。本文基于 Ubuntu 22.04 系统环境，搭建 Nginx + vLLM-Semantic-Router + vLLM-Router","https:\u002F\u002Fimg2024.cnblogs.com\u002Fblog\u002F1379525\u002F202609\u002F1379525-20260910165534385-2022408098.png","\u002Fnews\u002F930",[45,7],{},[7],{"id":78,"kind":35,"targetType":36,"targetId":79,"subtype":35,"typeLabel":38,"title":80,"subtitle":40,"summary":81,"coverUrl":82,"badgeText":29,"href":83,"sourceName":40,"meta":84,"metrics":85,"tags":86,"resolved":48},"NEWS_ARTICLE:932",932,"2026年AI编程工具大全，33个主流工具一次看懂","事情是这样的，前两天看到一张图，是某个社区官网的「支持的工具」清单，我数了数，整整31个AI编程工具。 两年前这份清单撑死5个，现在直接31个，而且我居然每一个都认识。。。 干脆整理成一篇，顺手把最近字节的TraeWork和豆包工作也补了进来，凑成33个。 今天给大家推荐一遍，每个工具说说它是干什么","https:\u002F\u002Fimage.kjdaohang.com\u002Fimg\u002F20260909210838518.png","\u002Fnews\u002F932",[45,7],{},[7],{"id":88,"kind":35,"targetType":36,"targetId":89,"subtype":35,"typeLabel":38,"title":90,"subtitle":40,"summary":91,"coverUrl":29,"badgeText":29,"href":92,"sourceName":40,"meta":93,"metrics":94,"tags":95,"resolved":48},"NEWS_ARTICLE:931",931,"SH 中文化样例数据使用手册","在数据库演示与 PoC 场景中，Oracle 自带的 SH 示例模式虽然经典，但英文维度数据往往让国内演示效果打折扣。笔者整理了一套方案：保留 SH 标准英文对象名，同时装载中文化维度数据，并按参数生成可复现的销售历史数据，方便个人测试与概念验证。 01 | 环境准备 使用前请确认满足以下条件： O","\u002Fnews\u002F931",[45],{},[],{"id":97,"kind":35,"targetType":36,"targetId":98,"subtype":35,"typeLabel":38,"title":99,"subtitle":40,"summary":100,"coverUrl":101,"badgeText":29,"href":102,"sourceName":40,"meta":103,"metrics":104,"tags":105,"resolved":48},"NEWS_ARTICLE:933",933,"写给 C++ 工程师的 OpenClaw.NET 上手指南：用你熟悉的 C++ 思维，跑起一个生产级 AI Agent","它像一个「基于 boost.asio + REST 端点的常驻服务」——对外是 HTTP \u002F WebSocket \u002F 各 IM 的 webhook，对内跑着一个能调工具、读写记忆、跨渠道对话的 AI Agent——只不过协程不用你手写 promise_type，内存不用你管 new\u002Fdelete。","https:\u002F\u002Fimg2024.cnblogs.com\u002Fblog\u002F510\u002F202609\u002F510-20260905070811424-1837144874.jpg","\u002Fnews\u002F933",[45,11],{},[11],{"id":107,"kind":35,"targetType":36,"targetId":108,"subtype":35,"typeLabel":38,"title":109,"subtitle":40,"summary":110,"coverUrl":111,"badgeText":29,"href":112,"sourceName":40,"meta":113,"metrics":114,"tags":115,"resolved":48},"NEWS_ARTICLE:935",935,"[Agent Memory \u002F 强化学习] MemPO源码学习笔记 ---（1）--- 总体","[Agent Memory \u002F 强化学习] MemPO源码学习笔记 （1） 总体 目录[Agent Memory \u002F 强化学习] MemPO源码学习笔记 （1） 总体0x00 概要0x01 基础 & 背景1.1 用RL训练记忆系统的要点1.2 主要难点1.3 主要思路1.4 RL训练方案1.","https:\u002F\u002Fimg2024.cnblogs.com\u002Fblog\u002F1850883\u002F202609\u002F1850883-20260906190854637-986949885.jpg","\u002Fnews\u002F935",[45,7],{},[7],{"id":117,"kind":35,"targetType":36,"targetId":118,"subtype":35,"typeLabel":38,"title":119,"subtitle":40,"summary":120,"coverUrl":29,"badgeText":29,"href":121,"sourceName":40,"meta":122,"metrics":123,"tags":124,"resolved":48},"NEWS_ARTICLE:934",934,"Memory 记忆设计讨论：为什么 Agent Memory 不能只靠向量数据库？","这是 Agent Memory 系列的第二篇文章。上一篇先给出了一个总体判断：Memory 是一套让 Agent 能够持续使用历史上下文的能力。 这篇只讨论一个经常出现的问题： 既然 Agent 需要从历史信息中找内容，接入向量数据库不就可以了吗？ 我的答案是：向量数据库很有用，但它只解决了召回问题","\u002Fnews\u002F934",[45],{},[],{"id":126,"kind":35,"targetType":36,"targetId":127,"subtype":35,"typeLabel":38,"title":128,"subtitle":40,"summary":129,"coverUrl":130,"badgeText":29,"href":131,"sourceName":40,"meta":132,"metrics":133,"tags":134,"resolved":48},"NEWS_ARTICLE:936",936,"Claude Code 8问，最狠难题怎么破","大家已经不太问「Claude Code 是什么」「MCP 怎么安装」这种入门题，更多是下面这种让人抓头发的问题，Skill 明明装了，为什么不触发；MCP 明明注册了，为什么 Claude Code 偏不用；代码图谱接上了，结果它照着旧地图改代码；PDF Token 省下来了，图表却答错了。 这才是","https:\u002F\u002Fimg2024.cnblogs.com\u002Fblog\u002F735767\u002F202609\u002F735767-20260910172715276-211335062.png","\u002Fnews\u002F936",[45],{},[],{"id":136,"kind":35,"targetType":36,"targetId":137,"subtype":35,"typeLabel":38,"title":138,"subtitle":40,"summary":139,"coverUrl":140,"badgeText":29,"href":141,"sourceName":40,"meta":142,"metrics":143,"tags":144,"resolved":48},"NEWS_ARTICLE:938",938,"[Python\u002F测试] pytest：简洁、可扩展的 Python 测试框架","0 序 缘起: pytest 做单元测试 1 概述 产品介绍 定位：pytest 是一个面向 Python 的开源测试框架，主打“用最小代价写出可读性高的测试”，同时可扩展到复杂的功能测试、集成测试和端到端测试。 核心理念： 用普通 assert 表达断言，由 pytest 提供失败时的详细信息。","https:\u002F\u002Fimg2024.cnblogs.com\u002Fblog\u002F1173617\u002F202609\u002F1173617-20260910183944954-2024428266.png","\u002Fnews\u002F938",[45],{},[],{"id":146,"kind":35,"targetType":36,"targetId":147,"subtype":35,"typeLabel":38,"title":148,"subtitle":40,"summary":149,"coverUrl":29,"badgeText":29,"href":150,"sourceName":40,"meta":151,"metrics":152,"tags":153,"resolved":48},"NEWS_ARTICLE:937",937,"实战攻防中SNI技术应用","SNI 是 TLS 协议里一个很基础的机制，但在攻防里它的意义很特殊——因为它是明文的。 一、SNI 是什么 SNI（Server Name Indication，服务器名称指示） 是 TLS 握手的扩展字段。客户端在建立 HTTPS 连接时，会在 ClientHello 里明文告诉服务器「我要访问","\u002Fnews\u002F937",[45],{},[],{"id":155,"kind":35,"targetType":36,"targetId":156,"subtype":35,"typeLabel":38,"title":157,"subtitle":40,"summary":158,"coverUrl":159,"badgeText":29,"href":160,"sourceName":40,"meta":161,"metrics":162,"tags":163,"resolved":48},"NEWS_ARTICLE:939",939,"面向AIGC的检索增强生成综述","原文：https:\u002F\u002Fmp.weixin.qq.com\u002Fs\u002FZ3355OEqzJBQvqXONxgO0g RAG往期文章推荐 RAG效果差？7个指标让你的准确率大幅提升 RAG评测完整指南：指标、测试和最佳实践 收藏！RAG核心工具大全: 7大解析工具+向量模型+数据库+检索排序 GraphRAG开","https:\u002F\u002Fimg2024.cnblogs.com\u002Fblog\u002F3719074\u002F202609\u002F3719074-20260910153551210-1121871745.png","\u002Fnews\u002F939",[45],{},[],{"id":165,"kind":35,"targetType":36,"targetId":166,"subtype":35,"typeLabel":38,"title":167,"subtitle":40,"summary":168,"coverUrl":169,"badgeText":29,"href":170,"sourceName":40,"meta":171,"metrics":172,"tags":173,"resolved":48},"NEWS_ARTICLE:941",941,"【UE】动画蓝图参数屏幕调试工具","动画蓝图参数屏幕调试工具 这里用到的动画蓝图基类为自定义的UBaseAnimInstance 引脚 类型 默认值 \u002F 用法 Key FName 手动填写参数名称，不要加冒号 Value String 变量、转换结果、拼接结果或字符串常量均可 Key Color Linear Color 白色，包含自","https:\u002F\u002Fimg2024.cnblogs.com\u002Fblog\u002F3614909\u002F202609\u002F3614909-20260910131827752-546469957.png","\u002Fnews\u002F941",[45],{},[],{"id":175,"kind":35,"targetType":36,"targetId":176,"subtype":35,"typeLabel":38,"title":177,"subtitle":40,"summary":178,"coverUrl":179,"badgeText":29,"href":180,"sourceName":40,"meta":181,"metrics":182,"tags":183,"resolved":48},"NEWS_ARTICLE:940",940,"LLama-Factory 实现大模型LoRA-SFT微调指南","LLaMA Factory 是一款开源、低代码、一站式大语言与多模态模型微调框架，用于降低大模型的微调落地门槛。框架兼容 Qwen、LLaMA、ChatGLM、LLaVA 等上百款主流开源模型，支持增量预训练、SFT 监督微调、DPO、KTO、ORPO 等多种训练对齐方案，原生集成 LoRA、QLo","https:\u002F\u002Fimg2024.cnblogs.com\u002Fblog\u002F1379525\u002F202609\u002F1379525-20260909140111199-832484524.png","\u002Fnews\u002F940",[45,7],{},[7],{"id":185,"kind":35,"targetType":36,"targetId":186,"subtype":35,"typeLabel":38,"title":187,"subtitle":40,"summary":188,"coverUrl":189,"badgeText":29,"href":190,"sourceName":40,"meta":191,"metrics":192,"tags":193,"resolved":48},"NEWS_ARTICLE:942",942,"软件不是从数据开始，而是从现实开始 | KDC 系列 01","当 AI 应用开始在运行时解释数据、选择工具并影响业务状态时，数据表示中的遗漏和歧义可能直接转化为行动风险。本文从退款案例出发，提出 Reality First 的观察顺序：先明确系统面对的领域现实，再讨论现实模型、数字表示以及验证反馈如何共同支撑软件的判断与行动。","https:\u002F\u002Fpic2.zhimg.com\u002F80\u002Fv2-aedcaa5b4794add531df0485993b1aed_720w.webp","\u002Fnews\u002F942",[45],{},[],{"id":195,"kind":35,"targetType":36,"targetId":196,"subtype":35,"typeLabel":38,"title":197,"subtitle":40,"summary":198,"coverUrl":199,"badgeText":29,"href":200,"sourceName":40,"meta":201,"metrics":202,"tags":203,"resolved":48},"NEWS_ARTICLE:943",943,"195 次合并、13 个新连接器！Apache SeaTunnel 8 月有哪些新进展？","Apache SeaTunnel 8 月社区持续高频迭代！本月共合并 195 次提交，60 位贡献者参与贡献，新增 13 个连接器，进一步拓展数据源生态；Zeta 可观测性全面升级，动态日志、Worker 资源监控、Job 状态追踪等能力持续增强；同时，Checkpoint 恢复、CDC Schem","https:\u002F\u002Fimg2024.cnblogs.com\u002Fblog\u002F3195851\u002F202609\u002F3195851-20260910103059041-1816823347.jpg","\u002Fnews\u002F943",[45],{},[],{"id":205,"kind":35,"targetType":36,"targetId":206,"subtype":35,"typeLabel":38,"title":207,"subtitle":40,"summary":208,"coverUrl":29,"badgeText":29,"href":209,"sourceName":40,"meta":210,"metrics":211,"tags":212,"resolved":48},"NEWS_ARTICLE:945",945,"CURD系统怎样做出技术含量：微内核改造","建「中心」还是「平台」？ 10年前在一个公司，公司里经常需要做线上活动。对开发来说，就是通过不同的活动页面进行注册、登录，之后会有一些个性化的小活动：领取小礼物、领取红包或者参加抽奖。 前端和后端疲于天天做各种换汤不换药的注册、登录。开发方式基本是把上次的代码拷贝一份，在此基础上做修改，来支撑新活动","\u002Fnews\u002F945",[45],{},[],{"id":214,"kind":35,"targetType":36,"targetId":215,"subtype":35,"typeLabel":38,"title":216,"subtitle":40,"summary":217,"coverUrl":29,"badgeText":29,"href":218,"sourceName":40,"meta":219,"metrics":220,"tags":221,"resolved":48},"NEWS_ARTICLE:944",944,"Text-to-SQL 已过时？DolphinX 正在重新定义 AI 问数","在推动工业场景 AI Agent 能力落地过程中，DolphinDB 选择智能问数作为具体切入点，基于 DolphinX 做了一个智能问数 Agent，尝试探索一套面向企业数据的智能问数链路应该如何构建。","\u002Fnews\u002F944",[45],{},[],{"id":223,"kind":35,"targetType":36,"targetId":224,"subtype":35,"typeLabel":38,"title":225,"subtitle":40,"summary":226,"coverUrl":29,"badgeText":29,"href":227,"sourceName":40,"meta":228,"metrics":229,"tags":230,"resolved":48},"NEWS_ARTICLE:947",947,"使用 C# 轻松管理 PDF 文件的用户权限","在文档安全管理中，通过编程控制 PDF 的访问权限（如禁止打印、禁止复制等）是常见需求。本文使用 Free Spire.PDF for .NET 免费库，基于其 PdfSecurityPolicy 类，介绍如何用简洁的 C# 代码完成权限设置。 两种密码的角色 用户密码（User Password）","\u002Fnews\u002F947",[45],{},[],200,1,20,10,[236,239,241,243,246,248,250,252,254,256,258,260,262,264,266,268,270,272,274,276],{"id":34,"kind":35,"title":39,"summary":41,"image":42,"href":43,"meta":237,"badge":38,"author":40,"tags":238},"2026 · 软件开发",[11],{"id":50,"kind":35,"title":52,"summary":53,"image":29,"href":54,"meta":237,"badge":38,"author":40,"tags":240},[11],{"id":59,"kind":35,"title":61,"summary":62,"image":29,"href":63,"meta":45,"badge":38,"author":40,"tags":242},[],{"id":68,"kind":35,"title":70,"summary":71,"image":72,"href":73,"meta":244,"badge":38,"author":40,"tags":245},"2026 · 人工智能",[7],{"id":78,"kind":35,"title":80,"summary":81,"image":82,"href":83,"meta":244,"badge":38,"author":40,"tags":247},[7],{"id":88,"kind":35,"title":90,"summary":91,"image":29,"href":92,"meta":45,"badge":38,"author":40,"tags":249},[],{"id":97,"kind":35,"title":99,"summary":100,"image":101,"href":102,"meta":237,"badge":38,"author":40,"tags":251},[11],{"id":107,"kind":35,"title":109,"summary":110,"image":111,"href":112,"meta":244,"badge":38,"author":40,"tags":253},[7],{"id":117,"kind":35,"title":119,"summary":120,"image":29,"href":121,"meta":45,"badge":38,"author":40,"tags":255},[],{"id":126,"kind":35,"title":128,"summary":129,"image":130,"href":131,"meta":45,"badge":38,"author":40,"tags":257},[],{"id":136,"kind":35,"title":138,"summary":139,"image":140,"href":141,"meta":45,"badge":38,"author":40,"tags":259},[],{"id":146,"kind":35,"title":148,"summary":149,"image":29,"href":150,"meta":45,"badge":38,"author":40,"tags":261},[],{"id":155,"kind":35,"title":157,"summary":158,"image":159,"href":160,"meta":45,"badge":38,"author":40,"tags":263},[],{"id":165,"kind":35,"title":167,"summary":168,"image":169,"href":170,"meta":45,"badge":38,"author":40,"tags":265},[],{"id":175,"kind":35,"title":177,"summary":178,"image":179,"href":180,"meta":244,"badge":38,"author":40,"tags":267},[7],{"id":185,"kind":35,"title":187,"summary":188,"image":189,"href":190,"meta":45,"badge":38,"author":40,"tags":269},[],{"id":195,"kind":35,"title":197,"summary":198,"image":199,"href":200,"meta":45,"badge":38,"author":40,"tags":271},[],{"id":205,"kind":35,"title":207,"summary":208,"image":29,"href":209,"meta":45,"badge":38,"author":40,"tags":273},[],{"id":214,"kind":35,"title":216,"summary":217,"image":29,"href":218,"meta":45,"badge":38,"author":40,"tags":275},[],{"id":223,"kind":35,"title":225,"summary":226,"image":29,"href":227,"meta":45,"badge":38,"author":40,"tags":277},[]]