<?xml version="1.0" encoding="UTF-8"?><?xml-stylesheet href="/rss/styles.xsl" type="text/xsl"?><rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom"><channel><title>紫牙的博客 - 林健(紫牙)前端架构师</title><description>林健(紫牙)的博客,又名紫牙的博客。11 年前端架构经验,专注于企业级中后台、微前端、跨端开发、性能优化与 AI 工程实践。</description><link>https://lovanya.github.io</link><language>zh-CN</language><atom:link href="https://lovanya.github.io/rss.xml" rel="self" type="application/rss+xml" xmlns:atom="http://www.w3.org/2005/Atom"/><item><title>Agent 架构：ReAct、Plan-and-Execute 与多 Agent 协作</title><link>https://lovanya.github.io/blog/agent-architecture-patterns</link><guid isPermaLink="true">https://lovanya.github.io/blog/agent-architecture-patterns</guid><description>从单步工具调用到完整 agent 思考循环——ReAct / Plan-and-Execute / Multi-Agent 三大主流架构与生产权衡</description><pubDate>Tue, 07 Jul 2026 00:00:00 GMT</pubDate><category>llm</category><category>ai-engineering</category><category>agent</category><author>紫牙(林健)</author></item><item><title>AI 时代的新 UI：当下与近未来的界面应该长什么样</title><link>https://lovanya.github.io/blog/ai-ui-new-era</link><guid isPermaLink="true">https://lovanya.github.io/blog/ai-ui-new-era</guid><description>Generative UI、Agent 优先、对话即操作——AI 时代 UI 设计的范式转移、实战模式与近未来趋势</description><pubDate>Tue, 07 Jul 2026 00:00:00 GMT</pubDate><category>llm</category><category>ai-engineering</category><category>ui</category><author>紫牙(林健)</author></item><item><title>给静态博客加上 AI 助手：Pagefind + Gemini 纯前端方案</title><link>https://lovanya.github.io/blog/build-ai-chatbot-astro</link><guid isPermaLink="true">https://lovanya.github.io/blog/build-ai-chatbot-astro</guid><description>纯前端方案给你的 Astro 博客加一个 AI 聊天悬浮机器人，无需后端部署，支持站内搜索、页面导航和 AI 对话。</description><pubDate>Wed, 08 Jul 2026 00:00:00 GMT</pubDate><category>astro</category><category>react</category><category>ai</category><category>pagefind</category><category>gemini</category><category>frontend</category><category>tutorial</category><author>紫牙(林健)</author></item><item><title>Function Calling 与 Tool Use：从协议到工程实现</title><link>https://lovanya.github.io/blog/function-calling-and-tools</link><guid isPermaLink="true">https://lovanya.github.io/blog/function-calling-and-tools</guid><description>LLM 怎么调用你的代码——JSON schema 设计、tool 选择、错误处理、Zod 强校验、生产工程陷阱</description><pubDate>Tue, 07 Jul 2026 00:00:00 GMT</pubDate><category>llm</category><category>ai-engineering</category><category>agent</category><author>紫牙(林健)</author></item><item><title>Embedding 模型与向量数据库选型实战</title><link>https://lovanya.github.io/blog/embedding-and-vector-db</link><guid isPermaLink="true">https://lovanya.github.io/blog/embedding-and-vector-db</guid><description>RAG 的两个核心依赖怎么选——BGE / OpenAI / Cohere / Voyage 选哪个？Pinecone / Milvus / pgvector / Qdrant / Weaviate 怎么挑？</description><pubDate>Tue, 07 Jul 2026 00:00:00 GMT</pubDate><category>llm</category><category>ai-engineering</category><category>embedding</category><author>紫牙(林健)</author></item><item><title>LLM 成本与性能优化：Caching、Streaming 与模型路由</title><link>https://lovanya.github.io/blog/llm-cost-and-performance</link><guid isPermaLink="true">https://lovanya.github.io/blog/llm-cost-and-performance</guid><description>Token 账单失控怎么治——prompt cache、semantic cache、流式优化、模型路由、自动降级 5 大优化手段</description><pubDate>Tue, 07 Jul 2026 00:00:00 GMT</pubDate><category>llm</category><category>ai-engineering</category><category>optimization</category><author>紫牙(林健)</author></item><item><title>LLM 应用的评估体系：Metrics、LLM-as-Judge 与 A/B 测试</title><link>https://lovanya.github.io/blog/llm-evaluation</link><guid isPermaLink="true">https://lovanya.github.io/blog/llm-evaluation</guid><description>当 observability 有了数据，怎么判断哪个 prompt 更好——评估指标、LLM-as-Judge 实战、回归测试集、Braintrust / Ragas 实践</description><pubDate>Tue, 07 Jul 2026 00:00:00 GMT</pubDate><category>llm</category><category>ai-engineering</category><category>evaluation</category><author>紫牙(林健)</author></item><item><title>大模型入门：Transformer 与 LLM 是怎么工作的</title><link>https://lovanya.github.io/blog/llm-fundamentals</link><guid isPermaLink="true">https://lovanya.github.io/blog/llm-fundamentals</guid><description>写给前端架构师的 LLM 底层速通——Transformer、训练三阶段、推理机制，一次讲清楚</description><pubDate>Fri, 03 Jul 2026 00:00:00 GMT</pubDate><category>llm</category><category>ai-engineering</category><category>transformer</category><author>紫牙(林健)</author></item><item><title>LLM 应用的 Guardrails：Prompt Injection 与输出安全</title><link>https://lovanya.github.io/blog/llm-guardrails</link><guid isPermaLink="true">https://lovanya.github.io/blog/llm-guardrails</guid><description>LLM 安全的 5 道防线——prompt injection 防御、Guardrails AI 实战、输入输出校验、PII 检测、危险操作审批</description><pubDate>Tue, 07 Jul 2026 00:00:00 GMT</pubDate><category>llm</category><category>ai-engineering</category><category>security</category><author>紫牙(林健)</author></item><item><title>LLM 应用的 Memory 系统设计</title><link>https://lovanya.github.io/blog/llm-memory-systems</link><guid isPermaLink="true">https://lovanya.github.io/blog/llm-memory-systems</guid><description>短期 / 长期 / 语义记忆怎么搭——三种记忆的工程实现、Mem0 架构、上下文窗口管理与生产坑</description><pubDate>Tue, 07 Jul 2026 00:00:00 GMT</pubDate><category>llm</category><category>ai-engineering</category><category>memory</category><author>紫牙(林健)</author></item><item><title>LLM 可观测性：Tracing、Logging 与调试</title><link>https://lovanya.github.io/blog/llm-observability</link><guid isPermaLink="true">https://lovanya.github.io/blog/llm-observability</guid><description>LLM 应用上线后怎么知道它在干什么——Langfuse 架构、OpenTelemetry GenAI 语义、token 成本监控、生产排错实战</description><pubDate>Tue, 07 Jul 2026 00:00:00 GMT</pubDate><category>llm</category><category>ai-engineering</category><category>observability</category><author>紫牙(林健)</author></item><item><title>从 Demo 到 Production：LLM 应用的工程化与团队协作</title><link>https://lovanya.github.io/blog/llm-production-engineering</link><guid isPermaLink="true">https://lovanya.github.io/blog/llm-production-engineering</guid><description>LLM 团队的 5 大工程化实践——prompt 版本化、CI/CD、灰度发布、多模型切换、团队工作流（Monorepo vs 多仓）</description><pubDate>Tue, 07 Jul 2026 00:00:00 GMT</pubDate><category>llm</category><category>ai-engineering</category><category>production</category><author>紫牙(林健)</author></item><item><title>Git 分支管理完全指南：从入门到团队协作</title><link>https://lovanya.github.io/blog/macos-git-branch</link><guid isPermaLink="true">https://lovanya.github.io/blog/macos-git-branch</guid><description>2026 年最新的 Git 分支管理指南，涵盖核心命令、分支策略、冲突解决和团队协作最佳实践。</description><pubDate>Mon, 12 Nov 2018 00:00:00 GMT</pubDate><category>git</category><category>macos</category><category>terminal</category><category>developer-tools</category><author>紫牙(林健)</author></item><item><title>RAG 检索增强生成：从基础 RAG 到生产级架构</title><link>https://lovanya.github.io/blog/rag-from-naive-to-production</link><guid isPermaLink="true">https://lovanya.github.io/blog/rag-from-naive-to-production</guid><description>不止是&quot;塞 context 进 prompt&quot;——RAG 的演进路线、advanced RAG 五大技法、生产架构的工程陷阱</description><pubDate>Tue, 07 Jul 2026 00:00:00 GMT</pubDate><category>llm</category><category>ai-engineering</category><category>rag</category><author>紫牙(林健)</author></item><item><title>Oh My Zsh 终极配置指南：从入门到高效开发</title><link>https://lovanya.github.io/blog/oh-my-zsh</link><guid isPermaLink="true">https://lovanya.github.io/blog/oh-my-zsh</guid><description>2026 年最新的 Oh My Zsh 安装指南（macOS + Windows）、主题、插件配置、性能优化和开发效率提升。</description><pubDate>Sat, 22 Dec 2018 00:00:00 GMT</pubDate><category>terminal</category><category>macos</category><category>windows</category><category>zsh</category><category>developer-tools</category><author>紫牙(林健)</author></item><item><title>闭源 vs 开源 LLM：架构师选型决策框架</title><link>https://lovanya.github.io/blog/llm-selection-framework</link><guid isPermaLink="true">https://lovanya.github.io/blog/llm-selection-framework</guid><description>不再&quot;凭感觉选模型&quot;——5 个评估维度、3 套主流选型、API vs 自托管成本算账、混合路由策略</description><pubDate>Fri, 03 Jul 2026 00:00:00 GMT</pubDate><category>llm</category><category>ai-engineering</category><category>selection</category><author>紫牙(林健)</author></item><item><title>Token、Context Window 与 Prompt Engineering 核心模式</title><link>https://lovanya.github.io/blog/tokens-context-and-prompts</link><guid isPermaLink="true">https://lovanya.github.io/blog/tokens-context-and-prompts</guid><description>把 LLM 底层知识落地为 prompt 设计技法 — tokenization、context 边界、5 个必会的 prompt 模式</description><pubDate>Fri, 03 Jul 2026 00:00:00 GMT</pubDate><category>llm</category><category>ai-engineering</category><category>prompt-engineering</category><author>紫牙(林健)</author></item><item><title>Starship 终端提示符安装与使用指南</title><link>https://lovanya.github.io/blog/starship-prompt</link><guid isPermaLink="true">https://lovanya.github.io/blog/starship-prompt</guid><description>Starship 是一个 Rust 编写的跨平台终端提示符，毫秒级启动。本文介绍在 macOS 和 Windows 上的安装和基本配置。</description><pubDate>Tue, 07 Jul 2026 00:00:00 GMT</pubDate><category>terminal</category><category>macos</category><category>windows</category><category>starship</category><category>developer-tools</category><author>紫牙(林健)</author></item></channel></rss>