Auto-sync
This commit is contained in:
75
wiki/log.md
75
wiki/log.md
@@ -18,3 +18,78 @@ Created/updated: 12 entity pages (DeepSeek, Qwen, Flux, Stable Diffusion, Hunyua
|
||||
- [7-ways-NotebookLM](sources/7-ways-NotebookLM.md)
|
||||
- Key claims: Source-Grounding 机制确保回答可溯源;Audio Overviews 支持被动学习;NotebookLM 可作项目管理系统与法律文档审查工具
|
||||
- Created: 2 concepts (Source-Grounding, 被动学习)
|
||||
|
||||
## [2026-04-15] ingest | Designing for Agentic AI
|
||||
- [Designing-for-Agentic-AI](sources/Designing-for-Agentic-AI.md)
|
||||
- Key claims: Agentic AI = 行动导向而非内容生成;5条设计原则:透明度、控制感、个性化、对话、预判;用户通过观察 AI 决策过程参与交互
|
||||
- Created: 1 concept (Agentic AI), 1 entity (Yuri Pessa)
|
||||
|
||||
## [2026-04-15] ingest | LLMs、RAG、AI Agent 三个到底什么区别
|
||||
- [LLMs-RAG-AI-Agent-三个到底什么区别](sources/LLMs-RAG-AI-Agent-三个到底什么区别.md)
|
||||
- Key claims: LLM = 天才大脑(推理);RAG = 随身图书馆助理(信息);AI Agent = 行动循环(执行);三者协同构成生产级 AI 系统
|
||||
- Created: concepts (LLM, RAG, AI Agent, AI Agent 循环)
|
||||
|
||||
## [2026-04-15] ingest | 大模型相关术语和框架总结
|
||||
- [大模型相关术语和框架总结](sources/大模型相关术语和框架总结.md)
|
||||
- Key claims: LLM ≥1B 参数门槛;MCP 协议是 LLM 与外部工具的标准化接口;大模型仅输出步骤不执行;RAG 将考试正确率从 60% 提升至 90%;vLLM 通过 PagedAttention 和连续批处理优化 GPU 利用率
|
||||
- Created: 8 concepts (MCP, vLLM, Token, Self-Healing Systems, RCA, Multi-Cloud Governance, 数据蒸馏), 0 entities (已有 DeepSeek/Manus/GPT-2/GPT-3)
|
||||
|
||||
## [2026-04-15] ingest | GitHub 上 5000 人收藏的 Vibe Coding 神级指南
|
||||
- [GitHub-Vibe-Coding-神级指南](sources/GitHub-Vibe-Coding-神级指南.md)
|
||||
- Key claims: Vibe Coding = 规划驱动 + 上下文固定 + AI 结对执行;开发者做导演,AI 工具承担体力活;推荐 Cursor + claude-opus-4.5-xhigh
|
||||
- Created: 5 entities (Cursor, Windsurf, Trae, Karpathy, vibe-coding-cn), 1 concept (Vibe Coding)
|
||||
|
||||
## [2026-04-15] ingest | How Agentic AI can help for Cloud DevOps
|
||||
- [How-Agentic-AI-for-Cloud-DevOps](sources/How-Agentic-AI-for-Cloud-DevOps.md)
|
||||
- Key claims: Agentic AI 赋能 7 大 DevOps 场景:自主检测修复、智能 IaC、成本优化、安全合规、日志分析、多租户管理、AI 增强决策
|
||||
- Created: 4 entities (AWS, GCP, Azure, Terraform, CloudWatch), 0 concepts (已有 Agentic AI)
|
||||
|
||||
## [2026-04-15] ingest | Multi-Agent System Reliability
|
||||
- [Multi-Agent-System-Reliability](sources/Multi-Agent-System-Reliability.md)
|
||||
- Key claims: 4种架构模式提升多 Agent 可靠性:Hierarchy/Consensus/Adversarial Debate/Knock-out;LLM 不可靠必须被约束验证淘汰;停止拟人化 LLM
|
||||
- Created: 4 concepts (Multi-Agent Hierarchy/Consensus/Adversarial Debate/Knock-out), 1 entity (Alex Ewerlöf)
|
||||
|
||||
## [2026-04-15] ingest | 如何写出完美的Prompt
|
||||
- [如何写出完美的Prompt(提示词)?](sources/如何写出完美的Prompt(提示词)?.md)
|
||||
- Key claims: Prompt = 人与AI协作协议;Prompt能力本质是结构化思维+精准表达;建立测试-反馈-优化闭环
|
||||
- Created: 3 concepts (结构化思维, 精准表达, Prompt工程)
|
||||
|
||||
## [2026-04-15] ingest | RAG从入门到精通
|
||||
- [RAG从入门到精通系列1:基础RAG](sources/RAG从入门到精通系列1:基础RAG.md)
|
||||
- Key claims: RAG = Indexing-Retrieval-Generation 三阶段;Embedding Model Context Window 512~8192 token 需文档切分;LangChain 简化 RAG 管道构建
|
||||
- Created: 3 concepts (RAG, Embedding, 向量数据库)
|
||||
|
||||
## [2026-04-15] ingest | YouTube RSS Feed
|
||||
- [How to Get the RSS Feed For Any YouTube Channel](sources/How to Get the RSS Feed For Any YouTube Channel.md)
|
||||
- Key claims: 通过 View Page Source 搜索 channel_id= 可获取 RSS Feed URL;无需第三方服务
|
||||
- Created: 1 concept (RSS Feed)
|
||||
|
||||
## [2026-04-15] ingest | Nano Banana 提示词框架
|
||||
- [Nano-Banana-提示词框架](sources/Nano-Banana-提示词框架.md)
|
||||
- Key claims: 9层结构化字段(Shot/Subject/Environment/Lighting/Camera/ColorGrade/Style/Quality/Negatives);negatives 是质量控制关键;camera 字段提供电影级构图控制
|
||||
- Created: 1 concept (Nano Banana), 1 entity (Google)
|
||||
|
||||
## [2026-04-15] ingest | Claude Code 调用方法总结
|
||||
- [Claude-Code调用方法总结](sources/Claude-Code调用方法总结.md)
|
||||
- Key claims: Print Mode = stdin 管道非交互模式(推荐);bypassPermissions 跳过所有确认;--add-dir 自动激活 SKILL;delegate_task 无法建立外部 CLI 通道
|
||||
- Created: 2 concepts (Print Mode, Skill加载), 2 entities (Claude Code, Hermes)
|
||||
|
||||
## [2026-04-15] ingest | 万字讲透OpenClaw Workspace
|
||||
- [万字讲透OpenClaw-Workspace深度解析](sources/万字讲透OpenClaw-Workspace深度解析.md)
|
||||
- Key claims: AGENTS.md 300-500字最佳;SOUL vs IDENTITY 分工明确;TOOLS.md 核心是"什么时候不用";memory/ 是真正长期记忆;7个核心文件配合实现可预期性
|
||||
- Created: 7 concepts (Workspace, AGENTS.md, SOUL.md, USER.md, TOOLS.md, IDENTITY.md, BOOTSTRAP.md), 2 entities (OpenClaw, DracoVibeCoding)
|
||||
|
||||
## [2026-04-15] ingest | 使用Claude自动生成N8N工作流
|
||||
- [使用Claude自动生成N8N工作流的实操教程](sources/使用Claude自动生成N8N工作流的实操教程.md)
|
||||
- Key claims: n8n-mcp 提供 543 节点结构化访问;Claude 自动生成工作流完成度 80%-90%;Opensea 模型 + extended thinking 提升质量
|
||||
- Created: 4 entities (n8n-mcp, czlonkowski, Claude, n8n), 1 concept (AI工作流自动生成)
|
||||
|
||||
## [2026-04-15] ingest | MCP在Cursor中的集成与应用
|
||||
- [MCP在Cursor中的集成与应用详解](sources/MCP在Cursor中的集成与应用详解.md)
|
||||
- Key claims: MCP 3种核心接口(GET/POST/Promise);Cursor 支持 SSE 和 Command 两种接入方式;Agent 模式自动执行工具链;Sequential Thinking 分步推理
|
||||
- Created: 4 entities (Composer, Sequential Thinking, Cursor, MCP), 3 concepts (Agent模式, MCP工具链, MCP)
|
||||
|
||||
## [2026-04-15] ingest | Google 5个Agent Skill设计模式
|
||||
- [Google-5个Agent-Skill设计模式-2026-03-19](sources/Google-5个Agent-Skill设计模式-2026-03-19.md)
|
||||
- Key claims: 5种设计模式(Tool Wrapper/Generator/Reviewer/Inversion/Pipeline);可组合使用;Anthropic:最好的 Skill 是工具箱而非提示词
|
||||
- Created: 1 entity (Google), 10 concepts (Agent Skill设计模式, Tool Wrapper, Generator, Reviewer, Inversion, Pipeline, 渐进式披露, etc.)
|
||||
|
||||
Reference in New Issue
Block a user