Auto-sync: 2026-04-17 19:17
This commit is contained in:
45
wiki/sources/autonomous-project-management.md
Normal file
45
wiki/sources/autonomous-project-management.md
Normal file
@@ -0,0 +1,45 @@
|
||||
---
|
||||
title: "Autonomous Project Management with Subagents"
|
||||
type: source
|
||||
tags: []
|
||||
date: 2026-04-17
|
||||
source_file: raw/Agent/usecases/autonomous-project-management.md
|
||||
---
|
||||
|
||||
## Source File
|
||||
- [[raw/Agent/usecases/autonomous-project-management.md]]
|
||||
|
||||
## Summary
|
||||
- 核心主题:去中心化的多 Subagent 项目管理模式,通过共享 STATE.yaml 文件协调任务,避免中央 orchestrator 瓶颈
|
||||
- 问题域:复杂项目的多任务并行管理,传统 orchestrator 模式导致主 Agent 成为流量瓶颈
|
||||
- 方法/机制:基于共享状态文件的去中心化协调,主 Agent 采用"CEO 模式"仅做策略决策
|
||||
- 结论/价值:多个 Subagent 并行工作,通过状态文件自驱协调,主会话保持精简
|
||||
|
||||
## Key Claims
|
||||
- STATE.yaml 作为单一真相源,替代中央 orchestrator 实现去中心化协调
|
||||
- 主会话采用 CEO 模式,仅负责任务分配和状态检查,不参与具体执行
|
||||
- Subagent 通过读写共享状态文件实现自主协调,无需中央调度
|
||||
- Git 提交 STATE.yaml 变更实现完整审计追踪
|
||||
|
||||
## Key Quotes
|
||||
> "Decentralized coordination: Agents read/write to a shared STATE.yaml file" — 核心设计原则
|
||||
> "Main session stays thin (CEO pattern—strategy only)" — 主会话保持精简
|
||||
> "File-based coordination scales better than message-passing" — STATE.yaml 优于消息传递
|
||||
|
||||
## Key Concepts
|
||||
- [[去中心化协调]]:通过共享状态文件实现多 Agent 自主协调的模式
|
||||
- [[Subagent 管理]]:使用 sessions_spawn/sessions_send 管理子代理的技术
|
||||
- [[项目状态管理]]:基于 STATE.yaml 的任务追踪和协调机制
|
||||
- [[CEO 模式]]:主 Agent 仅做策略决策,不执行具体任务的架构模式
|
||||
|
||||
## Key Entities
|
||||
- [[Nicholas Carlini]]:自主编码 Agent 方案的提出者,启发了去中心化协调模式
|
||||
- [[OpenClaw]]:支持 Subagent 管理的 AI Agent 工具
|
||||
|
||||
## Connections
|
||||
- [[Project State Management System]] ← extends ← [[去中心化协调]]
|
||||
- [[Multi-Agent Team]] ← uses ← [[Subagent 管理]]
|
||||
- [[Shared Memory]] ← relates_to ← [[项目状态管理]]
|
||||
|
||||
## Contradictions
|
||||
- 与传统 Orchestrator 模式冲突:传统模式依赖中央调度,本模式强调去中心化自驱
|
||||
42
wiki/sources/dynamic-dashboard-with-sub-agent-spawning.md
Normal file
42
wiki/sources/dynamic-dashboard-with-sub-agent-spawning.md
Normal file
@@ -0,0 +1,42 @@
|
||||
---
|
||||
title: "Dynamic Dashboard with Sub-agent Spawning"
|
||||
type: source
|
||||
tags: [dashboard, sub-agent, automation]
|
||||
date: 2026-04-17
|
||||
---
|
||||
|
||||
## Source File
|
||||
- [[raw/Agent/usecases/dynamic-dashboard.md]]
|
||||
|
||||
## Summary
|
||||
- 核心主题:实时动态仪表盘系统,通过 Sub-agent 并行获取多数据源数据
|
||||
- 问题域:传统静态仪表盘数据过时、构建成本高、API 速率限制问题
|
||||
- 方法/机制:OpenClaw spawn sub-agents 并行获取多数据源(GitHub、Twitter、Polymarket、系统健康),聚合结果后推送至 Discord 或生成 HTML 仪表盘,定时 Cron Jobs 自动更新
|
||||
- 结论/价值:快速实现实时多源数据监控,支持告警阈值和历史趋势分析
|
||||
|
||||
## Key Claims
|
||||
- Sub-agent 并行执行可避免阻塞并分散 API 负载
|
||||
- 定时任务实现自动化数据更新,无需人工干预
|
||||
- 历史数据存储支持趋势分析和可视化
|
||||
|
||||
## Key Quotes
|
||||
> "Building a custom dashboard takes weeks. By the time it's done, requirements have changed." — 痛点描述
|
||||
|
||||
## Key Concepts
|
||||
- [[Cron Jobs]]:定时任务调度机制,每 15 分钟触发数据更新
|
||||
- [[Sub-agent 管理]]:OpenClaw 的 sessions_spawn 会话管理,用于并行执行多数据源获取任务
|
||||
- [[Multi-Agent Team]]:多 Agent 协作架构,每个 Agent 有独立角色和任务
|
||||
|
||||
## Key Entities
|
||||
- [[OpenClaw]]:AI Agent 管理工具,负责 sub-agent 调度和任务协调
|
||||
- [[GitHub]]:代码托管平台,提供 stars、forks、issues 等指标
|
||||
- [[Discord]]:即时通讯平台,用于接收仪表盘更新推送
|
||||
|
||||
## Connections
|
||||
- [[Cron Jobs]] ← drives ← [[动态仪表盘]]
|
||||
- [[Sub-agent 管理]] ← enables ← [[动态仪表盘]]
|
||||
- [[动态仪表盘]] ← monitors ← [[GitHub]]
|
||||
- [[动态仪表盘]] ← monitors ← [[Discord]]
|
||||
|
||||
## Contradictions
|
||||
- (暂无)
|
||||
53
wiki/sources/habit-tracker-accountability-coach.md
Normal file
53
wiki/sources/habit-tracker-accountability-coach.md
Normal file
@@ -0,0 +1,53 @@
|
||||
---
|
||||
id: habit-tracker-accountability-coach
|
||||
title: Habit Tracker & Accountability Coach
|
||||
type: source
|
||||
tags: [agent-use-case, habit-tracking, accountability, telegram, automation]
|
||||
date: 2026-04-17
|
||||
---
|
||||
|
||||
## Source File
|
||||
- [[raw/Agent/usecases/habit-tracker-accountability-coach.md]]
|
||||
|
||||
## Summary
|
||||
- 核心主题:AI Agent 作为主动式习惯追踪与问责伙伴
|
||||
- 问题域:传统习惯追踪应用被动式提醒,用户容易忽略通知,行为改变需要主动问责
|
||||
- 方法/机制:通过 Telegram/SMS 定时主动检查、连续打卡追踪、自适应提醒语气、周报分析
|
||||
- 结论/价值:主动问责比被动提醒更有效,AI Agent 可实现零社交压力的个人问责系统
|
||||
|
||||
## Key Claims
|
||||
- 习惯追踪应用失败的根本原因是被动式交互,用户主动打开率低
|
||||
- 主动式问责(AI 直接询问完成情况)比推送通知更有效
|
||||
- 自适应语气(连续成功时鼓励,失败时温和提醒)提升长期坚持率
|
||||
- 习惯数量控制在 3-5 个可避免检查疲劳
|
||||
|
||||
## Key Quotes
|
||||
> "What actually works for behavior change is active accountability — someone (or something) that asks you directly, celebrates your wins, and nudges you when you slip."
|
||||
|
||||
> "A message that says 'Day 15, don't break it now' actually motivates."
|
||||
|
||||
## Key Concepts
|
||||
- [[Cron-Jobs]]:定时任务调度,AI Agent 通过 cron 实现每日定时检查
|
||||
- [[Preference-Learning]]:AI 通过交互学习用户偏好,持续优化提醒策略
|
||||
- [[工作流自动化]]:预定义自动化流程,AI Agent 定时执行检查并记录结果
|
||||
- [[上下文记忆]]:AI Agent 保留对话历史,追踪连续打卡天数和用户响应模式
|
||||
|
||||
## Key Entities
|
||||
- [[Telegram]]:消息推送渠道,通过 Telegram Bot API 实现每日检查
|
||||
- [[Twilio]]:SMS 替代渠道,提供短信通知能力
|
||||
- [[Google Sheets]]:可选可视化仪表盘,数据导出和可视化展示
|
||||
- [[OpenClaw]]:AI Agent 运行环境,支持定时任务和文件存储
|
||||
|
||||
## Connections
|
||||
- [[Health-Symptom-Tracker]] ← combines_with ← [[Habit-Tracker-Accountability-Coach]]
|
||||
- [[Custom-Morning-Brief]] ← similar_approach ← [[Habit-Tracker-Accountability-Coach]]
|
||||
- [[Cron-Jobs]] ← enables ← [[Habit-Tracker-Accountability-Coach]]
|
||||
|
||||
## Contradictions
|
||||
- 与传统习惯追踪应用(如 Habitica、Streaks)的核心区别在于主动 vs 被动交互模式
|
||||
|
||||
## Implementation Notes
|
||||
1. Telegram Bot API 配置相对简单,无需手机号验证
|
||||
2. Twilio 需要美国号码用于 SMS 发送,成本高于 Telegram
|
||||
3. 数据存储推荐使用本地 JSON 文件(~/habits/log.json),便于历史查询
|
||||
4. Google Sheets 集成可选,适合需要可视化数据的用户
|
||||
41
wiki/sources/latex-paper-writing.md
Normal file
41
wiki/sources/latex-paper-writing.md
Normal file
@@ -0,0 +1,41 @@
|
||||
---
|
||||
title: "LaTeX Paper Writing"
|
||||
type: source
|
||||
tags: []
|
||||
date: 2026-04-17
|
||||
---
|
||||
|
||||
## Source File
|
||||
- [[raw/Agent/usecases/latex-paper-writing.md]]
|
||||
|
||||
## Summary
|
||||
- 核心主题:使用 AI Agent 作为 LaTeX 写作助手,实现无本地安装的即时编译环境
|
||||
- 问题域:TeX Live 安装占用大量磁盘空间、编译错误调试繁琐、编辑器与 PDF 查看器切换破坏工作流
|
||||
- 方法/机制:通过 Prismer Docker 容器提供完整 TeX Live 环境,latex-compiler skill 提供 4 个工具(latex_compile、latex_preview、latex_templates、latex_get_template)
|
||||
- 结论/价值:可通过自然语言描述需求生成 LaTeX 源码并即时编译预览,支持多种模板和文献格式
|
||||
|
||||
## Key Claims
|
||||
- AI Agent 可作为 LaTeX 写作助手,根据描述生成对应的 LaTeX 源码
|
||||
- 无需本地安装 TeX Live,通过 Docker 容器即可获得完整 LaTeX 编译环境
|
||||
- 支持 pdflatex、xelatex、lualatex 三种编译引擎,xelatex 用于 CJK/中文支持
|
||||
|
||||
## Key Quotes
|
||||
> "Write LaTeX collaboratively with the agent — describe what you want and it generates the source" — 核心工作方式
|
||||
|
||||
> "Compile to PDF instantly with pdflatex, xelatex, or lualatex (no local TeX installation needed)" — 零本地安装承诺
|
||||
|
||||
## Key Concepts
|
||||
- [[LaTeX编译]]:通过 pdflatex/xelatex/lualatex 将 LaTeX 源码编译为 PDF
|
||||
- [[BibTeX/BibLaTeX]]:文献管理格式,可在 LaTeX 文档中引用 .bib 文件
|
||||
- [[LaTeX模板]]:article、IEEE、beamer、中文article 等预置模板
|
||||
|
||||
## Key Entities
|
||||
- [[Prismer]]:提供 Docker 容器化 LaTeX 编译环境的服务端项目
|
||||
- [[latex-compiler]]:Claude Code Skill,提供 4 个 LaTeX 相关工具
|
||||
|
||||
## Connections
|
||||
- [[latex-compiler]] ← provided_by ← [[Prismer]]
|
||||
- [[LaTeX模板]] ← used_in ← [[LaTeX编译]]
|
||||
|
||||
## Contradiction
|
||||
- (无)
|
||||
48
wiki/sources/local-crm-framework-with-denchclaw.md
Normal file
48
wiki/sources/local-crm-framework-with-denchclaw.md
Normal file
@@ -0,0 +1,48 @@
|
||||
---
|
||||
id: local-crm-framework-with-denchclaw
|
||||
title: Local CRM Framework with DenchClaw
|
||||
type: source
|
||||
tags: [CRM, OpenClaw, DenchClaw, Agent, 自动化]
|
||||
sources:
|
||||
- raw/Agent/usecases/local-crm-framework.md
|
||||
last_updated: 2026-04-17
|
||||
---
|
||||
|
||||
## Source File
|
||||
- [[raw/Agent/usecases/local-crm-framework.md]]
|
||||
|
||||
## Summary
|
||||
- 核心主题:使用 DenchClaw 框架将 OpenClaw 转变为本地 CRM 系统
|
||||
- 问题域:CRM 系统搭建繁琐、需要整合多种工具(数据库、UI、浏览器自动化、消息平台)
|
||||
- 方法/机制:单命令安装、DuckDB 本地数据库、Chrome 配置克隆、自然语言交互、文件驱动 UI
|
||||
- 结论/价值:提供 Cursor 级别的用户体验,一站式解决 CRM 搭建问题
|
||||
|
||||
## Key Claims
|
||||
- 单命令安装(`npx denchclaw`)即可完成数据库、Web UI、OpenClaw 配置、浏览器自动化和技能的完整安装
|
||||
- 自然语言 CRM 功能允许通过自然语言查询和操作数据,无需手动配置过滤器
|
||||
- Chrome 配置克隆使 Agent 继承用户认证状态,可直接操作需要登录的网页
|
||||
- 文件系统优先的设计(所有设置、过滤器、视图存储为 YAML/Markdown 文件)使 Agent 能像修改代码一样直接修改 UI
|
||||
|
||||
## Key Quotes
|
||||
> "You want Cursor-level UX for your business operations, not a pile of shell scripts"
|
||||
|
||||
> "One `npx` command beats a weekend of setup" — 整个技术栈(数据库、Web UI、OpenClaw profile、gateway、browser、skills)通过单条命令自动安装和配置
|
||||
|
||||
## Key Concepts
|
||||
- [[DuckDB]]:嵌入式分析数据库,最小且性能最高的全功能 SQL 数据库,无服务器进程、无凭证、无网络依赖
|
||||
- [[Chrome-Profile-Cloning]]:克隆用户 Chrome 配置使 Agent 继承认证状态的技术
|
||||
- [[文件驱动 UI]]:将 UI 配置存储为文件,使 AI Agent 能直接读取和修改界面设置
|
||||
|
||||
## Key Entities
|
||||
- [[DenchClaw]]:开源框架,将 OpenClaw 转变为本地 CRM、销售自动化和生产力平台
|
||||
- [[OpenClaw]]:AI Agent 管理工具,DenchClaw 的底层技术
|
||||
- [[DuckDB]]:嵌入式分析数据库
|
||||
- [[Chrome]]:Google 浏览器,用于配置克隆
|
||||
|
||||
## Connections
|
||||
- [[DenchClaw]] ← builds_on ← [[OpenClaw]]
|
||||
- [[DenchClaw]] ← uses ← [[DuckDB]]
|
||||
- [[DenchClaw]] ← clones ← [[Chrome]]
|
||||
|
||||
## Contradictions
|
||||
- (暂无发现)
|
||||
50
wiki/sources/multi-channel-ai-customer-service-platform.md
Normal file
50
wiki/sources/multi-channel-ai-customer-service-platform.md
Normal file
@@ -0,0 +1,50 @@
|
||||
---
|
||||
title: "Multi-Channel AI Customer Service Platform"
|
||||
type: source
|
||||
tags: [AI, Customer Service, Automation, Multi-Channel]
|
||||
date: 2026-04-17
|
||||
---
|
||||
|
||||
## Source File
|
||||
- [[raw/Agent/usecases/multi-channel-customer-service.md]]
|
||||
|
||||
## Summary
|
||||
- 核心主题:多渠道AI客服平台,整合WhatsApp、Instagram、邮件和Google Reviews到统一收件箱
|
||||
- 问题域:小型企业多平台客户消息管理,人工客服成本高、响应慢
|
||||
- 方法/机制:AI自动回复 + 人工接管机制 + 测试模式 + 业务知识库训练
|
||||
- 结论/价值:将客服响应时间从4+小时缩短至2分钟内,80%咨询自动化处理
|
||||
|
||||
## Key Claims
|
||||
- AI驱动的统一收件箱可处理WhatsApp Business、Instagram DMs、Gmail和Google Reviews
|
||||
- 业务知识库训练使AI能够基于企业服务、价格和政策进行智能回复
|
||||
- 测试模式允许向客户演示系统而不影响真实客户
|
||||
- 清晰的人机交接规则定义升级触发条件,避免AI越权
|
||||
|
||||
## Key Quotes
|
||||
> "One restaurant reduced response time from 4+ hours to under 2 minutes, handling 80% of inquiries automatically." — Futurist Systems实际部署案例
|
||||
|
||||
## Key Concepts
|
||||
- [[统一收件箱]]:将多个渠道的客户消息汇聚到单一AI驱动界面的架构
|
||||
- [[AI自动回复]]:基于知识库和意图分类的智能响应机制
|
||||
- [[人工接管]]:复杂问题或特定触发条件下转由人工处理的机制
|
||||
- [[测试模式]]:演示模式,不发送真实消息仅记录日志
|
||||
- [[业务知识库]]:企业服务、价格、政策和FAQ的结构化数据存储
|
||||
- [[消息路由]]:根据渠道、意图分类和客户类型的智能分发逻辑
|
||||
|
||||
## Key Entities
|
||||
- [[WhatsApp Business API]]:WhatsApp官方商业API(通过360dialog或官方渠道)
|
||||
- [[Instagram Graph API]]:Meta Business Suite的Instagram消息API
|
||||
- [[Gmail]]:Google邮件服务(通过gog CLI集成)
|
||||
- [[Google Business Profile API]]:Google商家资料API,用于处理评论
|
||||
- [[Futurist Systems]]:部署该解决方案的技术咨询公司
|
||||
|
||||
## Connections
|
||||
- [[AI自动回复]] ← depends_on ← [[业务知识库]]
|
||||
- [[人工接管]] ← depends_on ← [[意图分类]]
|
||||
- [[Multi-Channel Assistant]] ← extends ← [[统一收件箱]]
|
||||
|
||||
## Contradictions
|
||||
- 与 [[Inbox De-clutter]] 冲突:
|
||||
- 冲突点:两者都处理消息,但Inbox De-clutter专注于Newsletter整理,本方案专注于多渠道客服
|
||||
- 当前观点:统一收件箱适合服务型企业
|
||||
- 对方观点:Newsletter整理适合内容创作者和营销人员
|
||||
45
wiki/sources/n8n-workflow-orchestration.md
Normal file
45
wiki/sources/n8n-workflow-orchestration.md
Normal file
@@ -0,0 +1,45 @@
|
||||
---
|
||||
title: "OpenClaw + n8n Workflow Orchestration"
|
||||
type: source
|
||||
tags: [ai-agent, n8n, workflow-automation, security]
|
||||
date: 2026-04-17
|
||||
---
|
||||
|
||||
## Source File
|
||||
- [[raw/Agent/usecases/n8n-workflow-orchestration.md]]
|
||||
|
||||
## Summary
|
||||
- 核心主题:AI Agent 与 n8n 工作流的编排模式,通过 webhook 代理实现凭证隔离和安全交互
|
||||
- 问题域:AI Agent 直接管理外部 API 凭证的安全风险
|
||||
- 方法/机制:代理模式、凭证隔离、可视化调试、锁定工作流
|
||||
- 结论/价值:三位一体——可观测性、安全性、性能优化
|
||||
|
||||
## Key Claims
|
||||
- OpenClaw 通过 n8n webhook 调用外部 API,永不接触 API 密钥
|
||||
- n8n 凭证存储在 n8n 内部,Agent 仅知道 webhook URL
|
||||
- 工作流可锁定,防止 Agent 悄悄修改 API 交互方式
|
||||
- 确定性子任务通过工作流执行,不消耗 LLM 推理 token
|
||||
|
||||
## Key Quotes
|
||||
> "三个复合问题:无可见性、凭证蔓延、Token 浪费" — 痛点描述
|
||||
|
||||
> "构建 → 测试 → 锁定周期是关键" — 关键工作流
|
||||
|
||||
## Key Concepts
|
||||
- [[Webhook]]: n8n 接收外部 HTTP POST 请求的触发器
|
||||
- [[Credential-Isolation]]: 凭证隔离,API 密钥存储在 n8n 而非 Agent 环境
|
||||
- [[Lockable-Workflow]]: 可锁定工作流,锁定后 Agent 无法修改 API 交互方式
|
||||
|
||||
## Key Entities
|
||||
- [[OpenClaw]]: AI Agent 管理工具
|
||||
- [[n8n]]: 开源工作流自动化工具
|
||||
|
||||
## Connections
|
||||
- [[OpenClaw]] ← delegates_to ← [[n8n]]
|
||||
- [[n8n]] ← connects_to ← External-Services
|
||||
|
||||
## Contradictions
|
||||
- 与直接 API 调用方式冲突:
|
||||
- 冲突点:凭证存储位置
|
||||
- 当前观点:通过 n8n 代理,凭证与 Agent 隔离
|
||||
- 对方观点:Agent 直接持有 API 密钥
|
||||
43
wiki/sources/overnight-mini-app-builder.md
Normal file
43
wiki/sources/overnight-mini-app-builder.md
Normal file
@@ -0,0 +1,43 @@
|
||||
---
|
||||
title: "Goal-Driven Autonomous Tasks"
|
||||
type: source
|
||||
tags: []
|
||||
date: 2026-04-17
|
||||
---
|
||||
|
||||
## Source File
|
||||
- [[raw/Agent/usecases/overnight-mini-app-builder.md]]
|
||||
|
||||
## Summary
|
||||
- 核心主题:AI Agent 自主目标驱动任务生成与执行
|
||||
- 问题域:个人目标分解与日常任务执行
|
||||
- 方法/机制:用户一次性输入目标 → Agent 每日自动生成任务 → 自主执行 → 追踪进度
|
||||
- 结论/价值:将 AI Agent 转变为自驱型员工,用户定义目标,Agent 负责分解和执行
|
||||
|
||||
## Key Claims
|
||||
- 用户只需 brain dump 目标,Agent 每天自动生成 4-5 个可完成的任务
|
||||
- Agent 可以每晚构建一个 MVP 级别的 mini-app 作为惊喜
|
||||
- Sub-agent 并发编辑同一文件会导致竞态条件,解决方法是分离任务文件(只读主文件 + 追加日志)
|
||||
|
||||
## Key Quotes
|
||||
> "Your AI agent is powerful but reactive — it only works when you tell it what to do. What if it knew your goals and proactively came up with tasks to move you closer to them every single day?" — 核心痛点阐述
|
||||
> "The brain dump is everything. The more context you give about your goals, the better the agent's daily tasks will be." — 关键洞察
|
||||
|
||||
## Key Concepts
|
||||
- [[Brain Dump]]:一次性输入所有目标、使命和任务的上下文设定方式
|
||||
- [[每日任务生成]]:Agent 每天自动生成 4-5 个可执行任务的工作流
|
||||
- [[Mini-App Builder]]:Agent 每晚构建 MVP 级别应用的工作流
|
||||
- [[竞态条件]]:Sub-agent 并发编辑共享文件导致的静默失败问题
|
||||
- [[任务文件分离模式]]:分离只读主文件和追加日志文件来解决竞态条件
|
||||
|
||||
## Key Entities
|
||||
- [[OpenClaw]]:AI Agent 管理工具,本工作流的核心执行引擎
|
||||
- [[Alex Finn]]:YouTube 创作者,灵感来源
|
||||
|
||||
## Connections
|
||||
- [[Multi-Agent Specialized Team (Solo Founder Setup)]] ← similar_concept ← [[Goal-Driven Autonomous Tasks]]
|
||||
- [[Dynamic Dashboard with Sub-agent Spawning]] ← uses ← [[Subagent]]
|
||||
- [[Habit Tracker & Accountability Coach]] ← extends ← [[Goal-Driven Autonomous Tasks]]
|
||||
|
||||
## Contradictions
|
||||
- (暂无)
|
||||
41
wiki/sources/polymarket-autopilot.md
Normal file
41
wiki/sources/polymarket-autopilot.md
Normal file
@@ -0,0 +1,41 @@
|
||||
---
|
||||
title: "Polymarket Autopilot: Automated Paper Trading"
|
||||
type: source
|
||||
tags: [polymarket, autopilot, paper-trading, prediction-market]
|
||||
date: 2026-04-17
|
||||
---
|
||||
|
||||
## Source File
|
||||
- [[raw/Agent/usecases/polymarket-autopilot.md]]
|
||||
|
||||
## Summary
|
||||
- 核心主题:AI Agent 自动化模拟交易预测市场
|
||||
- 问题域:预测市场机会监控、交易策略执行、组合绩效追踪
|
||||
- 方法/机制:通过 Cron Jobs 定时执行,调用 Polymarket API 获取市场数据,使用 TAIL(趋势跟随)、BONDING(逆势)、SPREAD(价差套利)三种策略执行模拟交易,记录每日绩效并推送 Discord 报告
|
||||
- 结论/价值:在无风险环境下测试和优化交易策略,学习市场模式并自适应调整参数
|
||||
|
||||
## Key Claims
|
||||
- AI Agent 可通过 API 持续监控预测市场并自动执行模拟交易策略
|
||||
- TAIL 策略在强趋势(概率 >60% + 成交量飙升)时有效
|
||||
- BONDING 策略在市场对新闻过度反应时(突然下跌 >10%)有效
|
||||
- SPREAD 策略在 YES+NO 概率之和 >1.05 时存在套利机会
|
||||
|
||||
## Key Quotes
|
||||
> "Manually monitoring prediction markets for arbitrage opportunities and executing trades is time-consuming and requires constant attention." — 手动监控预测市场的痛点
|
||||
|
||||
## Key Concepts
|
||||
- [[Paper Trading]]:模拟交易,使用虚拟资金测试策略无需承担真实风险
|
||||
- [[Cron Jobs]]:定时任务调度,AI Agent 每 15 分钟执行一次市场扫描
|
||||
- [[Discord Integration]]:通过 Discord Webhook 实现每日交易报告推送
|
||||
- [[Sub-agent Spawning]]:在高峰期并行分析多个市场
|
||||
|
||||
## Key Entities
|
||||
- [[Polymarket]]:预测市场平台,提供 API 获取市场价格、成交量、价差数据
|
||||
|
||||
## Connections
|
||||
- [[Cron Jobs]] ← 驱动 ← [[Polymarket Autopilot]]
|
||||
- [[Discord]] ← 通知 ← [[Polymarket Autopilot]]
|
||||
- [[Subagent 管理]] ← 实现 ← [[Polymarket Autopilot]](并行市场分析)
|
||||
|
||||
## Contradictions
|
||||
- (暂无)
|
||||
44
wiki/sources/pre-build-idea-validator.md
Normal file
44
wiki/sources/pre-build-idea-validator.md
Normal file
@@ -0,0 +1,44 @@
|
||||
---
|
||||
id: pre-build-idea-validator
|
||||
title: "Pre-Build Idea Validator"
|
||||
type: source
|
||||
tags: [agent, mcp, openclaw, idea-validation]
|
||||
date: 2026-04-17
|
||||
---
|
||||
|
||||
## Source File
|
||||
- [[raw/Agent/usecases/pre-build-idea-validator.md]]
|
||||
|
||||
## Summary
|
||||
- 核心主题:AI Agent 项目启动前的创意验证机制
|
||||
- 问题域:独立开发者经常在饱和领域构建重复项目,浪费大量时间
|
||||
- 方法/机制:通过 MCP server 扫描 5 个真实数据源(GitHub、Hacker News、npm、PyPI、Product Hunt),计算竞争度评分
|
||||
- 结论/价值:基于真实数据而非 LLM 猜测,帮助开发者做出"是否继续"的数据驱动决策
|
||||
|
||||
## Key Claims
|
||||
- 在构建前验证创意可避免最昂贵的错误:解决已被解决的问题
|
||||
- reality_signal > 70 应停止并讨论,30-70 显示结果并建议细分角度,< 30 可继续构建
|
||||
- 高评分意味着"差异化或不要做",低评分意味着存在真正的空白地带
|
||||
|
||||
## Key Quotes
|
||||
> "You tell your agent 'build me an AI code review tool' and it happily spends 6 hours coding. Meanwhile, 143,000+ repos already exist on GitHub — the top one has 53,000 stars." — 痛点描述
|
||||
|
||||
## Key Concepts
|
||||
- [[Pre-Build Idea Validation]]:在编写代码前验证项目创意是否已存在
|
||||
- [[reality_signal]]:竞争度评分(0-100),基于真实数据计算
|
||||
|
||||
## Key Entities
|
||||
- [[OpenClaw]]:AI Agent 管理工具,支持 MCP 集成
|
||||
- [[idea-reality-mcp]]:扫描真实数据源的 MCP server
|
||||
- [[GitHub]]:代码托管平台
|
||||
- [[Hacker News]]:科技新闻聚合平台
|
||||
- [[Product Hunt]]:产品发布平台
|
||||
|
||||
## Connections
|
||||
- [[OpenClaw]] ← uses ← [[idea-reality-mcp]]
|
||||
- [[idea-reality-mcp]] ← scans ← [[GitHub]]
|
||||
- [[idea-reality-mcp]] ← scans ← [[Hacker News]]
|
||||
- [[idea-reality-mcp]] ← scans ← [[Product Hunt]]
|
||||
|
||||
## Contradictions
|
||||
- (暂无发现冲突)
|
||||
48
wiki/sources/second-brain.md
Normal file
48
wiki/sources/second-brain.md
Normal file
@@ -0,0 +1,48 @@
|
||||
---
|
||||
title: "Second Brain"
|
||||
type: source
|
||||
tags: []
|
||||
date: 2026-04-17
|
||||
---
|
||||
|
||||
## Source File
|
||||
- [[raw/Agent/usecases/second-brain.md]]
|
||||
|
||||
## Summary
|
||||
- 核心主题:AI Agent 作为个人记忆捕获系统
|
||||
- 问题域:笔记工具的"信息黑洞"问题——收集但不使用
|
||||
- 方法/机制:通过即时通讯(Telegram/Discord/iMessage)零摩擦捕获,OpenClaw 永久存储,Next.js 搜索界面
|
||||
- 结论/价值:捕获如发短信般简单,检索如搜索般便捷
|
||||
|
||||
## Key Claims
|
||||
- 零摩擦捕获是第二大脑的核心——不需要打开应用、选择文件夹或添加标签,直接发短信即可
|
||||
- OpenClaw 的记忆系统是累积的——告诉它的每一条信息都会被永久记住
|
||||
- 可以用手机发短信,Agent 在电脑上构建——对话就是界面
|
||||
|
||||
## Key Quotes
|
||||
> "Capture should be as easy as texting, and retrieval should be as easy as searching" — 核心洞察
|
||||
|
||||
## Key Concepts
|
||||
- [[Second Brain]]:通过 AI Agent 实现的个人知识管理系统
|
||||
- [[零摩擦捕获]]:无需组织结构,直接文本发送即可存储的工作流
|
||||
- [[Memory System]]:OpenClaw 的永久记忆机制,按日期滚动存储跨会话信息
|
||||
|
||||
## Key Entities
|
||||
- [[OpenClaw]]:AI Agent 管理工具,提供记忆系统功能
|
||||
- [[Telegram]]:即时通讯平台,用于消息捕获(已有 entity)
|
||||
- [[Discord]]:即时通讯平台,用于消息捕获
|
||||
- [[Next.js]]:React 框架,用于构建搜索界面
|
||||
- [[Alex Finn]](已有):YouTube 创作者,Second Brain 用例视频的作者
|
||||
- [[Building a Second Brain]]:Tiago Forte 创建的知识管理方法论
|
||||
|
||||
## Connections
|
||||
- [[Second Brain]] ← enables ← [[零摩擦捕获]]
|
||||
- [[OpenClaw]] ← provides ← [[Memory System]]
|
||||
- [[Alex Finn]] ← inspired ← [[Second Brain]]
|
||||
- [[Telegram]] ← feeds ← [[OpenClaw]]
|
||||
|
||||
## Contradictions
|
||||
- 与 [[印象笔记]] 冲突:
|
||||
- 冲突点:组织方式
|
||||
- 当前观点:无需文件夹、无需标签,文本+搜索
|
||||
- 对方观点:依赖文件夹和标签的层级结构,容易导致"存而不读"
|
||||
40
wiki/sources/todoist-task-manager-agent-task-visibility.md
Normal file
40
wiki/sources/todoist-task-manager-agent-task-visibility.md
Normal file
@@ -0,0 +1,40 @@
|
||||
---
|
||||
title: "Todoist Task Manager: Agent Task Visibility"
|
||||
type: source
|
||||
tags: []
|
||||
date: 2026-04-17
|
||||
---
|
||||
|
||||
## Source File
|
||||
- [[raw/Agent/usecases/todoist-task-manager.md]]
|
||||
|
||||
## Summary
|
||||
- 核心主题:AI Agent 任务可视化与进度追踪
|
||||
- 问题域:长时间运行的 Agent 工作流缺乏透明度,用户无法追踪任务进度
|
||||
- 方法/机制:通过 Todoist API 将 Agent 内部推理和进度日志同步到任务管理工具
|
||||
- 结论/价值:最大化复杂 Agent 工作流的透明度,实时展示任务状态、子步骤完成情况和阻塞点
|
||||
|
||||
## Key Claims
|
||||
- Agent 工作流复杂度与用户信息需求呈正相关,多步骤任务需要外部化进度追踪
|
||||
- 通过 Todoist Section 区分任务状态(In Progress/Waiting/Done)实现任务阶段可视化
|
||||
- 任务描述中记录 Agent 内部 Plan 实现推理外部化
|
||||
- 子步骤完成通过任务评论实时追加,实现进度流式更新
|
||||
|
||||
## Key Quotes
|
||||
> "When agents run complex, multi-step tasks (like building a full-stack app or performing deep research), the user often loses track of what the agent is currently doing, what steps have been completed, and where the agent might be stuck." — 痛点描述
|
||||
|
||||
## Key Concepts
|
||||
- [[Agent-Task-Visibility]]:AI Agent 任务对用户的透明化展示机制
|
||||
- [[Task-Automation]]:将手动任务创建过程转化为系统自动执行的机制
|
||||
|
||||
## Key Entities
|
||||
- [[Todoist]]:流行任务管理应用,提供 REST API 用于任务和评论操作
|
||||
- [[OpenClaw]]:AI Agent 管理工具,可执行 shell 命令和文件系统操作
|
||||
|
||||
## Connections
|
||||
- [[OpenClaw]] ← uses → [[Todoist]]
|
||||
- [[Agent-Task-Visibility]] ← implements → [[Task-Automation]]
|
||||
- [[Todoist]] ← provides → [[Agent-Task-Visibility]]
|
||||
|
||||
## Contradictions
|
||||
- (暂无)
|
||||
Reference in New Issue
Block a user