Update nexus: fix conflicts and sync local changes

This commit is contained in:
Shen Wei
2026-04-26 12:06:50 +08:00
parent 191797c01b
commit f09834b5a5
2443 changed files with 254323 additions and 255154 deletions

View File

@@ -1,114 +1,114 @@
# Hermes 自定义技能说明
> 创建日期2026-04-20
> 更新日期2026-04-20
---
## 技能总览
| 技能名 | 用途 | 调用方式 |
| --------------------------------- | --------------------------------------------- | ------------- |
| `blogwatcher-daily` | RSS 订阅监控 + 每日笔记生成 | cronjob |
| `claude-code-executor` | 用 TMUX 启动 Claude Code 并委托任务 | delegate_task |
| `claude-code-infographic-prompts` | 调用 Claude Code 的 baoyu-infographic 生成信息图提示词 | 手动调用 |
| `marker-pdf-to-markdown` | PDF 转 Markdownmarker_single 单文件 / marker 批量) | terminal |
| `whisper-audio-to-text` | 音频/视频转文字Whisper | terminal |
---
## 1. claude-code-executor
**路径**`~/.hermes/skills/custom/claude-code-executor/SKILL.md`
**用途**:通过 TMUX 启动 Claude CodebypassPermissions 模式),发送任务指令,监控完成状态。
**触发关键词**:用户说"请用 Claude Code 做 xxx"、"用 Claude Code 执行 xxx"
**核心流程**
1. 启动 TMUX session `claude-task`
2. 等 8 秒后发送 EnterbypassPermissions 确认)
3. 发送完整任务指令
4. 监控 `done:` 输出
5. 清理 tmux session
**SSH 到 ubuntu2 命令**`ssh ubuntu2`
---
## 2. claude-code-infographic-build
**路径**`~/.hermes/skills/custom/claude-code-infographic-build/SKILL.md`
**用途**:调用 Claude Code 的 `baoyu-infographic` 技能,为笔记内容生成信息图提示词,并进一步生成图片添加到原始笔记。
**完整三步流程**
1. **步骤①**baoyu-infographic 生成三部分提示词 → 保存到 Hermes/xingzhi/
2. **步骤②**:在同一 Claude Code session 中baoyu-imagine 生成图片 → 保存到 Hermes/xingzhi/
3. **步骤③**obsidian skill 将图片引用添加到原始笔记
**三部分提示词格式**
1. **系统提示词** — Image Specifications + Core Principles
2. **风格锁定提示词** — Layout Guidelines + Style Guidelines
3. **内容结构提示词** — Text Requirements + 具体内容
**布局参考**`~/.claude/skills/baoyu-infographic/references/layouts/`
- `hub-spoke` — 适合 mind-map标注了 "Best For: Mind maps"
- `circular-flow` — 适合循环流程
- `venn` — 适合交集关系
**风格参考**`~/.claude/skills/baoyu-infographic/references/styles/`
- `corporate-memphis` — 商务孟菲斯风格
- `chalkboard` — 粉笔黑板风格
- `cyberpunk-neon` — 赛博霓虹风格
- `warm` / `notion` / `minimal` / `blueprint` / `watercolor` / `elegant`
**文件命名**`[主题]-infographic-prompts-YYYY-MM-DD.md`
---
## 3. blogwatcher-daily
**路径**`~/.hermes/skills/custom/blogwatcher-daily/SKILL.md`
**用途**RSS 订阅监控 + 每日笔记生成。使用 RSSHub + feedparser 抓取 31 个订阅,自动去重并存入 SQLite新文章追加写入 Markdown 笔记。
---
## 4. marker-pdf-to-markdown
**路径**`~/.hermes/skills/custom/marker-pdf-to-markdown/SKILL.md`
**用途**PDF 转 Markdown/HTML/JSON 高精度工具,支持 OCR、表格、公式识别。
**安装位置**ubuntu2 服务器(`ssh ubuntu2`
| 命令 | 用途 | 输入 |
|------|------|------|
| `marker_single` | 单文件转换,默认输出到 PDF 同目录 | 单个 PDF 文件 |
| `marker` | 批量转换,需文件夹输入 | 文件夹路径 |
常用选项:`--output_dir``--output_format json/markdown/html``--no-images``--use_llm``--page_range`
---
## 5. whisper-audio-to-text
**路径**`~/.hermes/skills/custom/whisper-audio-to-text/SKILL.md`
**用途**:使用 OpenAI Whisper 将音频/视频转换为文字(转录或翻译)。
**安装位置**Mac miniGPU 加速、ubuntu1、ubuntu2CPU 模式),均直接可用。
---
## 相关 Claude Code 技能(`~/.claude/skills/`
| 技能名 | 用途 |
|--------|------|
| `baoyu-article-illustrator` | 文章配图(分析文章结构 + 生成配图) |
| `baoyu-infographic` | 信息图生成 |
| `baoyu-cover-image` | 封面图生成 |
| `fireworks-tech-graph` | 技术图生成 |
这些 Claude Code 技能通过 `claude-code-executor` 启动 Claude Code CLI 后调用。
# Hermes 自定义技能说明
> 创建日期2026-04-20
> 更新日期2026-04-20
---
## 技能总览
| 技能名 | 用途 | 调用方式 |
| --------------------------------- | --------------------------------------------- | ------------- |
| `blogwatcher-daily` | RSS 订阅监控 + 每日笔记生成 | cronjob |
| `claude-code-executor` | 用 TMUX 启动 Claude Code 并委托任务 | delegate_task |
| `claude-code-infographic-prompts` | 调用 Claude Code 的 baoyu-infographic 生成信息图提示词 | 手动调用 |
| `marker-pdf-to-markdown` | PDF 转 Markdownmarker_single 单文件 / marker 批量) | terminal |
| `whisper-audio-to-text` | 音频/视频转文字Whisper | terminal |
---
## 1. claude-code-executor
**路径**`~/.hermes/skills/custom/claude-code-executor/SKILL.md`
**用途**:通过 TMUX 启动 Claude CodebypassPermissions 模式),发送任务指令,监控完成状态。
**触发关键词**:用户说"请用 Claude Code 做 xxx"、"用 Claude Code 执行 xxx"
**核心流程**
1. 启动 TMUX session `claude-task`
2. 等 8 秒后发送 EnterbypassPermissions 确认)
3. 发送完整任务指令
4. 监控 `done:` 输出
5. 清理 tmux session
**SSH 到 ubuntu2 命令**`ssh ubuntu2`
---
## 2. claude-code-infographic-build
**路径**`~/.hermes/skills/custom/claude-code-infographic-build/SKILL.md`
**用途**:调用 Claude Code 的 `baoyu-infographic` 技能,为笔记内容生成信息图提示词,并进一步生成图片添加到原始笔记。
**完整三步流程**
1. **步骤①**baoyu-infographic 生成三部分提示词 → 保存到 Hermes/xingzhi/
2. **步骤②**:在同一 Claude Code session 中baoyu-imagine 生成图片 → 保存到 Hermes/xingzhi/
3. **步骤③**obsidian skill 将图片引用添加到原始笔记
**三部分提示词格式**
1. **系统提示词** — Image Specifications + Core Principles
2. **风格锁定提示词** — Layout Guidelines + Style Guidelines
3. **内容结构提示词** — Text Requirements + 具体内容
**布局参考**`~/.claude/skills/baoyu-infographic/references/layouts/`
- `hub-spoke` — 适合 mind-map标注了 "Best For: Mind maps"
- `circular-flow` — 适合循环流程
- `venn` — 适合交集关系
**风格参考**`~/.claude/skills/baoyu-infographic/references/styles/`
- `corporate-memphis` — 商务孟菲斯风格
- `chalkboard` — 粉笔黑板风格
- `cyberpunk-neon` — 赛博霓虹风格
- `warm` / `notion` / `minimal` / `blueprint` / `watercolor` / `elegant`
**文件命名**`[主题]-infographic-prompts-YYYY-MM-DD.md`
---
## 3. blogwatcher-daily
**路径**`~/.hermes/skills/custom/blogwatcher-daily/SKILL.md`
**用途**RSS 订阅监控 + 每日笔记生成。使用 RSSHub + feedparser 抓取 31 个订阅,自动去重并存入 SQLite新文章追加写入 Markdown 笔记。
---
## 4. marker-pdf-to-markdown
**路径**`~/.hermes/skills/custom/marker-pdf-to-markdown/SKILL.md`
**用途**PDF 转 Markdown/HTML/JSON 高精度工具,支持 OCR、表格、公式识别。
**安装位置**ubuntu2 服务器(`ssh ubuntu2`
| 命令 | 用途 | 输入 |
|------|------|------|
| `marker_single` | 单文件转换,默认输出到 PDF 同目录 | 单个 PDF 文件 |
| `marker` | 批量转换,需文件夹输入 | 文件夹路径 |
常用选项:`--output_dir``--output_format json/markdown/html``--no-images``--use_llm``--page_range`
---
## 5. whisper-audio-to-text
**路径**`~/.hermes/skills/custom/whisper-audio-to-text/SKILL.md`
**用途**:使用 OpenAI Whisper 将音频/视频转换为文字(转录或翻译)。
**安装位置**Mac miniGPU 加速、ubuntu1、ubuntu2CPU 模式),均直接可用。
---
## 相关 Claude Code 技能(`~/.claude/skills/`
| 技能名 | 用途 |
|--------|------|
| `baoyu-article-illustrator` | 文章配图(分析文章结构 + 生成配图) |
| `baoyu-infographic` | 信息图生成 |
| `baoyu-cover-image` | 封面图生成 |
| `fireworks-tech-graph` | 技术图生成 |
这些 Claude Code 技能通过 `claude-code-executor` 启动 Claude Code CLI 后调用。

View File

@@ -1,169 +1,169 @@
# Chrome DevTools MCP - Infographic Prompts
Generated: 2026-04-20
Layout: circular-flow
Style: chalkboard
Aspect: 16:9
---
## Part 1: System Prompt
```
Create a professional infographic with the following specifications:
## Image Specifications
- Type: Infographic
- Layout: circular-flow (cyclic process showing continuous steps arranged in a circle)
- Style: chalkboard (chalk on black board aesthetic)
- Aspect Ratio: 16:9
- Language: English
## Core Principles
- Follow the circular-flow layout structure: arrange information nodes in a circle with directional arrows
- Each section represents a tool category as a node on the circle
- Use chalk-drawn style: imperfect hand-drawn lines, chalk dust effects, white/yellow/pink/blue chalk colors
- Black chalkboard background (#1A1A1A)
- Center can hold the main concept "Chrome DevTools MCP"
- Maintain authentic chalk texture on all elements
- Use circular arrangement to show the tool workflow cycle
- Clear visual hierarchy with color variety
## Text Requirements
- Main titles prominent and readable in chalk white
- Key concepts emphasized with chalk yellow/pink
- Labels clear and appropriately sized
```
---
## Part 2: Style Locking Prompt
```
## Style Guidelines - chalkboard
### Background
- Color: Chalkboard Black (#1A1A1A) or Dark Green-Black (#1C2B1C)
- Texture: Realistic chalkboard texture with subtle scratches, dust particles
### Typography
- Hand-drawn chalk lettering with visible chalk texture
- White or bright colored chalk for emphasis
### Color Palette
- Background: #1A1A1A (Chalkboard Black)
- Primary Text: #F5F5F5 (Chalk White)
- Accent 1: #FFE566 (Chalk Yellow) - highlights, emphasis
- Accent 2: #FF9999 (Chalk Pink) - secondary highlights
- Accent 3: #66B3FF (Chalk Blue) - links, connections
- Accent 4: #90EE90 (Chalk Green) - success indicators
### Visual Elements
- Hand-drawn chalk illustrations with sketchy, imperfect lines
- Chalk dust effects around text and key elements
- Doodles: stars, arrows, underlines, circles
- Connection lines with hand-drawn feel
- Directional arrows showing cycle flow
### Do
- Maintain authentic chalk texture on all elements
- Use imperfect, hand-drawn quality
- Add subtle chalk dust and smudge effects
- Create visual hierarchy with color variety
- Include playful doodles and annotations
### Don't
- Use perfect geometric shapes
- Create clean digital-looking lines
- Add photorealistic elements
- Use gradients or glossy effects
```
---
## Part 3: Content Structure Prompt
```
## Content Structure - circular-flow
### Center Element
- "Chrome DevTools MCP" - main concept in center of circle
### Circle Nodes (6 categories, clockwise flow):
1. INPUT AUTOMATION (9 tools)
- click, drag, fill, fill_form, handle_dialog, hover, press_key, type_text, upload_file
- Chalk Pink icon/node
2. NAVIGATION AUTOMATION (6 tools)
- close_page, list_pages, navigate_page, new_page, select_page, wait_for
- Chalk Blue icon/node
3. EMULATION (2 tools)
- emulate, resize_page
- Chalk Yellow icon/node
4. PERFORMANCE (4 tools)
- performance_analyze_insight, performance_start_trace, performance_stop_trace, take_memory_snapshot
- Chalk Green icon/node
5. NETWORK (2 tools)
- get_network_request, list_network_requests
- Chalk Pink icon/node
6. DEBUGGING (6 tools)
- evaluate_script, get_console_message, lighthouse_audit, list_console_messages, take_screenshot, take_snapshot
- Chalk Blue icon/node
### Supported Clients (around the outer edge):
- Claude Code, Cline, Cursor, VS Code, Copilot, Codex, Gemini, JetBrains, Kiro, Windsurf, Amp, Antigravity, Command Code, Factory, Mistral, OpenCode, Qoder, Warp
### Configuration Options (bottom section):
- --headless, --slim, --autoConnect, --browser-url, --channel, --viewport, --isolated, --user-data-dir
### Arrows
- Curved directional arrows connecting each node in clockwise direction
- Showing continuous workflow cycle
### Labels in English
- All text labels in English
- Use chalk white for main text, chalk yellow for emphasis
```
---
## Visual Layout Diagram
```
[INPUT]
↙ ↘
[DEBUGGING] ← CENTER → [NAVIGATION]
↗ ↣
↙ ↘
[NETWORK] ←─────────────────────→ [EMULATION]
↘ ↙
↗ ↣
[PERFORMANCE] →
[???]
Actually circular flow (clockwise):
┌─────────────────────────────────────────┐
│ │
│ [EMULATION] → [PERFORMANCE] │
│ ↗ ↗ │
│ │ │ │
│ [DEBUGGING] [NAVIGATION] │
│ │ │ │
│ ↖ ↘ │
│ [NETWORK] ← [INPUT] │
│ │
│ Center: "Chrome DevTools MCP" │
└─────────────────────────────────────────┘
```
---
## Final Prompt Summary
# Chrome DevTools MCP - Infographic Prompts
Generated: 2026-04-20
Layout: circular-flow
Style: chalkboard
Aspect: 16:9
---
## Part 1: System Prompt
```
Create a professional infographic with the following specifications:
## Image Specifications
- Type: Infographic
- Layout: circular-flow (cyclic process showing continuous steps arranged in a circle)
- Style: chalkboard (chalk on black board aesthetic)
- Aspect Ratio: 16:9
- Language: English
## Core Principles
- Follow the circular-flow layout structure: arrange information nodes in a circle with directional arrows
- Each section represents a tool category as a node on the circle
- Use chalk-drawn style: imperfect hand-drawn lines, chalk dust effects, white/yellow/pink/blue chalk colors
- Black chalkboard background (#1A1A1A)
- Center can hold the main concept "Chrome DevTools MCP"
- Maintain authentic chalk texture on all elements
- Use circular arrangement to show the tool workflow cycle
- Clear visual hierarchy with color variety
## Text Requirements
- Main titles prominent and readable in chalk white
- Key concepts emphasized with chalk yellow/pink
- Labels clear and appropriately sized
```
---
## Part 2: Style Locking Prompt
```
## Style Guidelines - chalkboard
### Background
- Color: Chalkboard Black (#1A1A1A) or Dark Green-Black (#1C2B1C)
- Texture: Realistic chalkboard texture with subtle scratches, dust particles
### Typography
- Hand-drawn chalk lettering with visible chalk texture
- White or bright colored chalk for emphasis
### Color Palette
- Background: #1A1A1A (Chalkboard Black)
- Primary Text: #F5F5F5 (Chalk White)
- Accent 1: #FFE566 (Chalk Yellow) - highlights, emphasis
- Accent 2: #FF9999 (Chalk Pink) - secondary highlights
- Accent 3: #66B3FF (Chalk Blue) - links, connections
- Accent 4: #90EE90 (Chalk Green) - success indicators
### Visual Elements
- Hand-drawn chalk illustrations with sketchy, imperfect lines
- Chalk dust effects around text and key elements
- Doodles: stars, arrows, underlines, circles
- Connection lines with hand-drawn feel
- Directional arrows showing cycle flow
### Do
- Maintain authentic chalk texture on all elements
- Use imperfect, hand-drawn quality
- Add subtle chalk dust and smudge effects
- Create visual hierarchy with color variety
- Include playful doodles and annotations
### Don't
- Use perfect geometric shapes
- Create clean digital-looking lines
- Add photorealistic elements
- Use gradients or glossy effects
```
---
## Part 3: Content Structure Prompt
```
## Content Structure - circular-flow
### Center Element
- "Chrome DevTools MCP" - main concept in center of circle
### Circle Nodes (6 categories, clockwise flow):
1. INPUT AUTOMATION (9 tools)
- click, drag, fill, fill_form, handle_dialog, hover, press_key, type_text, upload_file
- Chalk Pink icon/node
2. NAVIGATION AUTOMATION (6 tools)
- close_page, list_pages, navigate_page, new_page, select_page, wait_for
- Chalk Blue icon/node
3. EMULATION (2 tools)
- emulate, resize_page
- Chalk Yellow icon/node
4. PERFORMANCE (4 tools)
- performance_analyze_insight, performance_start_trace, performance_stop_trace, take_memory_snapshot
- Chalk Green icon/node
5. NETWORK (2 tools)
- get_network_request, list_network_requests
- Chalk Pink icon/node
6. DEBUGGING (6 tools)
- evaluate_script, get_console_message, lighthouse_audit, list_console_messages, take_screenshot, take_snapshot
- Chalk Blue icon/node
### Supported Clients (around the outer edge):
- Claude Code, Cline, Cursor, VS Code, Copilot, Codex, Gemini, JetBrains, Kiro, Windsurf, Amp, Antigravity, Command Code, Factory, Mistral, OpenCode, Qoder, Warp
### Configuration Options (bottom section):
- --headless, --slim, --autoConnect, --browser-url, --channel, --viewport, --isolated, --user-data-dir
### Arrows
- Curved directional arrows connecting each node in clockwise direction
- Showing continuous workflow cycle
### Labels in English
- All text labels in English
- Use chalk white for main text, chalk yellow for emphasis
```
---
## Visual Layout Diagram
```
[INPUT]
↙ ↘
[DEBUGGING] ← CENTER → [NAVIGATION]
↗ ↣
↙ ↘
[NETWORK] ←─────────────────────→ [EMULATION]
↘ ↙
↗ ↣
[PERFORMANCE] →
[???]
Actually circular flow (clockwise):
┌─────────────────────────────────────────┐
│ │
│ [EMULATION] → [PERFORMANCE] │
│ ↗ ↗ │
│ │ │ │
│ [DEBUGGING] [NAVIGATION] │
│ │ │ │
│ ↖ ↘ │
│ [NETWORK] ← [INPUT] │
│ │
│ Center: "Chrome DevTools MCP" │
└─────────────────────────────────────────┘
```
---
## Final Prompt Summary
Generate a chalkboard-style infographic showing Chrome DevTools MCP tool categories arranged in a circular flow pattern. The circle has 6 nodes representing the tool categories (Input Automation, Navigation, Emulation, Performance, Network, Debugging). Each node displays the tool count and key tool names in chalk-style lettering. The center shows "Chrome DevTools MCP" as the main concept. Arrows show clockwise flow indicating the continuous nature of browser automation workflow. Use authentic chalkboard aesthetic with black background (#1A1A1A), chalk white text, and colorful chalk accents (yellow, pink, blue, green) for visual hierarchy. Include chalk dust effects and hand-drawn imperfect lines throughout.

View File

@@ -1,63 +1,63 @@
# Hermes Custom Skills - Infographic Image Prompt
## Image Specifications
- **Type**: Infographic
- **Layout**: Circular Flow (cyclic process showing continuous recurring steps)
- **Style**: Chalkboard (black chalkboard background with colorful chalk drawing style)
- **Aspect Ratio**: 16:9
- **Language**: English
## Core Principles
- Follow the circular-flow layout structure precisely for information architecture
- Apply chalkboard style aesthetics consistently throughout
- Create stylistically similar alternatives for any specific figures
- Keep information concise, highlight keywords and core concepts
- Use ample whitespace for visual clarity
- Maintain clear visual hierarchy
## Layout Guidelines
- Circular arrangement with 5 skill nodes evenly spaced around the circle
- Arrows showing clockwise direction flow
- No clear start/end (continuous cycle)
- Center holds main concept: "Hermes Automation Ecosystem"
- Title at top: "Hermes Custom Skills - Automation Workflows"
- Step labels at each node with skill names and taglines
- Brief descriptions near nodes
## Style Guidelines
- **Background**: Chalkboard Black (#1A1A1A) with realistic chalkboard texture, subtle scratches, dust particles, faint eraser marks
- **Typography**: Hand-drawn chalk lettering style with visible chalk texture, imperfect baseline, white or bright colored chalk for emphasis
- **Color Palette**:
- Background: #1A1A1A (Chalkboard Black)
- Primary Text: #F5F5F5 (Chalk White)
- Accent 1: #FFE566 (Chalk Yellow) - for skill names
- Accent 2: #FF9999 (Chalk Pink) - for highlights
- Accent 3: #66B3FF (Chalk Blue) - for connections
- Accent 4: #90EE90 (Chalk Green) - for success indicators
- **Visual Elements**: Hand-drawn chalk illustrations with sketchy imperfect lines, chalk dust effects, doodles (stars, arrows, circles), stick figures, connection lines with hand-drawn feel
## Content
**Title**: Hermes Custom Skills - Automation Workflows
**Center Concept**: Hermes Automation Ecosystem
**5 Skill Nodes (clockwise from top)**:
1. **blogwatcher-daily** - RSS Subscription Monitor - RSS monitoring + daily note generation - cronjob
2. **claude-code-executor** - Claude Code Task Delegation - Launch Claude Code via TMUX, send tasks, monitor completion - delegate_task
3. **claude-code-infographic-prompts** - Infographic Prompt Generator - Generate 3-part prompts for baoyu-infographic skill - manual
4. **marker-pdf-to-markdown** - PDF to Markdown Converter - High-precision PDF conversion with OCR, table, formula recognition - terminal
5. **whisper-audio-to-text** - Audio/Video Transcription - OpenAI Whisper for speech-to-text - terminal
**Bottom Note**: "5 Custom Skills | Powering Hermes Automation"
**Visual Elements**:
- Chalk dust particles around text
- Hand-drawn arrows connecting skills in clockwise cycle
- Small doodle icons (stars, checkmarks) near each skill
# Hermes Custom Skills - Infographic Image Prompt
## Image Specifications
- **Type**: Infographic
- **Layout**: Circular Flow (cyclic process showing continuous recurring steps)
- **Style**: Chalkboard (black chalkboard background with colorful chalk drawing style)
- **Aspect Ratio**: 16:9
- **Language**: English
## Core Principles
- Follow the circular-flow layout structure precisely for information architecture
- Apply chalkboard style aesthetics consistently throughout
- Create stylistically similar alternatives for any specific figures
- Keep information concise, highlight keywords and core concepts
- Use ample whitespace for visual clarity
- Maintain clear visual hierarchy
## Layout Guidelines
- Circular arrangement with 5 skill nodes evenly spaced around the circle
- Arrows showing clockwise direction flow
- No clear start/end (continuous cycle)
- Center holds main concept: "Hermes Automation Ecosystem"
- Title at top: "Hermes Custom Skills - Automation Workflows"
- Step labels at each node with skill names and taglines
- Brief descriptions near nodes
## Style Guidelines
- **Background**: Chalkboard Black (#1A1A1A) with realistic chalkboard texture, subtle scratches, dust particles, faint eraser marks
- **Typography**: Hand-drawn chalk lettering style with visible chalk texture, imperfect baseline, white or bright colored chalk for emphasis
- **Color Palette**:
- Background: #1A1A1A (Chalkboard Black)
- Primary Text: #F5F5F5 (Chalk White)
- Accent 1: #FFE566 (Chalk Yellow) - for skill names
- Accent 2: #FF9999 (Chalk Pink) - for highlights
- Accent 3: #66B3FF (Chalk Blue) - for connections
- Accent 4: #90EE90 (Chalk Green) - for success indicators
- **Visual Elements**: Hand-drawn chalk illustrations with sketchy imperfect lines, chalk dust effects, doodles (stars, arrows, circles), stick figures, connection lines with hand-drawn feel
## Content
**Title**: Hermes Custom Skills - Automation Workflows
**Center Concept**: Hermes Automation Ecosystem
**5 Skill Nodes (clockwise from top)**:
1. **blogwatcher-daily** - RSS Subscription Monitor - RSS monitoring + daily note generation - cronjob
2. **claude-code-executor** - Claude Code Task Delegation - Launch Claude Code via TMUX, send tasks, monitor completion - delegate_task
3. **claude-code-infographic-prompts** - Infographic Prompt Generator - Generate 3-part prompts for baoyu-infographic skill - manual
4. **marker-pdf-to-markdown** - PDF to Markdown Converter - High-precision PDF conversion with OCR, table, formula recognition - terminal
5. **whisper-audio-to-text** - Audio/Video Transcription - OpenAI Whisper for speech-to-text - terminal
**Bottom Note**: "5 Custom Skills | Powering Hermes Automation"
**Visual Elements**:
- Chalk dust particles around text
- Hand-drawn arrows connecting skills in clockwise cycle
- Small doodle icons (stars, checkmarks) near each skill
- Eraser smudge textures as subtle background variation

View File

@@ -1,119 +1,119 @@
# Hermes Custom Skills Infographic Prompts
**Date**: 2026-04-20
**Layout**: circular-flow
**Style**: chalkboard
**Aspect**: 16:9
---
## Part 1: System Prompt (Image Specifications + Core Principles)
**Image Specifications**
- **Type**: Infographic
- **Layout**: Circular Flow (cyclic process showing continuous recurring steps)
- **Style**: Chalkboard (black chalkboard background with colorful chalk drawing style)
- **Aspect Ratio**: 16:9
- **Language**: English
**Core Principles**
- Follow the circular-flow layout structure precisely for information architecture
- Apply chalkboard style aesthetics consistently throughout
- Create stylistically similar alternatives for any specific figures
- Keep information concise, highlight keywords and core concepts
- Use ample whitespace for visual clarity
- Maintain clear visual hierarchy
---
## Part 2: Style Locking Prompts (Layout Guidelines + Style Guidelines)
**Layout Guidelines (Circular Flow)**
- Circular arrangement with steps around the circle
- Arrows showing direction (clockwise flow)
- No clear start/end (continuous cycle)
- Center can hold main concept
- Steps around the circle: 5 skill nodes evenly spaced
- Icons per step representing each skill
- Title at top
- Step labels at each node
- Brief descriptions near nodes
- Center concept: "Hermes Automation Ecosystem"
**Style Guidelines (Chalkboard)**
- **Background**: Chalkboard Black (#1A1A1A) with realistic chalkboard texture, subtle scratches, dust particles, faint eraser marks
- **Typography**: Hand-drawn chalk lettering style with visible chalk texture, imperfect baseline, white or bright colored chalk for emphasis
- **Color Palette**:
- Background: #1A1A1A (Chalkboard Black)
- Primary Text: #F5F5F5 (Chalk White)
- Accent 1: #FFE566 (Chalk Yellow) - for skill names
- Accent 2: #FF9999 (Chalk Pink) - for highlights
- Accent 3: #66B3FF (Chalk Blue) - for connections
- Accent 4: #90EE90 (Chalk Green) - for success indicators
- **Visual Elements**: Hand-drawn chalk illustrations with sketchy imperfect lines, chalk dust effects, doodles (stars, arrows, circles), stick figures, connection lines with hand-drawn feel
- **Do**: Maintain authentic chalk texture, use imperfect hand-drawn quality, add chalk dust effects, create visual hierarchy with color variety
- **Don't**: Use perfect geometric shapes, create clean digital-looking lines, add photorealistic elements, use gradients
---
## Part 3: Content Structure Prompts (Text Requirements + Structured Content)
**Text Requirements**
- All text must match chalkboard style treatment
- Main titles should be prominent and readable
- Key concepts should be visually emphasized
- Labels should be clear and appropriately sized
- Use English for all text content
**Structured Content**
**Title**: Hermes Custom Skills - Automation Workflows
**Center Concept**: Hermes Automation Ecosystem (in the middle of the circle)
**5 Skill Nodes (clockwise from top)**:
1. **blogwatcher-daily**
- Tagline: RSS Subscription Monitor
- Function: RSS monitoring + daily note generation
- Call method: cronjob
- Icon: RSS feed / newspaper symbol
2. **claude-code-executor**
- Tagline: Claude Code Task Delegation
- Function: Launch Claude Code via TMUX, send tasks, monitor completion
- Call method: delegate_task
- Icon: robot / terminal symbol
3. **claude-code-infographic-prompts**
- Tagline: Infographic Prompt Generator
- Function: Generate 3-part prompts for baoyu-infographic skill
- Call method: manual
- Icon: palette / image symbol
4. **marker-pdf-to-markdown**
- Tagline: PDF to Markdown Converter
- Function: High-precision PDF conversion with OCR, table, formula recognition
- Call method: terminal
- Icon: document conversion symbol
5. **whisper-audio-to-text**
- Tagline: Audio/Video Transcription
- Function: OpenAI Whisper for speech-to-text
- Call method: terminal
- Icon: microphone / audio wave symbol
**Connection Arrows**: Clockwise arrows showing continuous workflow between skills
**Bottom Note**: "5 Custom Skills | Powering Hermes Automation"
---
## Visual Elements to Include
- Wooden chalkboard frame border (optional)
- Chalk dust particles around text
- Hand-drawn arrows connecting skills in cycle
- Small doodle icons (stars, checkmarks) near each skill
- Mathematical/formula-style decorations (fitting the chalkboard theme)
# Hermes Custom Skills Infographic Prompts
**Date**: 2026-04-20
**Layout**: circular-flow
**Style**: chalkboard
**Aspect**: 16:9
---
## Part 1: System Prompt (Image Specifications + Core Principles)
**Image Specifications**
- **Type**: Infographic
- **Layout**: Circular Flow (cyclic process showing continuous recurring steps)
- **Style**: Chalkboard (black chalkboard background with colorful chalk drawing style)
- **Aspect Ratio**: 16:9
- **Language**: English
**Core Principles**
- Follow the circular-flow layout structure precisely for information architecture
- Apply chalkboard style aesthetics consistently throughout
- Create stylistically similar alternatives for any specific figures
- Keep information concise, highlight keywords and core concepts
- Use ample whitespace for visual clarity
- Maintain clear visual hierarchy
---
## Part 2: Style Locking Prompts (Layout Guidelines + Style Guidelines)
**Layout Guidelines (Circular Flow)**
- Circular arrangement with steps around the circle
- Arrows showing direction (clockwise flow)
- No clear start/end (continuous cycle)
- Center can hold main concept
- Steps around the circle: 5 skill nodes evenly spaced
- Icons per step representing each skill
- Title at top
- Step labels at each node
- Brief descriptions near nodes
- Center concept: "Hermes Automation Ecosystem"
**Style Guidelines (Chalkboard)**
- **Background**: Chalkboard Black (#1A1A1A) with realistic chalkboard texture, subtle scratches, dust particles, faint eraser marks
- **Typography**: Hand-drawn chalk lettering style with visible chalk texture, imperfect baseline, white or bright colored chalk for emphasis
- **Color Palette**:
- Background: #1A1A1A (Chalkboard Black)
- Primary Text: #F5F5F5 (Chalk White)
- Accent 1: #FFE566 (Chalk Yellow) - for skill names
- Accent 2: #FF9999 (Chalk Pink) - for highlights
- Accent 3: #66B3FF (Chalk Blue) - for connections
- Accent 4: #90EE90 (Chalk Green) - for success indicators
- **Visual Elements**: Hand-drawn chalk illustrations with sketchy imperfect lines, chalk dust effects, doodles (stars, arrows, circles), stick figures, connection lines with hand-drawn feel
- **Do**: Maintain authentic chalk texture, use imperfect hand-drawn quality, add chalk dust effects, create visual hierarchy with color variety
- **Don't**: Use perfect geometric shapes, create clean digital-looking lines, add photorealistic elements, use gradients
---
## Part 3: Content Structure Prompts (Text Requirements + Structured Content)
**Text Requirements**
- All text must match chalkboard style treatment
- Main titles should be prominent and readable
- Key concepts should be visually emphasized
- Labels should be clear and appropriately sized
- Use English for all text content
**Structured Content**
**Title**: Hermes Custom Skills - Automation Workflows
**Center Concept**: Hermes Automation Ecosystem (in the middle of the circle)
**5 Skill Nodes (clockwise from top)**:
1. **blogwatcher-daily**
- Tagline: RSS Subscription Monitor
- Function: RSS monitoring + daily note generation
- Call method: cronjob
- Icon: RSS feed / newspaper symbol
2. **claude-code-executor**
- Tagline: Claude Code Task Delegation
- Function: Launch Claude Code via TMUX, send tasks, monitor completion
- Call method: delegate_task
- Icon: robot / terminal symbol
3. **claude-code-infographic-prompts**
- Tagline: Infographic Prompt Generator
- Function: Generate 3-part prompts for baoyu-infographic skill
- Call method: manual
- Icon: palette / image symbol
4. **marker-pdf-to-markdown**
- Tagline: PDF to Markdown Converter
- Function: High-precision PDF conversion with OCR, table, formula recognition
- Call method: terminal
- Icon: document conversion symbol
5. **whisper-audio-to-text**
- Tagline: Audio/Video Transcription
- Function: OpenAI Whisper for speech-to-text
- Call method: terminal
- Icon: microphone / audio wave symbol
**Connection Arrows**: Clockwise arrows showing continuous workflow between skills
**Bottom Note**: "5 Custom Skills | Powering Hermes Automation"
---
## Visual Elements to Include
- Wooden chalkboard frame border (optional)
- Chalk dust particles around text
- Hand-drawn arrows connecting skills in cycle
- Small doodle icons (stars, checkmarks) near each skill
- Mathematical/formula-style decorations (fitting the chalkboard theme)
- Eraser smudge textures as subtle background variation

View File

@@ -1,72 +1,72 @@
---
title: "llm-wiki-sync Circular Flow"
topic: technical
data_type: cycle
complexity: moderate
point_count: 7
source_language: zh
user_language: en
---
## Main Topic
A cyclical knowledge management pipeline that transforms raw notes into structured wiki pages through continuous LLM-powered ingestion, extraction, and reuse.
## Learning Objectives
After viewing this infographic, the viewer will understand:
1. The continuous circular flow of llm-wiki-sync from raw notes to reusable knowledge
2. The key extraction outputs: Summary, Claims, Entities, Concepts, Connections
3. How feedback and reuse complete the cycle back to new raw material
## Target Audience
- **Knowledge Level**: Intermediate technical audience
- **Context**: Developers and knowledge workers interested in AI-powered knowledge management
- **Expectations**: Clear understanding of the llm-wiki-sync pipeline and its cyclical nature
## Content Type Analysis
- **Data Structure**: Cyclic process with recurring steps
- **Key Relationships**: Raw → Ingest → Extract → Source Page → Graph/Site → Reuse → Raw (feedback loop)
- **Visual Opportunities**: Circular flow with nodes for each stage, arrows showing direction, center concept
## Key Data Points (Verbatim)
### Core Pipeline Steps
1. **Raw Note** - Original document in raw/ folder
2. **Ingest** - LLM analyzes and extracts structured information
3. **Extract** - Summary, Claims, Quotes, Entities, Concepts, Connections
4. **Source Page** - Structured wiki/sources/ page with frontmatter
5. **Graph & Site** - graph.json and Quartz static site generation
6. **Reuse** - Synthesize, query, create new content from structured knowledge
7. **Feedback Loop** - New raw notes created from reused knowledge
### Extraction Outputs
- **Summary**: 核心主题, 问题域, 方法/机制, 结论/价值
- **Key Claims**: Verifiable assertions extracted from text
- **Key Entities**: LaunchDarkly, HP, Christian Dior, etc.
- **Key Concepts**: RTO, RPO, Feature Flag, Kill Switch, Gradual Rollout
- **Connections**: depends_on, enables, provides relationships
### Key Quotes
- "RTO is about speed: how fast you get back online. RPO is about data: how much you can afford to lose."
- "Deploy != Release. Feature flags change this. You can deploy code to production without releasing it to users."
## Layout × Style Signals
- Content type: cycle → circular-flow
- Tone: technical educational → chalkboard
- Audience: developers → clear, legible, professional
- Complexity: moderate → balanced density with clear visual hierarchy
## Design Instructions (from user input)
- **Layout**: circular-flow (NOT linear - must emphasize recurring cycle)
- **Style**: chalkboard (dark background, hand-drawn chalk accents)
- **Aspect**: 16:9 landscape
- **Language**: English
- Circular flow showing: raw note -> ingest -> extract -> source page -> graph/static site -> reuse/feedback -> knowledge base
- Include core extraction outputs as recurring nodes/callouts
- Keep text concise and legible
- Dark chalkboard background with hand-drawn chalk accents
- Avoid clutter, make cycle visually clear and publication-ready
## Recommended Combinations
1. **circular-flow + chalkboard** (Recommended): Perfect match for cycle/process content with chalkboard aesthetic
2. **hub-spoke + technical-schematic**: For emphasizing central concepts with technical precision
---
title: "llm-wiki-sync Circular Flow"
topic: technical
data_type: cycle
complexity: moderate
point_count: 7
source_language: zh
user_language: en
---
## Main Topic
A cyclical knowledge management pipeline that transforms raw notes into structured wiki pages through continuous LLM-powered ingestion, extraction, and reuse.
## Learning Objectives
After viewing this infographic, the viewer will understand:
1. The continuous circular flow of llm-wiki-sync from raw notes to reusable knowledge
2. The key extraction outputs: Summary, Claims, Entities, Concepts, Connections
3. How feedback and reuse complete the cycle back to new raw material
## Target Audience
- **Knowledge Level**: Intermediate technical audience
- **Context**: Developers and knowledge workers interested in AI-powered knowledge management
- **Expectations**: Clear understanding of the llm-wiki-sync pipeline and its cyclical nature
## Content Type Analysis
- **Data Structure**: Cyclic process with recurring steps
- **Key Relationships**: Raw → Ingest → Extract → Source Page → Graph/Site → Reuse → Raw (feedback loop)
- **Visual Opportunities**: Circular flow with nodes for each stage, arrows showing direction, center concept
## Key Data Points (Verbatim)
### Core Pipeline Steps
1. **Raw Note** - Original document in raw/ folder
2. **Ingest** - LLM analyzes and extracts structured information
3. **Extract** - Summary, Claims, Quotes, Entities, Concepts, Connections
4. **Source Page** - Structured wiki/sources/ page with frontmatter
5. **Graph & Site** - graph.json and Quartz static site generation
6. **Reuse** - Synthesize, query, create new content from structured knowledge
7. **Feedback Loop** - New raw notes created from reused knowledge
### Extraction Outputs
- **Summary**: 核心主题, 问题域, 方法/机制, 结论/价值
- **Key Claims**: Verifiable assertions extracted from text
- **Key Entities**: LaunchDarkly, HP, Christian Dior, etc.
- **Key Concepts**: RTO, RPO, Feature Flag, Kill Switch, Gradual Rollout
- **Connections**: depends_on, enables, provides relationships
### Key Quotes
- "RTO is about speed: how fast you get back online. RPO is about data: how much you can afford to lose."
- "Deploy != Release. Feature flags change this. You can deploy code to production without releasing it to users."
## Layout × Style Signals
- Content type: cycle → circular-flow
- Tone: technical educational → chalkboard
- Audience: developers → clear, legible, professional
- Complexity: moderate → balanced density with clear visual hierarchy
## Design Instructions (from user input)
- **Layout**: circular-flow (NOT linear - must emphasize recurring cycle)
- **Style**: chalkboard (dark background, hand-drawn chalk accents)
- **Aspect**: 16:9 landscape
- **Language**: English
- Circular flow showing: raw note -> ingest -> extract -> source page -> graph/static site -> reuse/feedback -> knowledge base
- Include core extraction outputs as recurring nodes/callouts
- Keep text concise and legible
- Dark chalkboard background with hand-drawn chalk accents
- Avoid clutter, make cycle visually clear and publication-ready
## Recommended Combinations
1. **circular-flow + chalkboard** (Recommended): Perfect match for cycle/process content with chalkboard aesthetic
2. **hub-spoke + technical-schematic**: For emphasizing central concepts with technical precision
3. **bento-grid + craft-handmade**: For multiple topic overview with friendly handmade feel

View File

@@ -1,158 +1,158 @@
Create a professional infographic following these specifications:
## Image Specifications
- **Type**: Infographic
- **Layout**: circular-flow (Cyclic process showing continuous or recurring steps)
- **Style**: chalkboard (Black chalkboard background with colorful chalk drawing style)
- **Aspect Ratio**: 16:9
- **Language**: English
## Core Principles
- Follow the layout structure precisely for information architecture
- Apply style aesthetics consistently throughout
- If content involves sensitive or copyrighted figures, create stylistically similar alternatives
- Keep information concise, highlight keywords and core concepts
- Use ample whitespace for visual clarity
- Maintain clear visual hierarchy
## Text Requirements
- All text must match the specified style treatment
- Main titles should be prominent and readable
- Key concepts should be visually emphasized
- Labels should be clear and appropriately sized
- Use the specified language for all text content
## Layout Guidelines
- Circular arrangement
- Steps around the circle
- Arrows showing direction
- No clear start/end (continuous)
- Center can hold main concept
- Circle or ring shape
- Directional arrows
- Step nodes evenly spaced
- Icons per step
- Optional center element
## Style Guidelines
- **Background**: Chalkboard Black (#1A1A1A) or Dark Green-Black (#1C2B1C)
- **Texture**: Realistic chalkboard texture with subtle scratches, dust particles, and faint eraser marks
- **Typography**: Hand-drawn chalk lettering style with visible chalk texture. Imperfect baseline adds authenticity.
- **Color Palette**:
- Background: Chalkboard Black (#1A1A1A)
- Primary Text: Chalk White (#F5F5F5)
- Accent 1: Chalk Yellow (#FFE566)
- Accent 2: Chalk Pink (#FF9999)
- Accent 3: Chalk Blue (#66B3FF)
- Accent 4: Chalk Green (#90EE90)
- Accent 5: Chalk Orange (#FFB366)
- **Visual Elements**:
- Hand-drawn chalk illustrations with sketchy, imperfect lines
- Chalk dust effects around text and key elements
- Doodles: stars, arrows, underlines, circles, checkmarks
- Eraser smudges and chalk residue textures
- Wooden frame border optional
- Stick figures and simple icons
- Connection lines with hand-drawn feel
- **Style Rules**:
- Maintain authentic chalk texture on all elements
- Use imperfect, hand-drawn quality throughout
- Add subtle chalk dust and smudge effects
- Create visual hierarchy with color variety
- Include playful doodles and annotations
- DO NOT use perfect geometric shapes
- DO NOT create clean digital-looking lines
---
Generate the infographic based on the content below:
# llm-wiki-sync: Turning Scattered Notes into a Reusable Knowledge Base
## Overview
A cyclical pipeline showing how raw notes are continuously transformed through LLM-powered ingestion into structured wiki pages, then feedback into the knowledge base for reuse.
## The Knowledge Pipeline Cycle (Center Concept)
The llm-wiki-sync pipeline operates as a continuous cycle, not a linear process.
7 stages in the cycle: Raw Note → Ingest → Extract → Source Page → Graph/Site → Reuse → Feedback Loop
## Circular Flow Diagram with 7 Stages:
1. **Raw Note** (Stage 1)
- Original documents stored in raw/ folder
- Contains unprocessed information awaiting structure
- Icon: Stack of paper/note icon
2. **Ingest** (Stage 2)
- LLM analyzes and extracts structured information
- Hermes skill triggers Claude Code for ingestion
- Context check against wiki/index.md prevents duplicates
- Icon: Brain/processing icon
3. **Extract** (Stage 3)
- Six key elements extracted from each document:
- Summary (核心主题, 问题域, 方法/机制, 结论/价值)
- Key Claims (Verifiable assertions)
- Key Quotes (Preserved citations)
- Key Entities (LaunchDarkly, HP, etc.)
- Key Concepts (RTO, RPO, Feature Flag, etc.)
- Connections (depends_on, enables, provides)
- Icon: Six circles/callouts
4. **Source Page** (Stage 4)
- Written to wiki/sources/<slug>.md
- Contains frontmatter and standard sections
- Links use [[PageName]] format
- Icon: Document/page icon
5. **Graph & Site** (Stage 5)
- graph.json: Machine-readable graph structure
- graph.html: Interactive visualization
- Quartz: Static site generation
- Icon: Network/graph icon
6. **Reuse** (Stage 6)
- Query, Synthesize, Write, Connect
- Icon: Multiple arrows pointing outward
7. **Feedback Loop** (Stage 7)
- New insights become new raw notes
- Cycle continues indefinitely
- Icon: Circular arrow completing the cycle
## Key Quotes (to include as callouts):
- "RTO is about speed: how fast you get back online. RPO is about data: how much you can afford to lose."
- "Deploy != Release. Feature flags change this. You can deploy code to production without releasing it to users."
## Design Requirements:
- Circular flow with 7 stages evenly spaced around a circle
- Clockwise arrow direction
- Center contains: "llm-wiki-sync" as main concept
- Each stage is a node with icon + label
- Extraction outputs (6 items) shown as callouts or inner ring
- Dark chalkboard background with hand-drawn chalk accents
- Chalk colors for visual hierarchy
- Imperfect, sketchy lines throughout
- Publication-ready quality
- NO clutter - only essential elements
- Clear hierarchy: title > headlines > labels > descriptions
## Text Labels (in English):
- Headline: "The Knowledge Pipeline Cycle"
- Subhead: "How llm-wiki-sync transforms scattered notes into a reusable knowledge base"
- Stage labels: "Raw Note", "Ingest", "Extract", "Source Page", "Graph & Site", "Reuse", "Feedback"
- Center: "llm-wiki-sync"
- Extraction labels: "Summary", "Claims", "Quotes", "Entities", "Concepts", "Connections"
## Key Constraints:
- 16:9 aspect ratio (landscape)
- All text in English
- Chalkboard style (dark background, chalk-like hand-drawn elements)
- Circular flow layout (NOT linear)
- Publication-ready, visually clear
Create a professional infographic following these specifications:
## Image Specifications
- **Type**: Infographic
- **Layout**: circular-flow (Cyclic process showing continuous or recurring steps)
- **Style**: chalkboard (Black chalkboard background with colorful chalk drawing style)
- **Aspect Ratio**: 16:9
- **Language**: English
## Core Principles
- Follow the layout structure precisely for information architecture
- Apply style aesthetics consistently throughout
- If content involves sensitive or copyrighted figures, create stylistically similar alternatives
- Keep information concise, highlight keywords and core concepts
- Use ample whitespace for visual clarity
- Maintain clear visual hierarchy
## Text Requirements
- All text must match the specified style treatment
- Main titles should be prominent and readable
- Key concepts should be visually emphasized
- Labels should be clear and appropriately sized
- Use the specified language for all text content
## Layout Guidelines
- Circular arrangement
- Steps around the circle
- Arrows showing direction
- No clear start/end (continuous)
- Center can hold main concept
- Circle or ring shape
- Directional arrows
- Step nodes evenly spaced
- Icons per step
- Optional center element
## Style Guidelines
- **Background**: Chalkboard Black (#1A1A1A) or Dark Green-Black (#1C2B1C)
- **Texture**: Realistic chalkboard texture with subtle scratches, dust particles, and faint eraser marks
- **Typography**: Hand-drawn chalk lettering style with visible chalk texture. Imperfect baseline adds authenticity.
- **Color Palette**:
- Background: Chalkboard Black (#1A1A1A)
- Primary Text: Chalk White (#F5F5F5)
- Accent 1: Chalk Yellow (#FFE566)
- Accent 2: Chalk Pink (#FF9999)
- Accent 3: Chalk Blue (#66B3FF)
- Accent 4: Chalk Green (#90EE90)
- Accent 5: Chalk Orange (#FFB366)
- **Visual Elements**:
- Hand-drawn chalk illustrations with sketchy, imperfect lines
- Chalk dust effects around text and key elements
- Doodles: stars, arrows, underlines, circles, checkmarks
- Eraser smudges and chalk residue textures
- Wooden frame border optional
- Stick figures and simple icons
- Connection lines with hand-drawn feel
- **Style Rules**:
- Maintain authentic chalk texture on all elements
- Use imperfect, hand-drawn quality throughout
- Add subtle chalk dust and smudge effects
- Create visual hierarchy with color variety
- Include playful doodles and annotations
- DO NOT use perfect geometric shapes
- DO NOT create clean digital-looking lines
---
Generate the infographic based on the content below:
# llm-wiki-sync: Turning Scattered Notes into a Reusable Knowledge Base
## Overview
A cyclical pipeline showing how raw notes are continuously transformed through LLM-powered ingestion into structured wiki pages, then feedback into the knowledge base for reuse.
## The Knowledge Pipeline Cycle (Center Concept)
The llm-wiki-sync pipeline operates as a continuous cycle, not a linear process.
7 stages in the cycle: Raw Note → Ingest → Extract → Source Page → Graph/Site → Reuse → Feedback Loop
## Circular Flow Diagram with 7 Stages:
1. **Raw Note** (Stage 1)
- Original documents stored in raw/ folder
- Contains unprocessed information awaiting structure
- Icon: Stack of paper/note icon
2. **Ingest** (Stage 2)
- LLM analyzes and extracts structured information
- Hermes skill triggers Claude Code for ingestion
- Context check against wiki/index.md prevents duplicates
- Icon: Brain/processing icon
3. **Extract** (Stage 3)
- Six key elements extracted from each document:
- Summary (核心主题, 问题域, 方法/机制, 结论/价值)
- Key Claims (Verifiable assertions)
- Key Quotes (Preserved citations)
- Key Entities (LaunchDarkly, HP, etc.)
- Key Concepts (RTO, RPO, Feature Flag, etc.)
- Connections (depends_on, enables, provides)
- Icon: Six circles/callouts
4. **Source Page** (Stage 4)
- Written to wiki/sources/<slug>.md
- Contains frontmatter and standard sections
- Links use [[PageName]] format
- Icon: Document/page icon
5. **Graph & Site** (Stage 5)
- graph.json: Machine-readable graph structure
- graph.html: Interactive visualization
- Quartz: Static site generation
- Icon: Network/graph icon
6. **Reuse** (Stage 6)
- Query, Synthesize, Write, Connect
- Icon: Multiple arrows pointing outward
7. **Feedback Loop** (Stage 7)
- New insights become new raw notes
- Cycle continues indefinitely
- Icon: Circular arrow completing the cycle
## Key Quotes (to include as callouts):
- "RTO is about speed: how fast you get back online. RPO is about data: how much you can afford to lose."
- "Deploy != Release. Feature flags change this. You can deploy code to production without releasing it to users."
## Design Requirements:
- Circular flow with 7 stages evenly spaced around a circle
- Clockwise arrow direction
- Center contains: "llm-wiki-sync" as main concept
- Each stage is a node with icon + label
- Extraction outputs (6 items) shown as callouts or inner ring
- Dark chalkboard background with hand-drawn chalk accents
- Chalk colors for visual hierarchy
- Imperfect, sketchy lines throughout
- Publication-ready quality
- NO clutter - only essential elements
- Clear hierarchy: title > headlines > labels > descriptions
## Text Labels (in English):
- Headline: "The Knowledge Pipeline Cycle"
- Subhead: "How llm-wiki-sync transforms scattered notes into a reusable knowledge base"
- Stage labels: "Raw Note", "Ingest", "Extract", "Source Page", "Graph & Site", "Reuse", "Feedback"
- Center: "llm-wiki-sync"
- Extraction labels: "Summary", "Claims", "Quotes", "Entities", "Concepts", "Connections"
## Key Constraints:
- 16:9 aspect ratio (landscape)
- All text in English
- Chalkboard style (dark background, chalk-like hand-drawn elements)
- Circular flow layout (NOT linear)
- Publication-ready, visually clear
- No clutter or excessive elements

View File

@@ -1,229 +1,229 @@
# llm-wiki-sync: Turning Scattered Notes into a Reusable Knowledge Base
## Overview
A cyclical pipeline showing how raw notes are continuously transformed through LLM-powered ingestion into structured wiki pages, then feedback into the knowledge base for reuse.
## Learning Objectives
The viewer will understand:
1. The continuous circular flow of llm-wiki-sync from raw notes to reusable knowledge
2. The six key extraction outputs: Summary, Claims, Quotes, Entities, Concepts, Connections
3. How feedback and reuse complete the cycle back to new raw material
---
## Section 1: The Circular Flow (Center Concept)
**Key Concept**: The llm-wiki-sync pipeline operates as a continuous cycle, not a linear process.
**Content**:
- 7 stages in the cycle: Raw Note → Ingest → Extract → Source Page → Graph/Site → Reuse → Feedback Loop
- Each stage feeds into the next, with feedback returning to the beginning
- The cycle is continuous and self-reinforcing
**Visual Element**:
- Type: circular flow diagram
- Subject: 7 stages arranged in a circle with clockwise arrows
- Center label: "llm-wiki-sync Cycle"
- Treatment: chalk style with hand-drawn arrows connecting stages
**Text Labels**:
- Headline: "The Knowledge Pipeline Cycle"
- Stage labels: "Raw Note", "Ingest", "Extract", "Source Page", "Graph/Site", "Reuse", "Feedback"
- Center: "llm-wiki-sync"
---
## Section 2: Stage 1 — Raw Note (Input)
**Key Concept**: Raw notes are the starting point of the cycle.
**Content**:
- Original documents stored in raw/ folder
- Can be any format: markdown, text, research notes
- Contains unprocessed information awaiting structure
**Visual Element**:
- Type: document/note icon
- Subject: Stack of paper or note icon
- Treatment: Chalk sketch style
**Text Labels**:
- Label: "Raw Note"
- Description: "Original input"
---
## Section 3: Stage 2 — Ingest (LLM Analysis)
**Key Concept**: The LLM analyzes raw notes and extracts structured information.
**Content**:
- Hermes skill triggers Claude Code for ingestion
- LLM reads and analyzes the full document
- Context check against wiki/index.md prevents duplicates
**Visual Element**:
- Type: brain/processing icon
- Subject: Brain or gears with chalk lines
- Treatment: Hand-drawn chalk illustration
**Text Labels**:
- Label: "Ingest"
- Description: "LLM Analysis"
---
## Section 4: Stage 3 — Extract (Six Core Outputs)
**Key Concept**: Six key elements are extracted from each document.
**Content**:
1. **Summary**: 核心主题, 问题域, 方法/机制, 结论/价值
2. **Key Claims**: Verifiable assertions extracted from text
3. **Key Quotes**: Preserved citations for reference
4. **Key Entities**: Named people, companies, products (e.g., LaunchDarkly, HP)
5. **Key Concepts**: Abstract terms that can be reused (e.g., RTO, RPO, Feature Flag)
6. **Connections**: Relationships between elements (depends_on, enables, provides)
**Visual Element**:
- Type: 6 callout nodes around center
- Subject: Six boxes or bubbles representing extraction outputs
- Treatment: Chalk circles with icons inside each
**Text Labels**:
- Headline: "Extraction Outputs"
- Labels: "Summary", "Claims", "Quotes", "Entities", "Concepts", "Connections"
---
## Section 5: Stage 4 — Source Page (Structured Output)
**Key Concept**: Extracted information is written as a structured wiki source page.
**Content**:
- Written to wiki/sources/<slug>.md
- Contains frontmatter (id, title, type, tags, sources, last_updated)
- Standard sections: Summary, Key Claims, Key Quotes, Key Concepts, Key Entities, Connections, Contradictions
- Links use [[PageName]] format for interconnections
**Visual Element**:
- Type: document/page icon
- Subject: Page with visible structure headers
- Treatment: Chalk sketch with text lines
**Text Labels**:
- Label: "Source Page"
- Description: "wiki/sources/*.md"
---
## Section 6: Stage 5 — Graph & Static Site
**Key Concept**: Structured pages generate knowledge graphs and static websites.
**Content**:
- graph.json: Machine-readable graph structure
- graph.html: Interactive visualization
- Quartz: Static site generation for sharing/export
- Connections become edges in the knowledge graph
**Visual Element**:
- Type: network/graph icon
- Subject: Connected nodes representing knowledge graph
- Treatment: Chalk diagram with nodes and edges
**Text Labels**:
- Label: "Graph & Site"
- Description: "graph.json + Quartz"
---
## Section 7: Stage 6 — Reuse (Knowledge Application)
**Key Concept**: Structured knowledge enables multiple reuse scenarios.
**Content**:
- Query: Ask questions against the knowledge base
- Synthesize: Create new content from existing knowledge
- Write: Generate articles, reports from source material
- Connect: Link ideas across different source pages
**Visual Element**:
- Type: multiple arrows pointing outward
- Subject: Reuse scenarios as icons (question, document, pen)
- Treatment: Chalk illustration
**Text Labels**:
- Label: "Reuse"
- Sub-labels: "Query", "Synthesize", "Write", "Connect"
---
## Section 8: Stage 7 — Feedback Loop (Continuous Cycle)
**Key Concept**: Reuse generates new raw notes, completing the cycle.
**Content**:
- New insights from synthesis become new raw notes
- Updated knowledge feeds back to raw/ folder
- Cycle continues indefinitely
- Each iteration strengthens the knowledge base
**Visual Element**:
- Type: circular arrow
- Subject: Feedback loop arrow returning to Raw Note stage
- Treatment: Large chalk arrow completing the circle
**Text Labels**:
- Label: "Feedback Loop"
- Description: "New notes → Cycle repeats"
---
## Data Points (Verbatim)
### Key Quotes
- "RTO is about speed: how fast you get back online. RPO is about data: how much you can afford to lose."
- "Deploy != Release. Feature flags change this. You can deploy code to production without releasing it to users."
### Key Entities
- LaunchDarkly (Feature Flag management platform)
- HP (example enterprise)
- Christian Dior (example case)
### Key Concepts
- RTO (Recovery Time Objective)
- RPO (Recovery Point Objective)
- Feature Flag (特性开关)
- Kill Switch (紧急关闭机制)
- 渐进式发布 (Gradual Rollout)
---
## Design Instructions
### Layout Preferences
- Circular flow with 7 stages evenly spaced around a circle
- Clockwise arrow direction
- Center contains the main concept "llm-wiki-sync"
- Each stage is a node with icon + label
- Extraction outputs (6 items) shown as callouts or inner ring
### Style Preferences
- Chalkboard: Dark background (#1A1A1A)
- Hand-drawn chalk style for all elements
- Chalk colors: white, yellow, pink, blue, green, orange
- Imperfect, sketchy lines throughout
- Chalk dust effects for authenticity
### Text Requirements
- All text in English
- Legible font sizes (minimum 14pt for labels)
- Clear hierarchy: title > headlines > labels > descriptions
- Ample whitespace between stages
### Visual Clarity
- Avoid clutter - only essential elements
- Each stage should be clearly distinguishable
- Arrows should clearly indicate flow direction
# llm-wiki-sync: Turning Scattered Notes into a Reusable Knowledge Base
## Overview
A cyclical pipeline showing how raw notes are continuously transformed through LLM-powered ingestion into structured wiki pages, then feedback into the knowledge base for reuse.
## Learning Objectives
The viewer will understand:
1. The continuous circular flow of llm-wiki-sync from raw notes to reusable knowledge
2. The six key extraction outputs: Summary, Claims, Quotes, Entities, Concepts, Connections
3. How feedback and reuse complete the cycle back to new raw material
---
## Section 1: The Circular Flow (Center Concept)
**Key Concept**: The llm-wiki-sync pipeline operates as a continuous cycle, not a linear process.
**Content**:
- 7 stages in the cycle: Raw Note → Ingest → Extract → Source Page → Graph/Site → Reuse → Feedback Loop
- Each stage feeds into the next, with feedback returning to the beginning
- The cycle is continuous and self-reinforcing
**Visual Element**:
- Type: circular flow diagram
- Subject: 7 stages arranged in a circle with clockwise arrows
- Center label: "llm-wiki-sync Cycle"
- Treatment: chalk style with hand-drawn arrows connecting stages
**Text Labels**:
- Headline: "The Knowledge Pipeline Cycle"
- Stage labels: "Raw Note", "Ingest", "Extract", "Source Page", "Graph/Site", "Reuse", "Feedback"
- Center: "llm-wiki-sync"
---
## Section 2: Stage 1 — Raw Note (Input)
**Key Concept**: Raw notes are the starting point of the cycle.
**Content**:
- Original documents stored in raw/ folder
- Can be any format: markdown, text, research notes
- Contains unprocessed information awaiting structure
**Visual Element**:
- Type: document/note icon
- Subject: Stack of paper or note icon
- Treatment: Chalk sketch style
**Text Labels**:
- Label: "Raw Note"
- Description: "Original input"
---
## Section 3: Stage 2 — Ingest (LLM Analysis)
**Key Concept**: The LLM analyzes raw notes and extracts structured information.
**Content**:
- Hermes skill triggers Claude Code for ingestion
- LLM reads and analyzes the full document
- Context check against wiki/index.md prevents duplicates
**Visual Element**:
- Type: brain/processing icon
- Subject: Brain or gears with chalk lines
- Treatment: Hand-drawn chalk illustration
**Text Labels**:
- Label: "Ingest"
- Description: "LLM Analysis"
---
## Section 4: Stage 3 — Extract (Six Core Outputs)
**Key Concept**: Six key elements are extracted from each document.
**Content**:
1. **Summary**: 核心主题, 问题域, 方法/机制, 结论/价值
2. **Key Claims**: Verifiable assertions extracted from text
3. **Key Quotes**: Preserved citations for reference
4. **Key Entities**: Named people, companies, products (e.g., LaunchDarkly, HP)
5. **Key Concepts**: Abstract terms that can be reused (e.g., RTO, RPO, Feature Flag)
6. **Connections**: Relationships between elements (depends_on, enables, provides)
**Visual Element**:
- Type: 6 callout nodes around center
- Subject: Six boxes or bubbles representing extraction outputs
- Treatment: Chalk circles with icons inside each
**Text Labels**:
- Headline: "Extraction Outputs"
- Labels: "Summary", "Claims", "Quotes", "Entities", "Concepts", "Connections"
---
## Section 5: Stage 4 — Source Page (Structured Output)
**Key Concept**: Extracted information is written as a structured wiki source page.
**Content**:
- Written to wiki/sources/<slug>.md
- Contains frontmatter (id, title, type, tags, sources, last_updated)
- Standard sections: Summary, Key Claims, Key Quotes, Key Concepts, Key Entities, Connections, Contradictions
- Links use [[PageName]] format for interconnections
**Visual Element**:
- Type: document/page icon
- Subject: Page with visible structure headers
- Treatment: Chalk sketch with text lines
**Text Labels**:
- Label: "Source Page"
- Description: "wiki/sources/*.md"
---
## Section 6: Stage 5 — Graph & Static Site
**Key Concept**: Structured pages generate knowledge graphs and static websites.
**Content**:
- graph.json: Machine-readable graph structure
- graph.html: Interactive visualization
- Quartz: Static site generation for sharing/export
- Connections become edges in the knowledge graph
**Visual Element**:
- Type: network/graph icon
- Subject: Connected nodes representing knowledge graph
- Treatment: Chalk diagram with nodes and edges
**Text Labels**:
- Label: "Graph & Site"
- Description: "graph.json + Quartz"
---
## Section 7: Stage 6 — Reuse (Knowledge Application)
**Key Concept**: Structured knowledge enables multiple reuse scenarios.
**Content**:
- Query: Ask questions against the knowledge base
- Synthesize: Create new content from existing knowledge
- Write: Generate articles, reports from source material
- Connect: Link ideas across different source pages
**Visual Element**:
- Type: multiple arrows pointing outward
- Subject: Reuse scenarios as icons (question, document, pen)
- Treatment: Chalk illustration
**Text Labels**:
- Label: "Reuse"
- Sub-labels: "Query", "Synthesize", "Write", "Connect"
---
## Section 8: Stage 7 — Feedback Loop (Continuous Cycle)
**Key Concept**: Reuse generates new raw notes, completing the cycle.
**Content**:
- New insights from synthesis become new raw notes
- Updated knowledge feeds back to raw/ folder
- Cycle continues indefinitely
- Each iteration strengthens the knowledge base
**Visual Element**:
- Type: circular arrow
- Subject: Feedback loop arrow returning to Raw Note stage
- Treatment: Large chalk arrow completing the circle
**Text Labels**:
- Label: "Feedback Loop"
- Description: "New notes → Cycle repeats"
---
## Data Points (Verbatim)
### Key Quotes
- "RTO is about speed: how fast you get back online. RPO is about data: how much you can afford to lose."
- "Deploy != Release. Feature flags change this. You can deploy code to production without releasing it to users."
### Key Entities
- LaunchDarkly (Feature Flag management platform)
- HP (example enterprise)
- Christian Dior (example case)
### Key Concepts
- RTO (Recovery Time Objective)
- RPO (Recovery Point Objective)
- Feature Flag (特性开关)
- Kill Switch (紧急关闭机制)
- 渐进式发布 (Gradual Rollout)
---
## Design Instructions
### Layout Preferences
- Circular flow with 7 stages evenly spaced around a circle
- Clockwise arrow direction
- Center contains the main concept "llm-wiki-sync"
- Each stage is a node with icon + label
- Extraction outputs (6 items) shown as callouts or inner ring
### Style Preferences
- Chalkboard: Dark background (#1A1A1A)
- Hand-drawn chalk style for all elements
- Chalk colors: white, yellow, pink, blue, green, orange
- Imperfect, sketchy lines throughout
- Chalk dust effects for authenticity
### Text Requirements
- All text in English
- Legible font sizes (minimum 14pt for labels)
- Clear hierarchy: title > headlines > labels > descriptions
- Ample whitespace between stages
### Visual Clarity
- Avoid clutter - only essential elements
- Each stage should be clearly distinguishable
- Arrows should clearly indicate flow direction
- Publication-ready quality

View File

@@ -1,182 +1,182 @@
# Major Incident Definition - Infographic Prompts
# Generated: 2026-04-20
# Layout: Venn Diagram
# Style: Cyberpunk Neon
# Aspect: 16:9
# Language: English
---
## Part 1: System Prompt (Image Specifications + Core Principles)
Create a professional infographic following these specifications:
**Image Specifications**
- **Type**: Infographic
- **Layout**: Venn Diagram (venn-diagram)
- **Style**: Cyberpunk Neon
- **Aspect Ratio**: 16:9
- **Language**: English
**Core Principles**
- Follow the layout structure precisely for information architecture
- Apply style aesthetics consistently throughout
- If content involves sensitive or copyrighted figures, create stylistically similar alternatives
- Keep information concise, highlight keywords and core concepts
- Use ample whitespace for visual clarity
- Maintain clear visual hierarchy
**Text Requirements**
- All text must match the specified style treatment
- Main titles should be prominent and readable
- Key concepts should be visually emphasized
- Labels should be clear and appropriately sized
- Use English for all text content
---
## Part 2: Style Lock Prompt (Layout Guidelines + Style Guidelines)
### Layout Guidelines (Venn Diagram)
**Structure**: Three overlapping circles forming a central intersection area
**Visual Elements**:
- Three large overlapping circles representing different categories of Major Incident
- Central intersection zone highlighting the combined criteria
- Each circle contains key concepts specific to its category
- Outer labels for each circle region
- Connection lines from concepts to their respective regions
- Minimal text, emphasize visual grouping
**Composition**:
- Canvas divided into three equal sections for each circle
- 30-40% overlap area in center for intersection
- Ample negative space between elements
- Clear visual boundaries between overlapping regions
### Style Guidelines (Cyberpunk Neon)
**Color Palette**:
- Primary: Neon pink (#FF00FF), cyan (#00FFFF), electric blue
- Background: Deep black (#0A0A0A), dark purple gradients
- Accents: Neon glow effects, chrome reflections
**Visual Elements**:
- Glowing neon outlines on all circle boundaries
- Dark atmospheric backgrounds with subtle grid patterns
- Digital glitch effects on text
- Circuit patterns along connection lines
- Holographic elements in intersection zone
- Rain and reflections on edges
**Typography**:
- Glowing neon text (cyan for primary labels, pink for emphasis)
- Digital/tech monospace fonts
- Subtle flickering effects on key terms
- Outlined glow letters for main title
- All caps for section headers
**Effects**:
- Neon glow (2-3px bloom) around all text and shapes
- Gradient fills with 40-60% opacity
- Scanline overlay at 10% opacity
- Chromatic aberration on text edges
---
## Part 3: Content Structure Prompt (Text Requirements + Content)
### Main Title
**MAJOR INCIDENT DEFINITION**
*Highest Severity Level (S1/P1/Critical)*
### Three Venn Circles Content
**Circle 1: Business Impact**
- Total Service Outage
- Critical Feature Failure
- Data Corruption/Loss
- Security Breach
- Regulatory Compliance Risk
- High-Impact SLA Breach
**Circle 2: Incident Response**
- Immediate Actions (0-15 min)
- Automated Monitoring Alerts
- Incident Commander Assigned
- Major Incident Bridge
- Customer Communication
- Investigation & Mitigation (15-60 min)
- Root Cause Analysis
- Rollback/Hotfix
- Failover to Backup
- Workarounds
- Recovery & Post-Mortem (1-24h+)
- Full Service Restored
- RCA Report Published
- Long-Term Fixes
**Circle 3: Prevention Measures**
- High Availability Architecture
- Chaos Engineering & Load Testing
- Real-Time Monitoring & Alerting
- Automated Rollbacks
- Strict Change Management
- Security Hardening & Compliance
### Central Intersection (All Three Circles)
The intersection represents the critical overlap showing that Major Incidents require:
- Cross-team collaboration
- Immediate response (15-30 min SLA)
- Significant business impact
- Coordinated resolution
### Key Criteria Table (Bottom Section)
| Criteria | Description |
|----------|-------------|
| Scope | Multiple tenants/customers, entire platform |
| Business Criticality | Severe financial/reputational impact |
| Resolution Time | Immediate, 15-30 min acknowledgment |
| Workload Impact | Cloud Ops, DevOps, Security, Support |
| Regulatory Compliance | Legal/security obligations at risk |
### Footer Text
"Requires Swift, Coordinated Response to Minimize Downtime"
---
## Visual Composition for 16:9
- **Top Section** (10%): Title with neon glow effect
- **Middle Section** (70%): Three overlapping circles with content
- **Bottom Section** (20%): Criteria table in minimalist style
### Color Assignment per Circle
- Circle 1 (Business Impact): Cyan (#00FFFF) neon outline
- Circle 2 (Incident Response): Neon Pink (#FF00FF) neon outline
- Circle 3 (Prevention): Electric Blue (#0080FF) neon outline
- Intersection: White glow with purple tint
- Background: Deep black (#0A0A0A) with dark purple gradient
---
# Major Incident Definition - Infographic Prompts
# Generated: 2026-04-20
# Layout: Venn Diagram
# Style: Cyberpunk Neon
# Aspect: 16:9
# Language: English
---
## Part 1: System Prompt (Image Specifications + Core Principles)
Create a professional infographic following these specifications:
**Image Specifications**
- **Type**: Infographic
- **Layout**: Venn Diagram (venn-diagram)
- **Style**: Cyberpunk Neon
- **Aspect Ratio**: 16:9
- **Language**: English
**Core Principles**
- Follow the layout structure precisely for information architecture
- Apply style aesthetics consistently throughout
- If content involves sensitive or copyrighted figures, create stylistically similar alternatives
- Keep information concise, highlight keywords and core concepts
- Use ample whitespace for visual clarity
- Maintain clear visual hierarchy
**Text Requirements**
- All text must match the specified style treatment
- Main titles should be prominent and readable
- Key concepts should be visually emphasized
- Labels should be clear and appropriately sized
- Use English for all text content
---
## Part 2: Style Lock Prompt (Layout Guidelines + Style Guidelines)
### Layout Guidelines (Venn Diagram)
**Structure**: Three overlapping circles forming a central intersection area
**Visual Elements**:
- Three large overlapping circles representing different categories of Major Incident
- Central intersection zone highlighting the combined criteria
- Each circle contains key concepts specific to its category
- Outer labels for each circle region
- Connection lines from concepts to their respective regions
- Minimal text, emphasize visual grouping
**Composition**:
- Canvas divided into three equal sections for each circle
- 30-40% overlap area in center for intersection
- Ample negative space between elements
- Clear visual boundaries between overlapping regions
### Style Guidelines (Cyberpunk Neon)
**Color Palette**:
- Primary: Neon pink (#FF00FF), cyan (#00FFFF), electric blue
- Background: Deep black (#0A0A0A), dark purple gradients
- Accents: Neon glow effects, chrome reflections
**Visual Elements**:
- Glowing neon outlines on all circle boundaries
- Dark atmospheric backgrounds with subtle grid patterns
- Digital glitch effects on text
- Circuit patterns along connection lines
- Holographic elements in intersection zone
- Rain and reflections on edges
**Typography**:
- Glowing neon text (cyan for primary labels, pink for emphasis)
- Digital/tech monospace fonts
- Subtle flickering effects on key terms
- Outlined glow letters for main title
- All caps for section headers
**Effects**:
- Neon glow (2-3px bloom) around all text and shapes
- Gradient fills with 40-60% opacity
- Scanline overlay at 10% opacity
- Chromatic aberration on text edges
---
## Part 3: Content Structure Prompt (Text Requirements + Content)
### Main Title
**MAJOR INCIDENT DEFINITION**
*Highest Severity Level (S1/P1/Critical)*
### Three Venn Circles Content
**Circle 1: Business Impact**
- Total Service Outage
- Critical Feature Failure
- Data Corruption/Loss
- Security Breach
- Regulatory Compliance Risk
- High-Impact SLA Breach
**Circle 2: Incident Response**
- Immediate Actions (0-15 min)
- Automated Monitoring Alerts
- Incident Commander Assigned
- Major Incident Bridge
- Customer Communication
- Investigation & Mitigation (15-60 min)
- Root Cause Analysis
- Rollback/Hotfix
- Failover to Backup
- Workarounds
- Recovery & Post-Mortem (1-24h+)
- Full Service Restored
- RCA Report Published
- Long-Term Fixes
**Circle 3: Prevention Measures**
- High Availability Architecture
- Chaos Engineering & Load Testing
- Real-Time Monitoring & Alerting
- Automated Rollbacks
- Strict Change Management
- Security Hardening & Compliance
### Central Intersection (All Three Circles)
The intersection represents the critical overlap showing that Major Incidents require:
- Cross-team collaboration
- Immediate response (15-30 min SLA)
- Significant business impact
- Coordinated resolution
### Key Criteria Table (Bottom Section)
| Criteria | Description |
|----------|-------------|
| Scope | Multiple tenants/customers, entire platform |
| Business Criticality | Severe financial/reputational impact |
| Resolution Time | Immediate, 15-30 min acknowledgment |
| Workload Impact | Cloud Ops, DevOps, Security, Support |
| Regulatory Compliance | Legal/security obligations at risk |
### Footer Text
"Requires Swift, Coordinated Response to Minimize Downtime"
---
## Visual Composition for 16:9
- **Top Section** (10%): Title with neon glow effect
- **Middle Section** (70%): Three overlapping circles with content
- **Bottom Section** (20%): Criteria table in minimalist style
### Color Assignment per Circle
- Circle 1 (Business Impact): Cyan (#00FFFF) neon outline
- Circle 2 (Incident Response): Neon Pink (#FF00FF) neon outline
- Circle 3 (Prevention): Electric Blue (#0080FF) neon outline
- Intersection: White glow with purple tint
- Background: Deep black (#0A0A0A) with dark purple gradient
---
**Note**: Do not generate the actual image. Output only the prompt for image generation.

View File

@@ -1,256 +1,256 @@
# Wiki Sync 循环流程信息图 - 提示词
> 日期2026-04-20
> 布局circular-flow
> 风格chalkboard
---
## 一、系统提示词
```
Create a professional infographic following these specifications:
## Image Specifications
- **Type**: Infographic
- **Layout**: circular-flow
- **Style**: chalkboard
- **Aspect Ratio**: 16:9 (landscape)
- **Language**: 中文 (Simplified Chinese)
## Core Principles
- Follow the layout structure precisely for information architecture
- Apply style aesthetics consistently throughout
- If content involves sensitive or copyrighted figures, create stylistically similar alternatives
- Keep information concise, highlight keywords and core concepts
- Use ample whitespace for visual clarity
- Maintain clear visual hierarchy
## Text Requirements
- All text must match the specified style treatment
- Main titles should be prominent and readable
- Key concepts should be visually emphasized
- Labels should be clear and appropriately sized
- Use Chinese for all text content
## Layout Guidelines
### Circular Flow Layout
- 循环流程:展示持续运行的过程
- 环形排列:所有节点围绕中心点均布
- 方向箭头:显示顺时针或逆时针流动方向
- 无明确起点/终点(持续循环)
- 中心可放置核心概念或系统名称
### Wiki Sync 循环结构
1. 9个步骤节点均匀分布在圆环上
2. 循环箭头连接各步骤
3. 中心显示 "Wiki Sync" 系统名称
4. 标题位于圆环上方
## Style Guidelines
### Chalkboard Style
- 背景:黑板黑 (#1A1A1A) 或深绿黑板色 (#1C2B1C)
- 纹理:真实黑板质感,伴有划痕、粉尘、橡皮擦痕迹
- 线条手绘粉笔质感不完美的sketchy线条
- 颜色:仅使用粉笔色板
- 字体:手绘粉笔字体
- 装饰:手绘星星、下划线、箭头、圆圈、问号
```
---
## 二、风格锁定提示词
```
## Style Lock — 必须遵守
本信息图使用 chalkboard粉笔黑板风格。
### 背景
- 颜色:#1C2B1C深绿黑板或 #1A1A1A黑板黑
- 纹理:真实黑板纹理,细微划痕、粉尘粒子、橡皮擦痕迹
- 木质框边(可选):手绘木纹线条
### 粉笔线条质量
- 所有线条必须是手绘的、不完美的sketchy风格
- 轻微抖动和歪斜
- 不要干净的矢量线条
### 颜色色板(严格使用)
| 角色 | 颜色 | Hex |
|-----|------|-----|
| 主文字/轮廓 | 粉笔白 | #F5F5F5 |
| 高亮/下划线 | 粉笔黄 | #FFE566 |
| 次要高亮 | 粉笔粉 | #FF9999 |
| 技术元素/图表 | 粉笔蓝 | #66B3FF |
| 成功/自然 | 粉笔绿 | #90EE90 |
| 警告/能量 | 粉笔橙 | #FFB366 |
| 木框 | 棕色 | #8B6914 |
### 装饰元素
- 5-7角手绘星星不规则
- 手绘波浪下划线
- 手绘箭头(歪斜的笔触 + 箭头)
- 角落的粉笔问号、圆圈
- 底部散落的粉笔粉末
### 禁止出现
- 渐变
- 完美几何形状
- 写实元素
- 扁平数字图标
### 布局规则circular-flow
- 9个节点均布于圆环上
- 顺时针方向箭头连接各节点
- 中心放置核心概念
- 标题在圆环顶部
## Layout Guidelines - Circular Flow
### 环形结构
- 步骤节点9个步骤均匀分布在圆周上
- 方向:顺时针流动的箭头
- 中心:系统名称 "Wiki Sync"
- 标题:信息图主标题在顶部
### 节点内容
每个节点包含:
- 步骤编号(圆形框)
- 步骤名称(粉笔白大字)
- 简要描述1-2行较小字体
- 对应图标或符号
### 节点顺序Wiki Sync 流程)
1. Cron 触发
2. 加载 skill
3. 启动 TMUX
4. 启动 Claude Code
5. 发送任务指令
6. 执行 9 步 ingest
7. 解析 SLUG
8. 更新 manifest
9. Telegram 通知
```
---
## 三、具体内容提示词
```
INFOGRAPHIC: Wiki Sync 系统循环流程
## 主标题
- 位置:圆环顶部居中
- 内容:"Wiki Sync 自动同步系统"
- 字体:大型粉笔白(#F5F5F5手写体
- 装饰:粉笔黄(#FFE566双下划线波浪形手绘
- 两侧:粉笔粉(#FF9999小星星装饰
## 圆环中心
- 圆形区域放置系统图标或 "WS" 字样
- 颜色:粉笔蓝(#66B3FF轮廓
- 背景:深绿黑板色 #1C2B1C
## 循环节点(顺时针排列)
### 节点 1Cron 触发
- 编号圆圈:粉笔黄
- 步骤名:① Cron 触发
- 描述定时器启动每15分钟
- 图标:时钟或定时器手绘
### 节点 2加载 Skill
- 编号圆圈:粉笔黄
- 步骤名:② 加载 Skill
- 描述:加载 llm-wiki-sync
- 图标:齿轮或技能图标
### 节点 3检查待摄
- 编号圆圈:粉笔粉
- 步骤名:③ 待摄检查
- 描述sync.py --check
- 图标:文档或清单
### 节点 4启动 TMUX
- 编号圆圈:粉笔粉
- 步骤名:④ 启动 TMUX
- 描述:创建会话
- 图标:终端图标
### 节点 5Claude Code
- 编号圆圈:粉笔蓝
- 步骤名:⑤ Claude Code
- 描述bypassPermissions
- 图标AI/机器人
### 节点 6执行 Ingest
- 编号圆圈:粉笔蓝
- 步骤名:执行 Ingest
- 描述9步标准流程
- 图标:输入箭头
### 节点 7解析 SLUG
- 编号圆圈:粉笔绿
- 步骤名:⑦ 解析 SLUG
- 描述:从输出提取标识符
- 图标:标签/条形码
### 节点 8更新 Manifest
- 编号圆圈:粉笔绿
- 步骤名:更新 Manifest
- 描述:写入 JSON 状态
- 图标:数据保存
### 节点 9Telegram 通知
- 编号圆圈:粉笔橙
- 步骤名: Telegram 通知
- 描述:发送同步结果
- 图标:消息气泡
## 循环箭头
- 颜色:粉笔蓝(#66B3FF
- 风格:手绘歪斜箭头
- 方向:顺时针
- 每两个节点之间
## 底部装饰
- 左侧:粉笔粉 "Claude Code Agent"
- 右侧:粉笔黄 "持续运行 ∞"
- 底部边缘:散落粉笔粉末痕迹
- 角落:粉笔问号、圆圈装饰
## 整体效果
- 深绿黑板背景 (#1C2B1C)
- 所有文字使用粉笔质感手写体
- 节点框使用手绘不规则圆形/圆角矩形
- 箭头使用手绘歪斜线条
- 保持信息密度,流程清晰
```
---
## 四、使用说明
```
使用方法:
1. 使用支持 circular-flow 布局 + chalkboard 风格的 AI 图像生成工具
2. 建议比例16:9 (landscape)
3. 语言:中文(简体)
4. 参考本文件的三部分提示词组合使用
提示词组合优先级:
1. 第一部分(系统提示词)- 基础规范
2. 第二部分(风格锁定)- 必须遵守的约束
3. 具体内容提示词 - 实际图像内容
```
---
# Wiki Sync 循环流程信息图 - 提示词
> 日期2026-04-20
> 布局circular-flow
> 风格chalkboard
---
## 一、系统提示词
```
Create a professional infographic following these specifications:
## Image Specifications
- **Type**: Infographic
- **Layout**: circular-flow
- **Style**: chalkboard
- **Aspect Ratio**: 16:9 (landscape)
- **Language**: 中文 (Simplified Chinese)
## Core Principles
- Follow the layout structure precisely for information architecture
- Apply style aesthetics consistently throughout
- If content involves sensitive or copyrighted figures, create stylistically similar alternatives
- Keep information concise, highlight keywords and core concepts
- Use ample whitespace for visual clarity
- Maintain clear visual hierarchy
## Text Requirements
- All text must match the specified style treatment
- Main titles should be prominent and readable
- Key concepts should be visually emphasized
- Labels should be clear and appropriately sized
- Use Chinese for all text content
## Layout Guidelines
### Circular Flow Layout
- 循环流程:展示持续运行的过程
- 环形排列:所有节点围绕中心点均布
- 方向箭头:显示顺时针或逆时针流动方向
- 无明确起点/终点(持续循环)
- 中心可放置核心概念或系统名称
### Wiki Sync 循环结构
1. 9个步骤节点均匀分布在圆环上
2. 循环箭头连接各步骤
3. 中心显示 "Wiki Sync" 系统名称
4. 标题位于圆环上方
## Style Guidelines
### Chalkboard Style
- 背景:黑板黑 (#1A1A1A) 或深绿黑板色 (#1C2B1C)
- 纹理:真实黑板质感,伴有划痕、粉尘、橡皮擦痕迹
- 线条手绘粉笔质感不完美的sketchy线条
- 颜色:仅使用粉笔色板
- 字体:手绘粉笔字体
- 装饰:手绘星星、下划线、箭头、圆圈、问号
```
---
## 二、风格锁定提示词
```
## Style Lock — 必须遵守
本信息图使用 chalkboard粉笔黑板风格。
### 背景
- 颜色:#1C2B1C深绿黑板或 #1A1A1A黑板黑
- 纹理:真实黑板纹理,细微划痕、粉尘粒子、橡皮擦痕迹
- 木质框边(可选):手绘木纹线条
### 粉笔线条质量
- 所有线条必须是手绘的、不完美的sketchy风格
- 轻微抖动和歪斜
- 不要干净的矢量线条
### 颜色色板(严格使用)
| 角色 | 颜色 | Hex |
|-----|------|-----|
| 主文字/轮廓 | 粉笔白 | #F5F5F5 |
| 高亮/下划线 | 粉笔黄 | #FFE566 |
| 次要高亮 | 粉笔粉 | #FF9999 |
| 技术元素/图表 | 粉笔蓝 | #66B3FF |
| 成功/自然 | 粉笔绿 | #90EE90 |
| 警告/能量 | 粉笔橙 | #FFB366 |
| 木框 | 棕色 | #8B6914 |
### 装饰元素
- 5-7角手绘星星不规则
- 手绘波浪下划线
- 手绘箭头(歪斜的笔触 + 箭头)
- 角落的粉笔问号、圆圈
- 底部散落的粉笔粉末
### 禁止出现
- 渐变
- 完美几何形状
- 写实元素
- 扁平数字图标
### 布局规则circular-flow
- 9个节点均布于圆环上
- 顺时针方向箭头连接各节点
- 中心放置核心概念
- 标题在圆环顶部
## Layout Guidelines - Circular Flow
### 环形结构
- 步骤节点9个步骤均匀分布在圆周上
- 方向:顺时针流动的箭头
- 中心:系统名称 "Wiki Sync"
- 标题:信息图主标题在顶部
### 节点内容
每个节点包含:
- 步骤编号(圆形框)
- 步骤名称(粉笔白大字)
- 简要描述1-2行较小字体
- 对应图标或符号
### 节点顺序Wiki Sync 流程)
1. Cron 触发
2. 加载 skill
3. 启动 TMUX
4. 启动 Claude Code
5. 发送任务指令
6. 执行 9 步 ingest
7. 解析 SLUG
8. 更新 manifest
9. Telegram 通知
```
---
## 三、具体内容提示词
```
INFOGRAPHIC: Wiki Sync 系统循环流程
## 主标题
- 位置:圆环顶部居中
- 内容:"Wiki Sync 自动同步系统"
- 字体:大型粉笔白(#F5F5F5手写体
- 装饰:粉笔黄(#FFE566双下划线波浪形手绘
- 两侧:粉笔粉(#FF9999小星星装饰
## 圆环中心
- 圆形区域放置系统图标或 "WS" 字样
- 颜色:粉笔蓝(#66B3FF轮廓
- 背景:深绿黑板色 #1C2B1C
## 循环节点(顺时针排列)
### 节点 1Cron 触发
- 编号圆圈:粉笔黄
- 步骤名:① Cron 触发
- 描述定时器启动每15分钟
- 图标:时钟或定时器手绘
### 节点 2加载 Skill
- 编号圆圈:粉笔黄
- 步骤名:② 加载 Skill
- 描述:加载 llm-wiki-sync
- 图标:齿轮或技能图标
### 节点 3检查待摄
- 编号圆圈:粉笔粉
- 步骤名:③ 待摄检查
- 描述sync.py --check
- 图标:文档或清单
### 节点 4启动 TMUX
- 编号圆圈:粉笔粉
- 步骤名:④ 启动 TMUX
- 描述:创建会话
- 图标:终端图标
### 节点 5Claude Code
- 编号圆圈:粉笔蓝
- 步骤名:⑤ Claude Code
- 描述bypassPermissions
- 图标AI/机器人
### 节点 6执行 Ingest
- 编号圆圈:粉笔蓝
- 步骤名:执行 Ingest
- 描述9步标准流程
- 图标:输入箭头
### 节点 7解析 SLUG
- 编号圆圈:粉笔绿
- 步骤名:⑦ 解析 SLUG
- 描述:从输出提取标识符
- 图标:标签/条形码
### 节点 8更新 Manifest
- 编号圆圈:粉笔绿
- 步骤名:更新 Manifest
- 描述:写入 JSON 状态
- 图标:数据保存
### 节点 9Telegram 通知
- 编号圆圈:粉笔橙
- 步骤名: Telegram 通知
- 描述:发送同步结果
- 图标:消息气泡
## 循环箭头
- 颜色:粉笔蓝(#66B3FF
- 风格:手绘歪斜箭头
- 方向:顺时针
- 每两个节点之间
## 底部装饰
- 左侧:粉笔粉 "Claude Code Agent"
- 右侧:粉笔黄 "持续运行 ∞"
- 底部边缘:散落粉笔粉末痕迹
- 角落:粉笔问号、圆圈装饰
## 整体效果
- 深绿黑板背景 (#1C2B1C)
- 所有文字使用粉笔质感手写体
- 节点框使用手绘不规则圆形/圆角矩形
- 箭头使用手绘歪斜线条
- 保持信息密度,流程清晰
```
---
## 四、使用说明
```
使用方法:
1. 使用支持 circular-flow 布局 + chalkboard 风格的 AI 图像生成工具
2. 建议比例16:9 (landscape)
3. 语言:中文(简体)
4. 参考本文件的三部分提示词组合使用
提示词组合优先级:
1. 第一部分(系统提示词)- 基础规范
2. 第二部分(风格锁定)- 必须遵守的约束
3. 具体内容提示词 - 实际图像内容
```
---
*Generated: 2026-04-20*

View File

@@ -1,148 +1,148 @@
# Wiki Sync 系统信息图提示词
> 生成日期2026-04-20
> 用途:为 LLM Wiki 自动化同步系统生成信息图
---
## 第一部分系统提示词System Prompt
```
Create a professional infographic following these specifications:
## Image Specifications
- Type: Infographic
- Layout: hub-spoke
- Style: corporate-memphis
- Aspect Ratio: 16:9
- Language: 简体中文
## Core Principles
- Follow the hub-spoke layout structure precisely for information architecture
- Apply corporate-memphis style aesthetics consistently throughout
- If content involves sensitive or copyrighted figures, create stylistically similar alternatives
- Keep information concise, highlight keywords and core concepts
- Use ample whitespace for visual clarity
- Maintain clear visual hierarchy
## Text Requirements
- All text must match the specified style treatment
- Main titles should be prominent and readable
- Key concepts should be visually emphasized
- Labels should be clear and appropriately sized
- Use simplified Chinese for all text content
## Layout Guidelines
- Prominent central hub at the center of the canvas
- Clear spoke lines radiating outward from center
- Nodes at spoke ends with consistent styling
- Even distribution of spokes around the hub
- Icons representing each spoke item
- Brief labels near each node
## Style Guidelines
- Flat vector illustration style
- Bright, saturated colors: purple, orange, teal, yellow
- White or light pastel background
- Disproportionate human figures (optional)
- Abstract geometric shapes and floating elements
- Solid fills without outlines
- Clean sans-serif typography with bold headings
- Professional but friendly appearance
---
Generate the infographic based on the content below:
Text labels (in 简体中文):
```
---
## 第二部分风格锁定提示词Style Lock Prompt
```
## 布局规范hub-spoke
- 中心:突出显示核心主题"Wiki Sync 自动化同步系统"
- 辐射6条均匀分布的辐条连接中心与外圈节点
- 节点:每个辐条末端放置一个组件节点,使用统一的图标风格
- 连接:辐条使用浅灰色线条,带有标签
## 视觉元素corporate-memphis
- 背景:纯白色或浅奶油色 (#FFF8F0)
- 主色:明快的紫色 (#8B5CF6)、橙色 (#F97316)、青绿色 (#14B8A6)、黄色 (#FBBF24)
- 图形:扁平化几何图形填充,无描边
- 人物:抽象的不成比例人物剪影
- 装饰:漂浮的几何元素(圆形、方形、三角形)
- 字体:思源黑体或类似无衬线字体,标题加粗
## 图标风格
- 简化的线框图标
- 保持一致的线条粗细2-3px
- 圆角处理
```
---
## 第三部分具体内容提示词Content Prompt
```
INFOGRAPHIC: Wiki Sync 自动化同步系统
使用 hub-spoke 布局和 corporate-memphis 风格创建信息图。
## 中心主题Hub
- 主标题Wiki Sync 系统
- 副标题LLM Wiki 自动化同步解决方案
## 六个辐条节点Spokes
### 1. 核心组件Core
- 图标:齿轮/工具箱
- 标签sync.py CLI
- 描述manifest 管理、文件追踪
### 2. 工作流Workflow
- 图标:流程图/箭头
- 标签llm-wiki-sync
- 描述9步标准化执行流程
### 3. 定时任务Schedule
- 图标:时钟/日历
- 标签Cron Job
- 描述15分钟自动触发
### 4. 状态追踪Tracking
- 图标:清单/勾选
- 标签manifest.json
- 描述181篇文件状态记录
### 5. 交互模式Interaction
- 图标:终端/命令行
- 标签TMUX + Claude
- 描述bypassPermissions 启动
### 6. 交付通知Delivery
- 图标:消息/ telegram
- 标签Telegram Bot
- 描述5038825565 消息推送
## 底部补充信息
- 当前状态:已摄入 16 篇 / 待摄入 165 篇
- 关键规则:顺序执行 + SLUG 解析
```
---
## 使用说明
1.**第一部分** 作为系统提示词设置到图像生成模型会话开始时
2.**第二部分** 作为风格参考添加到提示词中
3.**第三部分** 作为具体内容提示词发送给图像生成模型
4. 生成的图像应呈现:
- 白色/浅奶油色背景
- 中心为"Wiki Sync 系统"标题
- 6个均匀分布的辐条节点
- 明快的紫、橙、青、黄配色
- 扁平化几何图标
# Wiki Sync 系统信息图提示词
> 生成日期2026-04-20
> 用途:为 LLM Wiki 自动化同步系统生成信息图
---
## 第一部分系统提示词System Prompt
```
Create a professional infographic following these specifications:
## Image Specifications
- Type: Infographic
- Layout: hub-spoke
- Style: corporate-memphis
- Aspect Ratio: 16:9
- Language: 简体中文
## Core Principles
- Follow the hub-spoke layout structure precisely for information architecture
- Apply corporate-memphis style aesthetics consistently throughout
- If content involves sensitive or copyrighted figures, create stylistically similar alternatives
- Keep information concise, highlight keywords and core concepts
- Use ample whitespace for visual clarity
- Maintain clear visual hierarchy
## Text Requirements
- All text must match the specified style treatment
- Main titles should be prominent and readable
- Key concepts should be visually emphasized
- Labels should be clear and appropriately sized
- Use simplified Chinese for all text content
## Layout Guidelines
- Prominent central hub at the center of the canvas
- Clear spoke lines radiating outward from center
- Nodes at spoke ends with consistent styling
- Even distribution of spokes around the hub
- Icons representing each spoke item
- Brief labels near each node
## Style Guidelines
- Flat vector illustration style
- Bright, saturated colors: purple, orange, teal, yellow
- White or light pastel background
- Disproportionate human figures (optional)
- Abstract geometric shapes and floating elements
- Solid fills without outlines
- Clean sans-serif typography with bold headings
- Professional but friendly appearance
---
Generate the infographic based on the content below:
Text labels (in 简体中文):
```
---
## 第二部分风格锁定提示词Style Lock Prompt
```
## 布局规范hub-spoke
- 中心:突出显示核心主题"Wiki Sync 自动化同步系统"
- 辐射6条均匀分布的辐条连接中心与外圈节点
- 节点:每个辐条末端放置一个组件节点,使用统一的图标风格
- 连接:辐条使用浅灰色线条,带有标签
## 视觉元素corporate-memphis
- 背景:纯白色或浅奶油色 (#FFF8F0)
- 主色:明快的紫色 (#8B5CF6)、橙色 (#F97316)、青绿色 (#14B8A6)、黄色 (#FBBF24)
- 图形:扁平化几何图形填充,无描边
- 人物:抽象的不成比例人物剪影
- 装饰:漂浮的几何元素(圆形、方形、三角形)
- 字体:思源黑体或类似无衬线字体,标题加粗
## 图标风格
- 简化的线框图标
- 保持一致的线条粗细2-3px
- 圆角处理
```
---
## 第三部分具体内容提示词Content Prompt
```
INFOGRAPHIC: Wiki Sync 自动化同步系统
使用 hub-spoke 布局和 corporate-memphis 风格创建信息图。
## 中心主题Hub
- 主标题Wiki Sync 系统
- 副标题LLM Wiki 自动化同步解决方案
## 六个辐条节点Spokes
### 1. 核心组件Core
- 图标:齿轮/工具箱
- 标签sync.py CLI
- 描述manifest 管理、文件追踪
### 2. 工作流Workflow
- 图标:流程图/箭头
- 标签llm-wiki-sync
- 描述9步标准化执行流程
### 3. 定时任务Schedule
- 图标:时钟/日历
- 标签Cron Job
- 描述15分钟自动触发
### 4. 状态追踪Tracking
- 图标:清单/勾选
- 标签manifest.json
- 描述181篇文件状态记录
### 5. 交互模式Interaction
- 图标:终端/命令行
- 标签TMUX + Claude
- 描述bypassPermissions 启动
### 6. 交付通知Delivery
- 图标:消息/ telegram
- 标签Telegram Bot
- 描述5038825565 消息推送
## 底部补充信息
- 当前状态:已摄入 16 篇 / 待摄入 165 篇
- 关键规则:顺序执行 + SLUG 解析
```
---
## 使用说明
1.**第一部分** 作为系统提示词设置到图像生成模型会话开始时
2.**第二部分** 作为风格参考添加到提示词中
3.**第三部分** 作为具体内容提示词发送给图像生成模型
4. 生成的图像应呈现:
- 白色/浅奶油色背景
- 中心为"Wiki Sync 系统"标题
- 6个均匀分布的辐条节点
- 明快的紫、橙、青、黄配色
- 扁平化几何图标
- 思源黑体加粗标题

View File

@@ -1,272 +1,272 @@
# Wiki Sync 系统搭建完整记录
> 日期2026-04-16
> 作者Hermes Agent
> 用途:记录 llm-wiki-agent 自动化同步系统的完整搭建过程
---
## 背景
用户希望将 Obsidian vault 中的 markdown 文件批量摄入到 LLM WikiKarpathy's LLM Wiki中。原有方案是手动逐篇执行效率低下。本次目标搭建自动化系统实现定时自动摄入。
---
## 系统架构
### 架构图
![Wiki Sync Architecture](IMG-20260418081458052.png)
### 同步时序图
![Wiki Sync Sequence Diagram](IMG-20260418081458100.png)
### Ingest 9 步流程图
![[IMG-20260418081458161.png]]
### 核心组件
| 组件 | 位置 | 职责 |
|------|------|------|
| **llm-wiki-sync skill** | `~/.hermes/skills/research/llm-wiki-sync/SKILL.md` | 执行模板和工作流定义 |
| **sync.py** | `/Users/weishen/Git/llm-wiki-agent/tools/sync.py` | manifest 管理、CLI 工具 |
| **manifest.json** | `/Users/weishen/Git/llm-wiki-agent/tools/manifest.json` | 文件状态追踪181篇 |
| **Hermes Cron Job** | 内部调度器 | 每 15 分钟触发一次摄入 |
## 一、初始探索(手动执行阶段)
### 1.1 发现 raw 文件
- 路径:`/Users/weishen/Workspace/nexus/raw/`Obsidian vault
- 通过 symlink 挂载到:`/Users/weishen/Git/llm-wiki-agent/raw/`
- 文件数182 个 markdown 文件
### 1.2 创建 manifest.json
手动扫描 raw 目录,生成初始 manifest
```
# 扫描 raw 目录,提取 frontmatter
for md in Path("raw").glob("**/*.md"):
# 提取 title, ingested, slug 等字段
manifest.append({...})
```
### 1.3 测试 /wiki-ingest
Claude Code 的 `/wiki-ingest` 命令执行 9 步流程:
读取 source 文档
读取 wiki/index.md 和 overview.md
生成 wiki/sources/<slug>.md
更新 wiki/index.md
更新 wiki/overview.md
创建/更新 Entity 页面
创建/更新 Concept 页面
检测并记录冲突
追加 wiki/log.md
## 二、遇到的问题和解决方案
### 问题 1stdin 交互问题
**现象**`claude --print` 模式在非交互环境下无法正常工作stdin 被占用导致命令卡住。
**解决方案**:使用 TMUX 交互模式启动 Claude Code
```bash
# 启动 TMUX session
tmux new-session -d -s wiki-ingest "cd /Users/weishen/Git/llm-wiki-agent && claude --permission-mode bypassPermissions"
# 等待启动完成
sleep 8 && tmux send-keys -t wiki-ingest Enter
# 发送任务
tmux send-keys -t wiki-ingest "请执行任务..."
```
### 问题 2manifest slug 与实际文件不匹配
**现象**manifest 中记录的 slug 和 source_path 与 LLM 实际生成的文件名不一致。
**原因**LLM 根据内容自动生成 slug而不是简单从文件名转换。
**解决方案**:要求 LLM 在任务完成后输出 `SLUG: xxx`,然后 Hermes 解析并更新 manifest
```python
def parse_slug_from_output(output: str) -> str:
"""从 LLM 输出中解析实际使用的 slug"""
match = re.search(r'SLUG:\s*([^\s]+)', output)
return match.group(1) if match else None
```
### 问题 3170 条 ingest 错误记录
**现象**manifest 中有 170 条记录标记为 `ingested=true` 但实际未成功。
**原因**:早期测试时使用 `--print` 模式失败但仍标记为成功。
**解决方案**:使用 `sync.py --reset-failed` 清理错误状态。
---
## 三、自建组件
### 3.1 sync.py CLI 工具
**位置**`/Users/weishen/Git/llm-wiki-agent/tools/sync.py`
**功能**
- `--pending`:列出待摄取文件
- `--check`:预览变化
- `--reset-failed`:重置失败记录
- `--bootstrap`:从现有 wiki sources 重建 manifest
**核心函数**
```python
def get_pending_files() -> list:
"""返回所有未摄入的文件"""
def mark_ingested(file_path: str, slug: str):
"""标记文件为已摄入"""
def reset_failed():
"""重置所有失败状态"""
def parse_slug_from_output(output: str) -> str:
"""从 LLM 输出解析 SLUG"""
```
### 3.2 llm-wiki-sync skill
**位置**`~/.hermes/skills/research/llm-wiki-sync/SKILL.md`
**版本**1.4.0
**核心内容**
- **角色分工**Hermes 编排流程 → Claude Code 执行 ingest
- **关键设计**TMUX 交互模式、顺序执行、SLUG 输出要求
- **TMUX 执行流程**:完整的启动、发送任务、监控、清理流程
- **Ingest Workflow 9 步**:标准化执行步骤
- **Cron Job 自动化**:使用 Hermes 原生 cron job
### 3.3 Hermes Cron Job
**创建命令**
```bash
cronjob create \
--name wiki-sync-15min \
--skill llm-wiki-sync \
--schedule "*/15 * * * *" \
--repeat 999999 \
--deliver "telegram:5038825565" \
--prompt "使用 llm-wiki-sync 技能执行一次 wiki 文章摄入..."
```
**配置**
- Job ID`98265b6998c5`
- Schedule`*/15 * * * *`(每 15 分钟)
- 交付方式TelegramID: 5038825565
- 技能llm-wiki-sync
---
## 四、执行流程(自动化阶段)
### 4.1 Cron Job 触发
1. 每 15 分钟00, 15, 30, 45 分)触发
2. 加载 llm-wiki-sync skill
3. 执行 skill 中的 prompt
### 4.2 实际执行步骤
```
1. 加载 llm-wiki-sync 技能
2. 检查 manifestpython tools/sync.py --pending
3. 启动 TMUX session
4. 启动 Claude CodebypassPermissions
5. 发送任务指令(含 SLUG 输出要求)
6. 监控任务完成tmux capture-pane
7. 解析 SLUG 并更新 manifest.json
8. 清理 TMUX session
9. 输出结果(自动发往 Telegram
```
### 4.3 示例输出
```
## Wiki Sync 完成
| 项目 | 结果 |
|------|------|
| 摄入文件 | raw/Home Office/用Docker中安装Navidrome.md |
| Slug | 用docker中安装navidrome |
| 状态 | ✅ 已完成 |
新增页面:
- wiki/sources/用docker中安装navidrome.md
- wiki/entities/Navidrome.md
- wiki/concepts/Docker-Compose.md
manifest 已更新ingested: true
剩余待摄入165 篇
```
---
## 五、当前状态
| 指标 | 值 |
|------|-----|
| 总文件数 | 181 篇 |
| 已摄入 | 16 篇 |
| 待摄入 | 165 篇 |
| Cron Job 状态 | 运行中 |
| 下次运行 | 18:15:00 |
---
## 六、关键命令速查
```bash
# 检查待摄取文件
cd /Users/weishen/Git/llm-wiki-agent && python tools/sync.py --pending
# 预览变化
python tools/sync.py --check
# 重置失败记录
python tools/sync.py --reset-failed
# 查看 cron job 状态
cronjob --list
# 手动触发 cron job
cronjob --run <job_id>
```
---
## 七、注意事项
1. **必须使用 TMUX**:不能用 subprocess 或 --print 模式
2. **必须顺序执行**:并发会触发 529 rate limit
3. **必须解析 SLUG**LLM 输出的实际 slug 用于更新 manifest
4. **交付方式**:使用 `telegram:5038825565` 发给用户
5. **保留 orphan**:不删除任何原始数据
---
## 八、扩展方向
- [ ] 添加错误重试机制529 时等待后重试)
- [ ] 支持批量摄入(改为每次 3-5 篇)
- [ ] 添加 webhook 通知(不只是 Telegram
- [ ] 统计摄入速率和成功率
---
# Wiki Sync 系统搭建完整记录
> 日期2026-04-16
> 作者Hermes Agent
> 用途:记录 llm-wiki-agent 自动化同步系统的完整搭建过程
---
## 背景
用户希望将 Obsidian vault 中的 markdown 文件批量摄入到 LLM WikiKarpathy's LLM Wiki中。原有方案是手动逐篇执行效率低下。本次目标搭建自动化系统实现定时自动摄入。
---
## 系统架构
### 架构图
![Wiki Sync Architecture](IMG-20260418081458052.png)
### 同步时序图
![Wiki Sync Sequence Diagram](IMG-20260418081458100.png)
### Ingest 9 步流程图
![[IMG-20260418081458161.png]]
### 核心组件
| 组件 | 位置 | 职责 |
|------|------|------|
| **llm-wiki-sync skill** | `~/.hermes/skills/research/llm-wiki-sync/SKILL.md` | 执行模板和工作流定义 |
| **sync.py** | `/Users/weishen/Git/llm-wiki-agent/tools/sync.py` | manifest 管理、CLI 工具 |
| **manifest.json** | `/Users/weishen/Git/llm-wiki-agent/tools/manifest.json` | 文件状态追踪181篇 |
| **Hermes Cron Job** | 内部调度器 | 每 15 分钟触发一次摄入 |
## 一、初始探索(手动执行阶段)
### 1.1 发现 raw 文件
- 路径:`/Users/weishen/Workspace/nexus/raw/`Obsidian vault
- 通过 symlink 挂载到:`/Users/weishen/Git/llm-wiki-agent/raw/`
- 文件数182 个 markdown 文件
### 1.2 创建 manifest.json
手动扫描 raw 目录,生成初始 manifest
```
# 扫描 raw 目录,提取 frontmatter
for md in Path("raw").glob("**/*.md"):
# 提取 title, ingested, slug 等字段
manifest.append({...})
```
### 1.3 测试 /wiki-ingest
Claude Code 的 `/wiki-ingest` 命令执行 9 步流程:
读取 source 文档
读取 wiki/index.md 和 overview.md
生成 wiki/sources/<slug>.md
更新 wiki/index.md
更新 wiki/overview.md
创建/更新 Entity 页面
创建/更新 Concept 页面
检测并记录冲突
追加 wiki/log.md
## 二、遇到的问题和解决方案
### 问题 1stdin 交互问题
**现象**`claude --print` 模式在非交互环境下无法正常工作stdin 被占用导致命令卡住。
**解决方案**:使用 TMUX 交互模式启动 Claude Code
```bash
# 启动 TMUX session
tmux new-session -d -s wiki-ingest "cd /Users/weishen/Git/llm-wiki-agent && claude --permission-mode bypassPermissions"
# 等待启动完成
sleep 8 && tmux send-keys -t wiki-ingest Enter
# 发送任务
tmux send-keys -t wiki-ingest "请执行任务..."
```
### 问题 2manifest slug 与实际文件不匹配
**现象**manifest 中记录的 slug 和 source_path 与 LLM 实际生成的文件名不一致。
**原因**LLM 根据内容自动生成 slug而不是简单从文件名转换。
**解决方案**:要求 LLM 在任务完成后输出 `SLUG: xxx`,然后 Hermes 解析并更新 manifest
```python
def parse_slug_from_output(output: str) -> str:
"""从 LLM 输出中解析实际使用的 slug"""
match = re.search(r'SLUG:\s*([^\s]+)', output)
return match.group(1) if match else None
```
### 问题 3170 条 ingest 错误记录
**现象**manifest 中有 170 条记录标记为 `ingested=true` 但实际未成功。
**原因**:早期测试时使用 `--print` 模式失败但仍标记为成功。
**解决方案**:使用 `sync.py --reset-failed` 清理错误状态。
---
## 三、自建组件
### 3.1 sync.py CLI 工具
**位置**`/Users/weishen/Git/llm-wiki-agent/tools/sync.py`
**功能**
- `--pending`:列出待摄取文件
- `--check`:预览变化
- `--reset-failed`:重置失败记录
- `--bootstrap`:从现有 wiki sources 重建 manifest
**核心函数**
```python
def get_pending_files() -> list:
"""返回所有未摄入的文件"""
def mark_ingested(file_path: str, slug: str):
"""标记文件为已摄入"""
def reset_failed():
"""重置所有失败状态"""
def parse_slug_from_output(output: str) -> str:
"""从 LLM 输出解析 SLUG"""
```
### 3.2 llm-wiki-sync skill
**位置**`~/.hermes/skills/research/llm-wiki-sync/SKILL.md`
**版本**1.4.0
**核心内容**
- **角色分工**Hermes 编排流程 → Claude Code 执行 ingest
- **关键设计**TMUX 交互模式、顺序执行、SLUG 输出要求
- **TMUX 执行流程**:完整的启动、发送任务、监控、清理流程
- **Ingest Workflow 9 步**:标准化执行步骤
- **Cron Job 自动化**:使用 Hermes 原生 cron job
### 3.3 Hermes Cron Job
**创建命令**
```bash
cronjob create \
--name wiki-sync-15min \
--skill llm-wiki-sync \
--schedule "*/15 * * * *" \
--repeat 999999 \
--deliver "telegram:5038825565" \
--prompt "使用 llm-wiki-sync 技能执行一次 wiki 文章摄入..."
```
**配置**
- Job ID`98265b6998c5`
- Schedule`*/15 * * * *`(每 15 分钟)
- 交付方式TelegramID: 5038825565
- 技能llm-wiki-sync
---
## 四、执行流程(自动化阶段)
### 4.1 Cron Job 触发
1. 每 15 分钟00, 15, 30, 45 分)触发
2. 加载 llm-wiki-sync skill
3. 执行 skill 中的 prompt
### 4.2 实际执行步骤
```
1. 加载 llm-wiki-sync 技能
2. 检查 manifestpython tools/sync.py --pending
3. 启动 TMUX session
4. 启动 Claude CodebypassPermissions
5. 发送任务指令(含 SLUG 输出要求)
6. 监控任务完成tmux capture-pane
7. 解析 SLUG 并更新 manifest.json
8. 清理 TMUX session
9. 输出结果(自动发往 Telegram
```
### 4.3 示例输出
```
## Wiki Sync 完成
| 项目 | 结果 |
|------|------|
| 摄入文件 | raw/Home Office/用Docker中安装Navidrome.md |
| Slug | 用docker中安装navidrome |
| 状态 | ✅ 已完成 |
新增页面:
- wiki/sources/用docker中安装navidrome.md
- wiki/entities/Navidrome.md
- wiki/concepts/Docker-Compose.md
manifest 已更新ingested: true
剩余待摄入165 篇
```
---
## 五、当前状态
| 指标 | 值 |
|------|-----|
| 总文件数 | 181 篇 |
| 已摄入 | 16 篇 |
| 待摄入 | 165 篇 |
| Cron Job 状态 | 运行中 |
| 下次运行 | 18:15:00 |
---
## 六、关键命令速查
```bash
# 检查待摄取文件
cd /Users/weishen/Git/llm-wiki-agent && python tools/sync.py --pending
# 预览变化
python tools/sync.py --check
# 重置失败记录
python tools/sync.py --reset-failed
# 查看 cron job 状态
cronjob --list
# 手动触发 cron job
cronjob --run <job_id>
```
---
## 七、注意事项
1. **必须使用 TMUX**:不能用 subprocess 或 --print 模式
2. **必须顺序执行**:并发会触发 529 rate limit
3. **必须解析 SLUG**LLM 输出的实际 slug 用于更新 manifest
4. **交付方式**:使用 `telegram:5038825565` 发给用户
5. **保留 orphan**:不删除任何原始数据
---
## 八、扩展方向
- [ ] 添加错误重试机制529 时等待后重试)
- [ ] 支持批量摄入(改为每次 3-5 篇)
- [ ] 添加 webhook 通知(不只是 Telegram
- [ ] 统计摄入速率和成功率
---
*End of Note*

View File

@@ -1,263 +1,263 @@
## SYSTEM PROMPT (Set this once at the start of the Gemini session)
```
You are an infographic generation assistant. Your job is to create 6 chalkboard-style
infographic cards that form a complete visual guide.
== GLOBAL STYLE RULES (apply to EVERY card, no exceptions) ==
Background:
- Dark green-black chalkboard: #1C2B1C
- Realistic chalkboard texture with subtle scratches, dust particles, faint eraser smudge marks
- Wooden frame border on all cards (hand-drawn wood grain lines in chalk brown/tan)
- NO gradients, NO perfect geometric shapes, NO photorealistic elements
Chalk Lines & Quality:
- ALL lines must be hand-drawn, imperfect, sketchy — slight wobble and variation
- Lines should look like real white/colored chalk on a blackboard
- NO clean digital vectors, NO sharp vector paths
Color Palette (strict — use ONLY these exact hex values):
- Chalk White: #F5F5F5 (main text, outlines)
- Chalk Yellow: #FFE566 (highlights, emphasis, underlines)
- Chalk Pink: #FF9999 (secondary highlights, icons)
- Chalk Blue: #66B3FF (diagrams, technical elements)
- Chalk Green: #90EE90 (success, nature, positive)
- Chalk Orange: #FFB366 (warnings, energy)
- Frame Brown: #8B6914 (wooden frame, hand-drawn)
Doodles & Decorative Elements:
- Small hand-drawn stars (5-7 points, imperfect)
- Hand-drawn underlines (slightly wavy)
- Hand-drawn arrows (sketchy shaft + arrowhead)
- Hand-drawn circles/ovals around key terms
- Hand-drawn checkmarks
- Scattered chalk dust particles near bottom/sides
Typography:
- All text hand-drawn chalk lettering style
- Imperfect baseline (letters slightly off horizontal)
- Mix of uppercase headers and lowercase body text for authenticity
- Visible chalk texture on letters
== CARD STRUCTURE (identical for all 6 cards) ==
Each card follows this layout:
┌──────────────────────────────────────────┐
│ [WOODEN FRAME BORDER] │
│ ┌────────────────────────────────────┐ │
│ │ CARD TITLE (large, chalk white) │ │
│ │ ~~underlined with accent color~~ │ │
│ ├────────────────────────────────────┤ │
│ │ [SECTION 1] │ [SECTION 2 if any] │ │
│ │ Header color │ │ │
│ │ Bullets │ │ │
│ │ Icon │ │ │
│ ├────────────────────────────────────┤ │
│ │ [Additional sections as needed] │ │
│ │ [Decorative doodles in corners] │ │
│ └────────────────────────────────────┘ │
└──────────────────────────────────────────┘
== CONSISTENCY RULES ==
1. Generate Card 1 first, send it to me
2. For Card 26, EXPLICITLY include this instruction:
"Follow the exact same chalkboard style as the previous card —
same background #1C2B1C, same chalk dust texture, same hand-drawn
line quality, same color hex values (#F5F5F5, #FFE566, #FF9999,
#66B3FF, #90EE90, #FFB366), same wooden frame border, same doodle
elements. Do NOT deviate from this style."
3. Aspect ratio: 16:9 for all cards
4. Each card should visually "belong" to the same set
== HOW TO USE THESE PROMPTS ==
1. Copy the SYSTEM PROMPT above and paste it at the start of your Gemini session
2. Then copy Prompt 1 and send it to Gemini Image Gen (Card 1)
3. Once Card 1 is generated, copy Prompt 2 but FIRST include the STYLE LOCK BLOCK
4. Repeat for all 6 cards, always referencing the previous card's style
5. Review each generated image: if chalk line quality or colors deviate,
regenerate with stronger style enforcement
```
---
## STYLE LOCK BLOCK (Prepend this to Prompts 26)
```
== STYLE LOCK — MANDATORY ==
This card MUST follow the EXACT same chalkboard style as the previously
generated card. Do not deviate.
Checklist — verify these match the previous card BEFORE generating:
□ Background color: #1C2B1C (dark green-black chalkboard)
□ Chalk texture: subtle scratches, dust, eraser smudges
□ Line quality: hand-drawn, imperfect, sketchy wobble — NO perfect vectors
□ Color hex values: #F5F5F5 (white), #FFE566 (yellow), #FF9999 (pink),
#66B3FF (blue), #90EE90 (green), #FFB366 (orange)
□ Frame: wooden border with hand-drawn wood grain
□ Doodles: stars, underlines, arrows, circles — all chalk-drawn
□ Typography: chalk lettering, imperfect baseline, chalk texture on letters
If ANY element does not match, regenerate with corrections.
```
---
## CARD 1 — Saleable & Security
```
== STYLE LOCK — MANDATORY ==
This card MUST follow the EXACT same chalkboard style as the previously
generated card. Do not deviate.
Checklist — verify these match the previous card BEFORE generating:
□ Background color: #1C2B1C (dark green-black chalkboard)
□ Chalk texture: subtle scratches, dust, eraser smudges
□ Line quality: hand-drawn, imperfect, sketchy wobble — NO perfect vectors
□ Color hex values: #F5F5F5 (white), #FFE566 (yellow), #FF9999 (pink),
#66B3FF (blue), #90EE90 (green), #FFB366 (orange)
□ Frame: wooden border with hand-drawn wood grain
□ Doodles: stars, underlines, arrows, circles — all chalk-drawn
□ Typography: chalk lettering, imperfect baseline, chalk texture on letters
If ANY element does not match, regenerate with corrections.
---
INFOGRAPHIC CARD 1: Saleable & Security
Create a single infographic card in chalkboard style with a dark green-black
background (#1C2B1C), realistic chalk dust texture, subtle eraser smudge marks,
and a wooden frame border with hand-drawn wood grain lines.
Card is 16:9 aspect ratio. All elements must look hand-drawn with real chalk —
imperfect sketchy lines, slight wobble, no clean vectors.
TITLE SECTION:
- "Saleable & Security" in large hand-drawn chalk white (#F5F5F5) uppercase
lettering, centered at top
- Double underline in chalk yellow (#FFE566), slightly wavy hand-drawn lines
- Small hand-drawn star doodles on each side of the title
TWO-COLUMN LAYOUT below title:
LEFT COLUMN — "Saleable" (header in chalk pink #FF9999, hand-drawn rectangle bar):
• Complete product definition in Control Tower
• SKUs clearly defined
• License generation strategy complete
Bullet markers: small chalk pink circles
Icon: hand-drawn chalk sketch of a product box with a small price tag label,
in chalk yellow on white outline
RIGHT COLUMN — "Security" (header in chalk blue #66B3FF, hand-drawn rectangle bar):
• Application self-defensibility capability
• WAF rules to protect cloud apps
• Defend against incorrect usage (accidental & purposeful)
Bullet markers: small chalk blue circles
Icon: hand-drawn chalk shield with a simple checkmark inside, outline in
chalk white, fill in semi-transparent chalk blue
FOOTER DECORATION:
- A hand-drawn chalk dividing line across the card width
- Two small doodle checkmarks at bottom right in chalk green (#90EE90)
- Scattered chalk dust particles along the bottom edge
NO gradients, NO sharp geometric shapes, NO flat digital icons.
```
## ![IMG-20260418172056603.png](app://77d327b60ef00764e25eb25542b49a223773/D:/Workspace/nexus/Asset/Attachment/raw/AI/%E5%A6%82%E4%BD%95%E8%AE%A9AI%E7%94%9F%E6%88%90%E9%A3%8E%E6%A0%BC%E4%B8%80%E8%87%B4%E7%9A%84%E5%9B%BE%E7%89%87/IMG-20260418172056603.png?1776503642747)
## CARD 2 — Cloud Deployment & Configuration
```
== STYLE LOCK — MANDATORY ==
This card MUST follow the EXACT same chalkboard style as the previously
generated card. Do not deviate.
Checklist — verify these match the previous card BEFORE generating:
□ Background color: #1C2B1C (dark green-black chalkboard)
□ Chalk texture: subtle scratches, dust, eraser smudges
□ Line quality: hand-drawn, imperfect, sketchy wobble — NO perfect vectors
□ Color hex values: #F5F5F5 (white), #FFE566 (yellow), #FF9999 (pink),
#66B3FF (blue), #90EE90 (green), #FFB366 (orange)
□ Frame: wooden border with hand-drawn wood grain
□ Doodles: stars, underlines, arrows, circles — all chalk-drawn
□ Typography: chalk lettering, imperfect baseline, chalk texture on letters
If ANY element does not match, regenerate with corrections.
---
INFOGRAPHIC CARD 2: Cloud Deployment & Configuration
Create a single infographic card in chalkboard style with a dark green-black
background (#1A1A1A), realistic chalk dust texture, subtle eraser smudge marks,
and a wooden frame border with hand-drawn wood grain lines.
Card is 16:9 aspect ratio. All elements must look hand-drawn with real chalk —
imperfect sketchy lines, slight wobble, no clean vectors.
TITLE SECTION:
- "Cloud Deployment & Configuration" in large hand-drawn chalk white (#F5F5F5)
uppercase lettering, centered at top
- Underline in chalk green (#90EE90), hand-drawn wavy line
- Small star doodle on left side of title
MAIN CONTENT AREA:
Header bar: "Deployment Requirements" in chalk blue (#66B3FF) hand-drawn
rectangle
Bullet list (chalk white text, chalk yellow bullet markers):
✔ Fully automated cloud platform deployment
✔ Web / API enabled configuration management
✔ All feature & functional configs through API interface
✔ Tenant capability enablement
Sub-section header: "Configuration Management" in chalk pink (#FF9999)
HAND-DRAWN FLOWCHART (center of card):
Three chalk blue boxes connected by sketchy chalk arrows:
[Cloud Platform] --arrow--> [API Gateway] --arrow--> [Tenant Config]
Each box: hand-drawn rectangle with slightly wavy edges, white #F5F5F5 outline
Text inside boxes: chalk white
Arrows: hand-drawn with slight wobble, chalk blue
ADDITIONAL ELEMENT:
Hand-drawn stick figure engineer icon (simple, chalk white) on the right side
holding a small chalk-drawn tablet/device
CORNER DOODLES:
- Top right: small hand-drawn cloud shape labeled "SaaS" in chalk orange (#FFB366)
- Bottom left: chalk pink circle with "API" text inside
- Scattered chalk dust particles near the wooden frame
NO gradients, NO sharp geometry, NO digital-looking elements.
```
## How to Use This File
```
SEQUENCE:
1. Gemini session start → paste the SYSTEM PROMPT
2. Send CARD 1 prompt → receive Card 1 image
3. Paste CARD 2 prompt (it includes STYLE LOCK BLOCK) → receive Card 2
4. Repeat for Cards 36
VERIFICATION after each card:
- Does background look like #1C2B1C dark green-black chalkboard? ✓/✗
- Do all lines look hand-drawn/sketchy? ✓/✗
- Are colors using the exact hex values? ✓/✗
- Is there a wooden frame border? ✓/✗
- Are doodles (stars, underlines, arrows) hand-drawn? ✓/✗
- Does it match the previous card visually? ✓/✗
If any check fails → regenerate with stronger style enforcement.
## SYSTEM PROMPT (Set this once at the start of the Gemini session)
```
You are an infographic generation assistant. Your job is to create 6 chalkboard-style
infographic cards that form a complete visual guide.
== GLOBAL STYLE RULES (apply to EVERY card, no exceptions) ==
Background:
- Dark green-black chalkboard: #1C2B1C
- Realistic chalkboard texture with subtle scratches, dust particles, faint eraser smudge marks
- Wooden frame border on all cards (hand-drawn wood grain lines in chalk brown/tan)
- NO gradients, NO perfect geometric shapes, NO photorealistic elements
Chalk Lines & Quality:
- ALL lines must be hand-drawn, imperfect, sketchy — slight wobble and variation
- Lines should look like real white/colored chalk on a blackboard
- NO clean digital vectors, NO sharp vector paths
Color Palette (strict — use ONLY these exact hex values):
- Chalk White: #F5F5F5 (main text, outlines)
- Chalk Yellow: #FFE566 (highlights, emphasis, underlines)
- Chalk Pink: #FF9999 (secondary highlights, icons)
- Chalk Blue: #66B3FF (diagrams, technical elements)
- Chalk Green: #90EE90 (success, nature, positive)
- Chalk Orange: #FFB366 (warnings, energy)
- Frame Brown: #8B6914 (wooden frame, hand-drawn)
Doodles & Decorative Elements:
- Small hand-drawn stars (5-7 points, imperfect)
- Hand-drawn underlines (slightly wavy)
- Hand-drawn arrows (sketchy shaft + arrowhead)
- Hand-drawn circles/ovals around key terms
- Hand-drawn checkmarks
- Scattered chalk dust particles near bottom/sides
Typography:
- All text hand-drawn chalk lettering style
- Imperfect baseline (letters slightly off horizontal)
- Mix of uppercase headers and lowercase body text for authenticity
- Visible chalk texture on letters
== CARD STRUCTURE (identical for all 6 cards) ==
Each card follows this layout:
┌──────────────────────────────────────────┐
│ [WOODEN FRAME BORDER] │
│ ┌────────────────────────────────────┐ │
│ │ CARD TITLE (large, chalk white) │ │
│ │ ~~underlined with accent color~~ │ │
│ ├────────────────────────────────────┤ │
│ │ [SECTION 1] │ [SECTION 2 if any] │ │
│ │ Header color │ │ │
│ │ Bullets │ │ │
│ │ Icon │ │ │
│ ├────────────────────────────────────┤ │
│ │ [Additional sections as needed] │ │
│ │ [Decorative doodles in corners] │ │
│ └────────────────────────────────────┘ │
└──────────────────────────────────────────┘
== CONSISTENCY RULES ==
1. Generate Card 1 first, send it to me
2. For Card 26, EXPLICITLY include this instruction:
"Follow the exact same chalkboard style as the previous card —
same background #1C2B1C, same chalk dust texture, same hand-drawn
line quality, same color hex values (#F5F5F5, #FFE566, #FF9999,
#66B3FF, #90EE90, #FFB366), same wooden frame border, same doodle
elements. Do NOT deviate from this style."
3. Aspect ratio: 16:9 for all cards
4. Each card should visually "belong" to the same set
== HOW TO USE THESE PROMPTS ==
1. Copy the SYSTEM PROMPT above and paste it at the start of your Gemini session
2. Then copy Prompt 1 and send it to Gemini Image Gen (Card 1)
3. Once Card 1 is generated, copy Prompt 2 but FIRST include the STYLE LOCK BLOCK
4. Repeat for all 6 cards, always referencing the previous card's style
5. Review each generated image: if chalk line quality or colors deviate,
regenerate with stronger style enforcement
```
---
## STYLE LOCK BLOCK (Prepend this to Prompts 26)
```
== STYLE LOCK — MANDATORY ==
This card MUST follow the EXACT same chalkboard style as the previously
generated card. Do not deviate.
Checklist — verify these match the previous card BEFORE generating:
□ Background color: #1C2B1C (dark green-black chalkboard)
□ Chalk texture: subtle scratches, dust, eraser smudges
□ Line quality: hand-drawn, imperfect, sketchy wobble — NO perfect vectors
□ Color hex values: #F5F5F5 (white), #FFE566 (yellow), #FF9999 (pink),
#66B3FF (blue), #90EE90 (green), #FFB366 (orange)
□ Frame: wooden border with hand-drawn wood grain
□ Doodles: stars, underlines, arrows, circles — all chalk-drawn
□ Typography: chalk lettering, imperfect baseline, chalk texture on letters
If ANY element does not match, regenerate with corrections.
```
---
## CARD 1 — Saleable & Security
```
== STYLE LOCK — MANDATORY ==
This card MUST follow the EXACT same chalkboard style as the previously
generated card. Do not deviate.
Checklist — verify these match the previous card BEFORE generating:
□ Background color: #1C2B1C (dark green-black chalkboard)
□ Chalk texture: subtle scratches, dust, eraser smudges
□ Line quality: hand-drawn, imperfect, sketchy wobble — NO perfect vectors
□ Color hex values: #F5F5F5 (white), #FFE566 (yellow), #FF9999 (pink),
#66B3FF (blue), #90EE90 (green), #FFB366 (orange)
□ Frame: wooden border with hand-drawn wood grain
□ Doodles: stars, underlines, arrows, circles — all chalk-drawn
□ Typography: chalk lettering, imperfect baseline, chalk texture on letters
If ANY element does not match, regenerate with corrections.
---
INFOGRAPHIC CARD 1: Saleable & Security
Create a single infographic card in chalkboard style with a dark green-black
background (#1C2B1C), realistic chalk dust texture, subtle eraser smudge marks,
and a wooden frame border with hand-drawn wood grain lines.
Card is 16:9 aspect ratio. All elements must look hand-drawn with real chalk —
imperfect sketchy lines, slight wobble, no clean vectors.
TITLE SECTION:
- "Saleable & Security" in large hand-drawn chalk white (#F5F5F5) uppercase
lettering, centered at top
- Double underline in chalk yellow (#FFE566), slightly wavy hand-drawn lines
- Small hand-drawn star doodles on each side of the title
TWO-COLUMN LAYOUT below title:
LEFT COLUMN — "Saleable" (header in chalk pink #FF9999, hand-drawn rectangle bar):
• Complete product definition in Control Tower
• SKUs clearly defined
• License generation strategy complete
Bullet markers: small chalk pink circles
Icon: hand-drawn chalk sketch of a product box with a small price tag label,
in chalk yellow on white outline
RIGHT COLUMN — "Security" (header in chalk blue #66B3FF, hand-drawn rectangle bar):
• Application self-defensibility capability
• WAF rules to protect cloud apps
• Defend against incorrect usage (accidental & purposeful)
Bullet markers: small chalk blue circles
Icon: hand-drawn chalk shield with a simple checkmark inside, outline in
chalk white, fill in semi-transparent chalk blue
FOOTER DECORATION:
- A hand-drawn chalk dividing line across the card width
- Two small doodle checkmarks at bottom right in chalk green (#90EE90)
- Scattered chalk dust particles along the bottom edge
NO gradients, NO sharp geometric shapes, NO flat digital icons.
```
## ![IMG-20260418172056603.png](app://77d327b60ef00764e25eb25542b49a223773/D:/Workspace/nexus/Asset/Attachment/raw/AI/%E5%A6%82%E4%BD%95%E8%AE%A9AI%E7%94%9F%E6%88%90%E9%A3%8E%E6%A0%BC%E4%B8%80%E8%87%B4%E7%9A%84%E5%9B%BE%E7%89%87/IMG-20260418172056603.png?1776503642747)
## CARD 2 — Cloud Deployment & Configuration
```
== STYLE LOCK — MANDATORY ==
This card MUST follow the EXACT same chalkboard style as the previously
generated card. Do not deviate.
Checklist — verify these match the previous card BEFORE generating:
□ Background color: #1C2B1C (dark green-black chalkboard)
□ Chalk texture: subtle scratches, dust, eraser smudges
□ Line quality: hand-drawn, imperfect, sketchy wobble — NO perfect vectors
□ Color hex values: #F5F5F5 (white), #FFE566 (yellow), #FF9999 (pink),
#66B3FF (blue), #90EE90 (green), #FFB366 (orange)
□ Frame: wooden border with hand-drawn wood grain
□ Doodles: stars, underlines, arrows, circles — all chalk-drawn
□ Typography: chalk lettering, imperfect baseline, chalk texture on letters
If ANY element does not match, regenerate with corrections.
---
INFOGRAPHIC CARD 2: Cloud Deployment & Configuration
Create a single infographic card in chalkboard style with a dark green-black
background (#1A1A1A), realistic chalk dust texture, subtle eraser smudge marks,
and a wooden frame border with hand-drawn wood grain lines.
Card is 16:9 aspect ratio. All elements must look hand-drawn with real chalk —
imperfect sketchy lines, slight wobble, no clean vectors.
TITLE SECTION:
- "Cloud Deployment & Configuration" in large hand-drawn chalk white (#F5F5F5)
uppercase lettering, centered at top
- Underline in chalk green (#90EE90), hand-drawn wavy line
- Small star doodle on left side of title
MAIN CONTENT AREA:
Header bar: "Deployment Requirements" in chalk blue (#66B3FF) hand-drawn
rectangle
Bullet list (chalk white text, chalk yellow bullet markers):
✔ Fully automated cloud platform deployment
✔ Web / API enabled configuration management
✔ All feature & functional configs through API interface
✔ Tenant capability enablement
Sub-section header: "Configuration Management" in chalk pink (#FF9999)
HAND-DRAWN FLOWCHART (center of card):
Three chalk blue boxes connected by sketchy chalk arrows:
[Cloud Platform] --arrow--> [API Gateway] --arrow--> [Tenant Config]
Each box: hand-drawn rectangle with slightly wavy edges, white #F5F5F5 outline
Text inside boxes: chalk white
Arrows: hand-drawn with slight wobble, chalk blue
ADDITIONAL ELEMENT:
Hand-drawn stick figure engineer icon (simple, chalk white) on the right side
holding a small chalk-drawn tablet/device
CORNER DOODLES:
- Top right: small hand-drawn cloud shape labeled "SaaS" in chalk orange (#FFB366)
- Bottom left: chalk pink circle with "API" text inside
- Scattered chalk dust particles near the wooden frame
NO gradients, NO sharp geometry, NO digital-looking elements.
```
## How to Use This File
```
SEQUENCE:
1. Gemini session start → paste the SYSTEM PROMPT
2. Send CARD 1 prompt → receive Card 1 image
3. Paste CARD 2 prompt (it includes STYLE LOCK BLOCK) → receive Card 2
4. Repeat for Cards 36
VERIFICATION after each card:
- Does background look like #1C2B1C dark green-black chalkboard? ✓/✗
- Do all lines look hand-drawn/sketchy? ✓/✗
- Are colors using the exact hex values? ✓/✗
- Is there a wooden frame border? ✓/✗
- Are doodles (stars, underlines, arrows) hand-drawn? ✓/✗
- Does it match the previous card visually? ✓/✗
If any check fails → regenerate with stronger style enforcement.
```

View File

@@ -1,51 +1,51 @@
---
title: 养龙虾5天血泪史我的AI Agent为什么总失忆OpenClaw 记忆调试全记录
type: metaphor
palette: warm
rendering: digital
text: title-subtitle
mood: balanced
---
请为这篇文章生成封面图。
## 视觉概念
以"金鱼记忆"为核心隐喻:一条卡通金鱼游在透明的水缸中,金鱼的大脑位置是一个空白的问号框架,周围漂浮着记忆碎片(文字泡泡、对话气泡、代码片段),碎片正在逐渐消散到缸外。底部有简单的齿轮和调试工具元素,暗示"调试"过程。
## 主视觉元素
- 金鱼:卡通风格,圆形身体,大眼睛,尾巴呈波浪形游动姿态
- 空白大脑:空心问号形状,位于金鱼头部位置
- 记忆碎片5-6个椭圆形气泡包含模糊的文字轮廓、对话符号、代码片段
- 水缸:简单几何圆形边框,内部有微妙的涟漪效果
- 调试工具:底部角落有小型齿轮、螺丝刀、代码括号图标
## 配色方案warm palette
- 主色:温暖橙色 #F5A623(用于金鱼身体)
- 辅色:柔和珊瑚色 #FF8C74(用于记忆碎片)
- 强调色:深琥珀色 #C47A2B(用于鱼鳍和调试工具)
- 背景:米白色 #FFF8F0 到浅杏色 #FFE8D6 渐变
- 文字色:深棕色 #4A3728
## 渲染风格digital
- 干净的矢量线条,精确边缘
- 平滑渐变,无粗糙纹理
- 轻微阴影创造层次感
- 像专业UI插图一样现代简洁
## 文字布局
- 标题:"养龙虾5天血泪史" 使用大号粗体字(金鱼上方右侧)
- 副标题:"我的AI Agent为什么总失忆" 使用较小字号(标题下方)
- 字体:现代无衬线体,清晰易读
- 文字颜色:深棕色 #4A3728
## 氛围balanced
- 中等对比度
- 正常饱和度
- 视觉重量平衡
---
title: 养龙虾5天血泪史我的AI Agent为什么总失忆OpenClaw 记忆调试全记录
type: metaphor
palette: warm
rendering: digital
text: title-subtitle
mood: balanced
---
请为这篇文章生成封面图。
## 视觉概念
以"金鱼记忆"为核心隐喻:一条卡通金鱼游在透明的水缸中,金鱼的大脑位置是一个空白的问号框架,周围漂浮着记忆碎片(文字泡泡、对话气泡、代码片段),碎片正在逐渐消散到缸外。底部有简单的齿轮和调试工具元素,暗示"调试"过程。
## 主视觉元素
- 金鱼:卡通风格,圆形身体,大眼睛,尾巴呈波浪形游动姿态
- 空白大脑:空心问号形状,位于金鱼头部位置
- 记忆碎片5-6个椭圆形气泡包含模糊的文字轮廓、对话符号、代码片段
- 水缸:简单几何圆形边框,内部有微妙的涟漪效果
- 调试工具:底部角落有小型齿轮、螺丝刀、代码括号图标
## 配色方案warm palette
- 主色:温暖橙色 #F5A623(用于金鱼身体)
- 辅色:柔和珊瑚色 #FF8C74(用于记忆碎片)
- 强调色:深琥珀色 #C47A2B(用于鱼鳍和调试工具)
- 背景:米白色 #FFF8F0 到浅杏色 #FFE8D6 渐变
- 文字色:深棕色 #4A3728
## 渲染风格digital
- 干净的矢量线条,精确边缘
- 平滑渐变,无粗糙纹理
- 轻微阴影创造层次感
- 像专业UI插图一样现代简洁
## 文字布局
- 标题:"养龙虾5天血泪史" 使用大号粗体字(金鱼上方右侧)
- 副标题:"我的AI Agent为什么总失忆" 使用较小字号(标题下方)
- 字体:现代无衬线体,清晰易读
- 文字颜色:深棕色 #4A3728
## 氛围balanced
- 中等对比度
- 正常饱和度
- 视觉重量平衡
- 温暖友好但不强烈

View File

@@ -1,22 +1,22 @@
## 信息图任务指令
请用**claude-code-infographic-build** 这个技能为`Hermes/xingzhi/Hermes自定义技能说明` 这篇笔记生成信息图。
- 语言:英文
- 布局circular-flow
- 风格chalkboard
- 比例16:9
## 封面图任务指令
请用**claude-code-executor**技能启动Claude Code
使用 **baoyu-cover-image** 技能为以下内容生成封面图提示词:
1. 读取 Hermes/xingzhi/wiki-sync-setup-2026-04-16.md 这篇笔记
2. 用**baoyu-cover-image** 生成一张封面图
- **类型 (Type)**`hero``conceptual``typography``metaphor``scene``minimal`
- **配色 (Palette)**`warm``elegant``cool``dark``earth``vivid``pastel``mono``retro`
- **渲染 (Rendering)**`flat-vector``hand-drawn``painterly``digital``pixel``chalk`
- **文字 (Text)**`none``title-only`(默认)、`title-subtitle``text-rich`
- **氛围 (Mood)**`subtle``balanced`(默认)、`bold`
3. 不要生成图片,只输出提示词
## 信息图任务指令
请用**claude-code-infographic-build** 这个技能为`Hermes/xingzhi/Hermes自定义技能说明` 这篇笔记生成信息图。
- 语言:英文
- 布局circular-flow
- 风格chalkboard
- 比例16:9
## 封面图任务指令
请用**claude-code-executor**技能启动Claude Code
使用 **baoyu-cover-image** 技能为以下内容生成封面图提示词:
1. 读取 Hermes/xingzhi/wiki-sync-setup-2026-04-16.md 这篇笔记
2. 用**baoyu-cover-image** 生成一张封面图
- **类型 (Type)**`hero``conceptual``typography``metaphor``scene``minimal`
- **配色 (Palette)**`warm``elegant``cool``dark``earth``vivid``pastel``mono``retro`
- **渲染 (Rendering)**`flat-vector``hand-drawn``painterly``digital``pixel``chalk`
- **文字 (Text)**`none``title-only`(默认)、`title-subtitle``text-rich`
- **氛围 (Mood)**`subtle``balanced`(默认)、`bold`
3. 不要生成图片,只输出提示词
4. 输出提示词到 Hermes/xingzhi/ 新建一个笔记

View File

@@ -1,127 +1,127 @@
**副标题**:如何把每一份笔记,通过 llm-wiki-sync 自动分析与提炼成结构化的页面、实体与概念,以便长期检索与复用。
---
## 前言与来源说明
本方案借鉴并整合了几条重要线索:
- Andrej Karpathy 对“LLM Wiki”理念的阐述将知识以可被大模型直接消费的结构化方式保存https://gist.github.com/karpathy/442a6bf555914893e9891c11519de94f
- 我们的实现基于开源项目 SamurAI 的 llm-wiki-agenthttps://github.com/SamurAIGPT/llm-wiki-agent在其基础上扩展了企业化的 ingest 流程、Cron 调度与 Hermes skillllm-wiki-sync
- 最终静态化展示使用 Quartzhttps://github.com/jackyzha0/quartz把生成的 wiki 内容导出为可浏览、可分享的静态站点。
下面重点介绍 llm-wiki-sync 如何把一篇笔记做结构化分析与提炼并用仓库中的实例wiki/sources/RTO-vs-RPO-Key-Differences-for-Modern-Disaster-Recovery作为逐项说明。
---
## llm-wiki-sync 的目标与核心能力
核心目标把原始文档raw/)自动转成结构化的 wiki source 页面抽取关键要素Summary、Key Claims、Key Quotes、Key Concepts、Key Entities、Connections并记录 ingest 日志、差异与审计信息,供后续检索、合成与内容再生产使用。
关键能力:
- 文本解析与语义压缩:把长文本压缩为 24 句高密度 summary。
- 声明抽取claim extraction识别文中明确的结论与可验证断言。
- 实体与概念抽取NER + linking识别人名/公司/工具/概念,并把它们标准化为 wiki 实体页([[Name]])。
- 关系发现connections把句子级别的语义关系转成图边A → depends_on → B
- 模板化输出固定页面头frontmatter+ 标准段落Summary / Key Claims / Quotes / Concepts / Entities / Connections / Contradictions
- 审计与可回滚:每次 ingest 都写入 wiki/log.md并可通过 git/checkpoint 回滚变更。
实现技术栈要点Hermesskill 调用、Claude Code / agent可选、llm-wiki-agent 基础脚本、以及最终的静态化工具 Quartz。
---
## 从笔记到 Source Page
仓库中的源页面wiki/sources/RTO-vs-RPO-Key-Differences-for-Modern-Disaster-Recovery.md
下面逐项展示 llm-wiki-sync 针对该文档所做的提取结果(摘自生成的 source 页面):
1) SummarySummary
- 核心主题RTO恢复时间目标与 RPO恢复点目标的定义、区别及在现代持续交付中的应用
- 问题域:灾难恢复规划、发布风险管控
- 方法/机制:通过 Feature Flag 实现秒级 RTO 和低 RPO
- 结论/价值预防优于恢复Feature Flag 将部署事故从灾难转为非事件
说明Summary 由模型将整篇文章的主旨、问题背景、关键方法与结论压缩为 24 条,便于快速检索与索引。
2) Key Claims断言提取
- RTO 衡量系统恢复速度:允许的最大停机时间
- RPO 衡量数据保护:可接受的最大数据丢失量
- 传统灾备聚焦硬件故障,现代风险更多来自代码变更(部署 bug、数据库迁移、AI 模型更新等)
- Feature Flag 将 RTO 从小时级降至秒级,同时保护 RPO
- 应用分层策略Critical / Important / Nice-to-have对应不同的 RTO/RPO 指标
说明断言提取用于建立事实层fact layer后续可自动化做一致性检查与冲突检测Contradictions 段)。
3) Key Quotes关键引用
- “RTO is about speed: how fast you get back online. RPO is about data: how much you can afford to lose.”
- “Deploy != Release. Feature flags change this. You can deploy code to production without releasing it to users.”
说明:保留可引用原句,便于在后续合成(如写作、演讲稿)中引用来源。
4) Key Concepts概念抽取
- RTORecovery Time Objective
- RPORecovery Point Objective
- Feature Flag特性开关
- Kill Switch紧急关闭机制
- 渐进式发布Gradual Rollout
说明:概念会被标准化为 wiki 的 concept 页面wiki/concepts/),用于聚合所有提到该概念的 source 页面。
5) Key Entities实体抽取
- LaunchDarklyFeature Flag 管理平台)
- HP示例企业
- Christian Dior示例企业 — 文档中作为案例提及)
说明:实体会被规范化为 wiki/entities/ 下的页面,并且 source 页面会在 sources 列表保留原始链接与引用。
6) Connections关系构建
- RTO ← depends_on ← Feature Flag
- RPO ← depends_on ← Feature Flag
- LaunchDarkly → provides → Feature Flag
- Feature Flag ← enables ← 渐进式发布
说明Connections 用于图谱构建graph/graph.json后续能在可视化页面graph.html展示节点与边。
7) Contradictions冲突检测
- 当前文档无明显与现有 wiki 冲突的声明若检测到冲突llm-wiki-sync 会把冲突条目列出并标注来源,供人工审查。
![[IMG-20260420160439552.png|872]]
![[IMG-20260420160439600.png]]
---
## llm-wiki-sync 的典型运行步骤(工程视角)
1. 读取 raw/<path>,解析 frontmatter/元数据(若缺失则询回填)。
2. 检索 wiki/index.md 与 overview.md获取上下文避免重复 ingest
3. 用 LLM 生成 Source PageSummary / Key Claims / Quotes / Concepts / Entities / Connections / Contradictions
4. 将结果写入 wiki/sources/<slug>.md并更新 wiki/index.md、wiki/overview.md如有新实体/概念也生成对应页面草稿。
5. 记录日志:在 wiki/log.md 追加 ingest 记录(时间、文件、摘要、状态)。
6. 可选:触发 graph 重建(增量或全量),并把 graph/graph.json、graph.html 更新到站点。
7. 通知:完成后通过 Hermes 的通知机制deliver=origin 或 Telegram 指定 chat告知负责人。
并发与配额注意:单文件 ingest 优先,批量操作分批(每批 2~3 篇);对接外部 agent/Claude Code 时避免并发超配额。
审计与回滚:每次 ingest 前执行 git branch 或 checkpoint如需回滚可用 git revert 或恢复 checkpoint。
---
## 总结与扩展
- llm-wiki-sync 把 Karpathy 关于 LLM Wiki 的理念落地为可执行的工程流程:把知识以结构化表征保存,使得大模型既是“读者”也是“执行者”。
- 在 Obsidian 中可以直接通过关系图graph view查看笔记间的关联在 llm-wiki-agent 中可以通过 wiki-graph 构建并在 graph.html / graph/graph.json 中可视化展示。
- 我们的实现基于 SamurAI 的 llm-wiki-agent并在其上加入了企业级的同步、审计与 Hermes skill 封装,最终通过 Quartz 静态站把生成的 wiki 内容对外展示与分享。
---
## Infographic Asset
![llm-wiki-sync Circular Flow Infographic](IMG-20260420160439662.png)
**Infographic**: The Knowledge Pipeline Cycle — circular flow showing how llm-wiki-sync transforms scattered notes into a reusable knowledge base.
- Layout: circular-flow (7-stage cycle)
- Style: chalkboard (dark background, hand-drawn chalk accents)
- Aspect ratio: 16:9
- Prompt file: `infographic/llm-wiki-sync-circular-flow/prompts/infographic.md`
- Image: `llm-wiki-sync-circular-flow-infographic.png`
**副标题**:如何把每一份笔记,通过 llm-wiki-sync 自动分析与提炼成结构化的页面、实体与概念,以便长期检索与复用。
---
## 前言与来源说明
本方案借鉴并整合了几条重要线索:
- Andrej Karpathy 对“LLM Wiki”理念的阐述将知识以可被大模型直接消费的结构化方式保存https://gist.github.com/karpathy/442a6bf555914893e9891c11519de94f
- 我们的实现基于开源项目 SamurAI 的 llm-wiki-agenthttps://github.com/SamurAIGPT/llm-wiki-agent在其基础上扩展了企业化的 ingest 流程、Cron 调度与 Hermes skillllm-wiki-sync
- 最终静态化展示使用 Quartzhttps://github.com/jackyzha0/quartz把生成的 wiki 内容导出为可浏览、可分享的静态站点。
下面重点介绍 llm-wiki-sync 如何把一篇笔记做结构化分析与提炼并用仓库中的实例wiki/sources/RTO-vs-RPO-Key-Differences-for-Modern-Disaster-Recovery作为逐项说明。
---
## llm-wiki-sync 的目标与核心能力
核心目标把原始文档raw/)自动转成结构化的 wiki source 页面抽取关键要素Summary、Key Claims、Key Quotes、Key Concepts、Key Entities、Connections并记录 ingest 日志、差异与审计信息,供后续检索、合成与内容再生产使用。
关键能力:
- 文本解析与语义压缩:把长文本压缩为 24 句高密度 summary。
- 声明抽取claim extraction识别文中明确的结论与可验证断言。
- 实体与概念抽取NER + linking识别人名/公司/工具/概念,并把它们标准化为 wiki 实体页([[Name]])。
- 关系发现connections把句子级别的语义关系转成图边A → depends_on → B
- 模板化输出固定页面头frontmatter+ 标准段落Summary / Key Claims / Quotes / Concepts / Entities / Connections / Contradictions
- 审计与可回滚:每次 ingest 都写入 wiki/log.md并可通过 git/checkpoint 回滚变更。
实现技术栈要点Hermesskill 调用、Claude Code / agent可选、llm-wiki-agent 基础脚本、以及最终的静态化工具 Quartz。
---
## 从笔记到 Source Page
仓库中的源页面wiki/sources/RTO-vs-RPO-Key-Differences-for-Modern-Disaster-Recovery.md
下面逐项展示 llm-wiki-sync 针对该文档所做的提取结果(摘自生成的 source 页面):
1) SummarySummary
- 核心主题RTO恢复时间目标与 RPO恢复点目标的定义、区别及在现代持续交付中的应用
- 问题域:灾难恢复规划、发布风险管控
- 方法/机制:通过 Feature Flag 实现秒级 RTO 和低 RPO
- 结论/价值预防优于恢复Feature Flag 将部署事故从灾难转为非事件
说明Summary 由模型将整篇文章的主旨、问题背景、关键方法与结论压缩为 24 条,便于快速检索与索引。
2) Key Claims断言提取
- RTO 衡量系统恢复速度:允许的最大停机时间
- RPO 衡量数据保护:可接受的最大数据丢失量
- 传统灾备聚焦硬件故障,现代风险更多来自代码变更(部署 bug、数据库迁移、AI 模型更新等)
- Feature Flag 将 RTO 从小时级降至秒级,同时保护 RPO
- 应用分层策略Critical / Important / Nice-to-have对应不同的 RTO/RPO 指标
说明断言提取用于建立事实层fact layer后续可自动化做一致性检查与冲突检测Contradictions 段)。
3) Key Quotes关键引用
- “RTO is about speed: how fast you get back online. RPO is about data: how much you can afford to lose.”
- “Deploy != Release. Feature flags change this. You can deploy code to production without releasing it to users.”
说明:保留可引用原句,便于在后续合成(如写作、演讲稿)中引用来源。
4) Key Concepts概念抽取
- RTORecovery Time Objective
- RPORecovery Point Objective
- Feature Flag特性开关
- Kill Switch紧急关闭机制
- 渐进式发布Gradual Rollout
说明:概念会被标准化为 wiki 的 concept 页面wiki/concepts/),用于聚合所有提到该概念的 source 页面。
5) Key Entities实体抽取
- LaunchDarklyFeature Flag 管理平台)
- HP示例企业
- Christian Dior示例企业 — 文档中作为案例提及)
说明:实体会被规范化为 wiki/entities/ 下的页面,并且 source 页面会在 sources 列表保留原始链接与引用。
6) Connections关系构建
- RTO ← depends_on ← Feature Flag
- RPO ← depends_on ← Feature Flag
- LaunchDarkly → provides → Feature Flag
- Feature Flag ← enables ← 渐进式发布
说明Connections 用于图谱构建graph/graph.json后续能在可视化页面graph.html展示节点与边。
7) Contradictions冲突检测
- 当前文档无明显与现有 wiki 冲突的声明若检测到冲突llm-wiki-sync 会把冲突条目列出并标注来源,供人工审查。
![[IMG-20260420160439552.png|872]]
![[IMG-20260420160439600.png]]
---
## llm-wiki-sync 的典型运行步骤(工程视角)
1. 读取 raw/<path>,解析 frontmatter/元数据(若缺失则询回填)。
2. 检索 wiki/index.md 与 overview.md获取上下文避免重复 ingest
3. 用 LLM 生成 Source PageSummary / Key Claims / Quotes / Concepts / Entities / Connections / Contradictions
4. 将结果写入 wiki/sources/<slug>.md并更新 wiki/index.md、wiki/overview.md如有新实体/概念也生成对应页面草稿。
5. 记录日志:在 wiki/log.md 追加 ingest 记录(时间、文件、摘要、状态)。
6. 可选:触发 graph 重建(增量或全量),并把 graph/graph.json、graph.html 更新到站点。
7. 通知:完成后通过 Hermes 的通知机制deliver=origin 或 Telegram 指定 chat告知负责人。
并发与配额注意:单文件 ingest 优先,批量操作分批(每批 2~3 篇);对接外部 agent/Claude Code 时避免并发超配额。
审计与回滚:每次 ingest 前执行 git branch 或 checkpoint如需回滚可用 git revert 或恢复 checkpoint。
---
## 总结与扩展
- llm-wiki-sync 把 Karpathy 关于 LLM Wiki 的理念落地为可执行的工程流程:把知识以结构化表征保存,使得大模型既是“读者”也是“执行者”。
- 在 Obsidian 中可以直接通过关系图graph view查看笔记间的关联在 llm-wiki-agent 中可以通过 wiki-graph 构建并在 graph.html / graph/graph.json 中可视化展示。
- 我们的实现基于 SamurAI 的 llm-wiki-agent并在其上加入了企业级的同步、审计与 Hermes skill 封装,最终通过 Quartz 静态站把生成的 wiki 内容对外展示与分享。
---
## Infographic Asset
![llm-wiki-sync Circular Flow Infographic](IMG-20260420160439662.png)
**Infographic**: The Knowledge Pipeline Cycle — circular flow showing how llm-wiki-sync transforms scattered notes into a reusable knowledge base.
- Layout: circular-flow (7-stage cycle)
- Style: chalkboard (dark background, hand-drawn chalk accents)
- Aspect ratio: 16:9
- Prompt file: `infographic/llm-wiki-sync-circular-flow/prompts/infographic.md`
- Image: `llm-wiki-sync-circular-flow-infographic.png`

File diff suppressed because one or more lines are too long

View File

@@ -1,222 +1,222 @@
---
title: django-tenants 完整配置指南
created: 2026-04-21
tags: [django, django-tenants, postgresql, saas, multi-tenant]
category: 技术笔记
---
# django-tenants 完整配置指南
## 一、安装依赖
pip install django-tenants psycopg2-binary django-jazzmin
## 二、项目目录结构
myproject/
├── config/
│ ├── settings/
│ │ ├── base.py
│ │ ├── development.py
│ │ └── production.py
│ ├── urls.py
│ └── wsgi.py
├── apps/
│ ├── tenants/
│ ├── subscription/
│ ├── accounts/
│ ├── listings/
│ ├── clients/
│ └── showings/
├── manage.py
└── requirements.txt
## 三、核心 Model租户与域名
- Company 继承 TenantMixin
- Domain 继承 DomainMixin
- 每一个中介公司 = 一个租户 = 一个独立 PostgreSQL Schema
- 每个公司可绑定多个域名/子域名
## 四、Settings 完整配置
关键点:
- SHARED_APPS 放公共 Schema 应用
- TENANT_APPS 放租户私有应用
- TENANT_MODEL = "tenants.Company"
- TENANT_DOMAIN_MODEL = "tenants.Domain"
- DATABASES 使用 django_tenants.postgresql_backend
- DATABASE_ROUTERS 使用 TenantSyncRouter
- TenantMainMiddleware 必须第一个
- ROOT_URLCONF / PUBLIC_SCHEMA_URLCONF 分离
- AUTH_USER_MODEL = accounts.User
## 五、URL 路由拆分
- config/urls_public.py公共域名、官网、注册、登录、超级管理后台
- config/urls_tenant.py租户子域名、租户后台、房源/客源/带看/员工模块
## 六、自定义 User Model跨租户关键
- User 继承 AbstractUser
- role 支持平台超管、公司管理员、门店经理、资深经纪人、经纪人、实习经纪人
- Branch 作为门店模型
## 七、初始化与常用命令
- createdb realestate_saas
- python manage.py migrate_schemas --shared
- python manage.py createsuperuser
- python manage.py shell 创建 Company 与 Domain
示例:
- schema_name = zuoan
- domain = zuoan.localhost
- 访问 http://zuoan.localhost:8000/admin/ 进入专属后台
## 八、本地开发配置hosts 文件)
- 127.0.0.1 localhost
- 127.0.0.1 zuoan.localhost
- 127.0.0.1 lianhe.localhost
- 127.0.0.1 xincheng.localhost
开发环境要点:
- ALLOWED_HOSTS 包含 .localhost
- 本地不用 HTTPS
## 九、数据隔离验证
使用 schema_context 切换 schema验证 Listing 等数据互相隔离。
## 下一步建议
推荐顺序:
1. 先做房源/客源/带看完整数据模型
2. 再做 Django Admin 深度定制Jazzmin 主题)
3. 最后补三级权限体系(总部/门店/经纪人)
---
# 上海房产中介 SaaS 系统规划
## 一、多租户架构选型
Django 多租户有三种主流方案,针对这个场景推荐独立 Schema 方案,也就是基于 PostgreSQL Schema 隔离的 django-tenants。
| 方案 | 原理 | 优点 | 缺点 | 适用场景 |
|---|---|---|---|---|
| 共享 Schema | 每张表加 tenant_id 字段 | 简单,运维成本低 | 数据隔离风险高 | 小规模、低安全需求 |
| 独立 Schema | PostgreSQL Schema 隔离 | 隔离好,性能佳 | 稍复杂 | 中介 SaaS推荐 |
| 独立数据库 | 每租户独立 DB | 最高隔离 | 运维成本极高 | 超高安全需求 |
推荐使用 django-tenants
pip install django-tenants
## 二、整体系统模块规划
SaaS 平台层:
- 租户注册
- 套餐管理
- 计费
- 超级后台
各中介公司 Tenant
- 房源管理
- 客源管理
- 员工 / 权限管理
- 带看管理
- 合同管理
- 报表 / BI 看板
- 渠道推广
- 财务佣金
- 消息 / 通知中心
核心 Django App 拆分:
- tenants租户管理公共 Schema
- accounts用户 / 员工 / 角色权限
- listings房源管理核心
- clients客源 / 客户跟进
- showings带看记录
- contracts合同管理
- commissions佣金 / 财务
- reports数据报表
- notifications消息通知
- subscription套餐订阅公共 Schema
## 三、房源模块数据模型(核心)
Listing 关键字段包括:
- 基础信息title、listing_type、status
- 上海地址结构district、street、community、building、floor、unit
- 房屋属性area、inner_area、layout、orientation、decoration、floor_total
- 价格price、price_unit
- 归属agent、source、exclusive
- 证件certificate_no
- 时间created_at、updated_at
## 四、技术栈推荐
后端:
- Django 5.x + DRF
- django-tenants
- django-guardian
- celery + redis
- django-filter
- PostgreSQL 15+
前端路径:
- 路径 ADjango Admin + 定制,最快上手,适合 MVP
- 路径 BHTMX + Alpine.js + Tailwind推荐中期方案
- 路径 CVue3 / React + DRF长期推荐
建议:先 A再 B最后 C。
## 五、租户路由设计
通过子域名区分租户:
- company-a.yourapp.com
- company-b.yourapp.com
- admin.yourapp.com
核心设置:
- TENANT_MODEL = "tenants.Company"
- TENANT_DOMAIN_MODEL = "tenants.Domain"
- SHARED_APPS 放公共应用
- TENANT_APPS 放租户私有应用
## 六、开发阶段规划
Phase 11-2 月MVP
- 租户注册 / 登录
- 房源 CRUD + 图片上传
- 客源基础管理
- Django Admin 后台
Phase 22-3 月)核心业务:
- 带看流程
- 合同模板 + 生成
- 员工角色权限
- 基础报表
Phase 33-4 月)增长功能:
- 佣金结算
- 渠道推广(链家 / 安居客对接)
- 微信小程序端
- BI 数据看板
Phase 4 商业化:
- 套餐 / 计费系统
- 多城市扩展
## 七、优先推进建议
如果要最快落地,建议优先顺序:
1. django-tenants 完整配置
2. 房源 / 客源 / 带看数据模型
3. Django Admin 深度定制
4. 权限系统设计
5. 前端升级到 HTMX 或 Vue
---
title: django-tenants 完整配置指南
created: 2026-04-21
tags: [django, django-tenants, postgresql, saas, multi-tenant]
category: 技术笔记
---
# django-tenants 完整配置指南
## 一、安装依赖
pip install django-tenants psycopg2-binary django-jazzmin
## 二、项目目录结构
myproject/
├── config/
│ ├── settings/
│ │ ├── base.py
│ │ ├── development.py
│ │ └── production.py
│ ├── urls.py
│ └── wsgi.py
├── apps/
│ ├── tenants/
│ ├── subscription/
│ ├── accounts/
│ ├── listings/
│ ├── clients/
│ └── showings/
├── manage.py
└── requirements.txt
## 三、核心 Model租户与域名
- Company 继承 TenantMixin
- Domain 继承 DomainMixin
- 每一个中介公司 = 一个租户 = 一个独立 PostgreSQL Schema
- 每个公司可绑定多个域名/子域名
## 四、Settings 完整配置
关键点:
- SHARED_APPS 放公共 Schema 应用
- TENANT_APPS 放租户私有应用
- TENANT_MODEL = "tenants.Company"
- TENANT_DOMAIN_MODEL = "tenants.Domain"
- DATABASES 使用 django_tenants.postgresql_backend
- DATABASE_ROUTERS 使用 TenantSyncRouter
- TenantMainMiddleware 必须第一个
- ROOT_URLCONF / PUBLIC_SCHEMA_URLCONF 分离
- AUTH_USER_MODEL = accounts.User
## 五、URL 路由拆分
- config/urls_public.py公共域名、官网、注册、登录、超级管理后台
- config/urls_tenant.py租户子域名、租户后台、房源/客源/带看/员工模块
## 六、自定义 User Model跨租户关键
- User 继承 AbstractUser
- role 支持平台超管、公司管理员、门店经理、资深经纪人、经纪人、实习经纪人
- Branch 作为门店模型
## 七、初始化与常用命令
- createdb realestate_saas
- python manage.py migrate_schemas --shared
- python manage.py createsuperuser
- python manage.py shell 创建 Company 与 Domain
示例:
- schema_name = zuoan
- domain = zuoan.localhost
- 访问 http://zuoan.localhost:8000/admin/ 进入专属后台
## 八、本地开发配置hosts 文件)
- 127.0.0.1 localhost
- 127.0.0.1 zuoan.localhost
- 127.0.0.1 lianhe.localhost
- 127.0.0.1 xincheng.localhost
开发环境要点:
- ALLOWED_HOSTS 包含 .localhost
- 本地不用 HTTPS
## 九、数据隔离验证
使用 schema_context 切换 schema验证 Listing 等数据互相隔离。
## 下一步建议
推荐顺序:
1. 先做房源/客源/带看完整数据模型
2. 再做 Django Admin 深度定制Jazzmin 主题)
3. 最后补三级权限体系(总部/门店/经纪人)
---
# 上海房产中介 SaaS 系统规划
## 一、多租户架构选型
Django 多租户有三种主流方案,针对这个场景推荐独立 Schema 方案,也就是基于 PostgreSQL Schema 隔离的 django-tenants。
| 方案 | 原理 | 优点 | 缺点 | 适用场景 |
|---|---|---|---|---|
| 共享 Schema | 每张表加 tenant_id 字段 | 简单,运维成本低 | 数据隔离风险高 | 小规模、低安全需求 |
| 独立 Schema | PostgreSQL Schema 隔离 | 隔离好,性能佳 | 稍复杂 | 中介 SaaS推荐 |
| 独立数据库 | 每租户独立 DB | 最高隔离 | 运维成本极高 | 超高安全需求 |
推荐使用 django-tenants
pip install django-tenants
## 二、整体系统模块规划
SaaS 平台层:
- 租户注册
- 套餐管理
- 计费
- 超级后台
各中介公司 Tenant
- 房源管理
- 客源管理
- 员工 / 权限管理
- 带看管理
- 合同管理
- 报表 / BI 看板
- 渠道推广
- 财务佣金
- 消息 / 通知中心
核心 Django App 拆分:
- tenants租户管理公共 Schema
- accounts用户 / 员工 / 角色权限
- listings房源管理核心
- clients客源 / 客户跟进
- showings带看记录
- contracts合同管理
- commissions佣金 / 财务
- reports数据报表
- notifications消息通知
- subscription套餐订阅公共 Schema
## 三、房源模块数据模型(核心)
Listing 关键字段包括:
- 基础信息title、listing_type、status
- 上海地址结构district、street、community、building、floor、unit
- 房屋属性area、inner_area、layout、orientation、decoration、floor_total
- 价格price、price_unit
- 归属agent、source、exclusive
- 证件certificate_no
- 时间created_at、updated_at
## 四、技术栈推荐
后端:
- Django 5.x + DRF
- django-tenants
- django-guardian
- celery + redis
- django-filter
- PostgreSQL 15+
前端路径:
- 路径 ADjango Admin + 定制,最快上手,适合 MVP
- 路径 BHTMX + Alpine.js + Tailwind推荐中期方案
- 路径 CVue3 / React + DRF长期推荐
建议:先 A再 B最后 C。
## 五、租户路由设计
通过子域名区分租户:
- company-a.yourapp.com
- company-b.yourapp.com
- admin.yourapp.com
核心设置:
- TENANT_MODEL = "tenants.Company"
- TENANT_DOMAIN_MODEL = "tenants.Domain"
- SHARED_APPS 放公共应用
- TENANT_APPS 放租户私有应用
## 六、开发阶段规划
Phase 11-2 月MVP
- 租户注册 / 登录
- 房源 CRUD + 图片上传
- 客源基础管理
- Django Admin 后台
Phase 22-3 月)核心业务:
- 带看流程
- 合同模板 + 生成
- 员工角色权限
- 基础报表
Phase 33-4 月)增长功能:
- 佣金结算
- 渠道推广(链家 / 安居客对接)
- 微信小程序端
- BI 数据看板
Phase 4 商业化:
- 套餐 / 计费系统
- 多城市扩展
## 七、优先推进建议
如果要最快落地,建议优先顺序:
1. django-tenants 完整配置
2. 房源 / 客源 / 带看数据模型
3. Django Admin 深度定制
4. 权限系统设计
5. 前端升级到 HTMX 或 Vue

View File

@@ -1,98 +1,98 @@
---
title: django-tenants 完整配置指南
created: 2026-04-21
tags: [django, django-tenants, postgresql, saas, multi-tenant]
category: 技术笔记
---
# django-tenants 完整配置指南
## 一、安装依赖
pip install django-tenants psycopg2-binary django-jazzmin
## 二、项目目录结构
myproject/
├── config/
│ ├── settings/
│ │ ├── base.py
│ │ ├── development.py
│ │ └── production.py
│ ├── urls.py
│ └── wsgi.py
├── apps/
│ ├── tenants/
│ ├── subscription/
│ ├── accounts/
│ ├── listings/
│ ├── clients/
│ └── showings/
├── manage.py
└── requirements.txt
## 三、核心 Model租户与域名
- Company 继承 TenantMixin
- Domain 继承 DomainMixin
- 每一个中介公司 = 一个租户 = 一个独立 PostgreSQL Schema
- 每个公司可绑定多个域名/子域名
## 四、Settings 完整配置
关键点:
- SHARED_APPS 放公共 Schema 应用
- TENANT_APPS 放租户私有应用
- TENANT_MODEL = "tenants.Company"
- TENANT_DOMAIN_MODEL = "tenants.Domain"
- DATABASES 使用 django_tenants.postgresql_backend
- DATABASE_ROUTERS 使用 TenantSyncRouter
- TenantMainMiddleware 必须第一个
- ROOT_URLCONF / PUBLIC_SCHEMA_URLCONF 分离
- AUTH_USER_MODEL = accounts.User
## 五、URL 路由拆分
- config/urls_public.py公共域名、官网、注册、登录、超级管理后台
- config/urls_tenant.py租户子域名、租户后台、房源/客源/带看/员工模块
## 六、自定义 User Model跨租户关键
- User 继承 AbstractUser
- role 支持平台超管、公司管理员、门店经理、资深经纪人、经纪人、实习经纪人
- Branch 作为门店模型
## 七、初始化与常用命令
- createdb realestate_saas
- python manage.py migrate_schemas --shared
- python manage.py createsuperuser
- python manage.py shell 创建 Company 与 Domain
示例:
- schema_name = zuoan
- domain = zuoan.localhost
- 访问 http://zuoan.localhost:8000/admin/ 进入专属后台
## 八、本地开发配置hosts 文件)
- 127.0.0.1 localhost
- 127.0.0.1 zuoan.localhost
- 127.0.0.1 lianhe.localhost
- 127.0.0.1 xincheng.localhost
开发环境要点:
- ALLOWED_HOSTS 包含 .localhost
- 本地不用 HTTPS
## 九、数据隔离验证
使用 schema_context 切换 schema验证 Listing 等数据互相隔离。
## 下一步建议
推荐顺序:
1. 先做房源/客源/带看完整数据模型
2. 再做 Django Admin 深度定制Jazzmin 主题)
3. 最后补三级权限体系(总部/门店/经纪人)
---
title: django-tenants 完整配置指南
created: 2026-04-21
tags: [django, django-tenants, postgresql, saas, multi-tenant]
category: 技术笔记
---
# django-tenants 完整配置指南
## 一、安装依赖
pip install django-tenants psycopg2-binary django-jazzmin
## 二、项目目录结构
myproject/
├── config/
│ ├── settings/
│ │ ├── base.py
│ │ ├── development.py
│ │ └── production.py
│ ├── urls.py
│ └── wsgi.py
├── apps/
│ ├── tenants/
│ ├── subscription/
│ ├── accounts/
│ ├── listings/
│ ├── clients/
│ └── showings/
├── manage.py
└── requirements.txt
## 三、核心 Model租户与域名
- Company 继承 TenantMixin
- Domain 继承 DomainMixin
- 每一个中介公司 = 一个租户 = 一个独立 PostgreSQL Schema
- 每个公司可绑定多个域名/子域名
## 四、Settings 完整配置
关键点:
- SHARED_APPS 放公共 Schema 应用
- TENANT_APPS 放租户私有应用
- TENANT_MODEL = "tenants.Company"
- TENANT_DOMAIN_MODEL = "tenants.Domain"
- DATABASES 使用 django_tenants.postgresql_backend
- DATABASE_ROUTERS 使用 TenantSyncRouter
- TenantMainMiddleware 必须第一个
- ROOT_URLCONF / PUBLIC_SCHEMA_URLCONF 分离
- AUTH_USER_MODEL = accounts.User
## 五、URL 路由拆分
- config/urls_public.py公共域名、官网、注册、登录、超级管理后台
- config/urls_tenant.py租户子域名、租户后台、房源/客源/带看/员工模块
## 六、自定义 User Model跨租户关键
- User 继承 AbstractUser
- role 支持平台超管、公司管理员、门店经理、资深经纪人、经纪人、实习经纪人
- Branch 作为门店模型
## 七、初始化与常用命令
- createdb realestate_saas
- python manage.py migrate_schemas --shared
- python manage.py createsuperuser
- python manage.py shell 创建 Company 与 Domain
示例:
- schema_name = zuoan
- domain = zuoan.localhost
- 访问 http://zuoan.localhost:8000/admin/ 进入专属后台
## 八、本地开发配置hosts 文件)
- 127.0.0.1 localhost
- 127.0.0.1 zuoan.localhost
- 127.0.0.1 lianhe.localhost
- 127.0.0.1 xincheng.localhost
开发环境要点:
- ALLOWED_HOSTS 包含 .localhost
- 本地不用 HTTPS
## 九、数据隔离验证
使用 schema_context 切换 schema验证 Listing 等数据互相隔离。
## 下一步建议
推荐顺序:
1. 先做房源/客源/带看完整数据模型
2. 再做 Django Admin 深度定制Jazzmin 主题)
3. 最后补三级权限体系(总部/门店/经纪人)