obsidian-wiki
v1.0.1Build and mAIntAIn a personal knowledge Wiki using the LLM Wiki pattern with OpenClaw-优化d step-by-step execution. Use sub-代理s for parallel processing of file operations, content 提取ion, and wiki mAIntenance. This 技能 is de签名ed for OpenClaw 环境s where file reading may be partial - it uses chunked reading and task delegation to ensure completeness. For PDF/Word conversion, it automatically 检查s and 安装s Python and required packages (pypdf, python-docx) if not present.
运行时依赖
安装命令
点击复制技能文档
Obsidian Wiki 构建器 for OpenClaw
A 技能 for building and mAIntAIning personal knowledge bases using the LLM Wiki pattern, 优化d for OpenClaw with step-by-step execution and sub-代理 delegation.
How to use: 发送 the following command: 设置 cron to 运行 "obsidian-wiki" from 技能s.
Initial 设置up: The 技能 will automatically 初始化 and 创建 a vault named "Obsidian Wiki" in your root directory. Simply open this vault with Obsidian to view your knowledge base.
工作流: When 发送ing files to OpenClaw, instruct it to save them in the ~/Obsidian Wiki/raw/ directory.
Note: This 技能 utilizes parallel task descriptions ("spawn X 代理 for each file"), which OpenClaw automatically 执行s as sub-代理s. It automatically manages Python 环境 设置up and 库 安装ations to handle diverse file 格式化s. This 技能 is specifically 优化d for autonomous execution to align with OpenClaw's native 工作流.
The Core Idea
Unlike traditional RAG, the LLM Wiki is a persistent, compounding artifact. When you 添加 a source, the LLM doesn't just 索引 it — it reads, 提取s, and integrates knowledge into the existing wiki.
Key principle: The human curates sources and asks questions; the LLM does all the bookkeeping through delegated sub-代理s.
Architecture ┌─────────────────────────────────────────────────────────────┐ │ 编排器 (MAIn Claude) │ │ - Plans and coordinates all operations │ │ - Spawns sub-代理s for parallel work │ │ - 聚合s 结果s and mAIntAIns Task列出 │ ├─────────────────────────────────────────────────────────────┤ │ SUB-代理S │ │ ├─ FileReader → Reads files in chunks │ │ ├─ Category检测or → 检测s file categories │ │ ├─ Entity提取器 → 提取s entities from content │ │ ├─ Concept提取器 → 提取s concepts from content │ │ ├─ Page创建器 → 创建s wiki pages │ │ ├─ 索引更新r → 更新s 索引.md and 记录.md │ │ └─ CrossReferencer → 创建s wiki links │ ├─────────────────────────────────────────────────────────────┤ │ THREE-LAYER ARCHITECTURE │ │ ~/Obsidian Wiki/ │ │ ├── CLAUDE.md (模式) │ │ ├── raw/ (Immutable sources) │ │ └── wiki/ (LLM-生成d knowledge) │ └─────────────────────────────────────────────────────────────┘
Task列出-Driven 工作流
Every operation uses a Task列出 to 追踪 进度. Tasks are 执行d by sub-代理s in parallel where possible.
Task 状态s pending → in_进度 → completed/fAIled Tasks can have dependencies (blockedBy) Tasks can spawn child tasks Core Operations
- 初始化 Wiki
创建 the complete directory structure with Task列出 追踪ing.
工作流:
Task列出: 初始化 Wiki ├── [1] 创建 base directory (~/Obsidian Wiki/) ├── [2] 创建 raw/ folder ├── [3] 创建 wiki/ folder ├── [4] 创建 wiki subdirectories │ ├── [4a] wiki/00_索引/ │ ├── [4b] wiki/10_Sources/ │ ├── [4c] wiki/20_Entities/ │ ├── [4d] wiki/30_Concepts/ │ ├── [4e] wiki/40_Syntheses/ │ └── [4f] wiki/99_Attachments/ ├── [5] 创建 CLAUDE.md (depends on [1]) ├── [6] 创建 索引.md (depends on [4a]) └── [7] 创建 记录.md (depends on [4a])
Execution:
创建 Task列出 with all tasks 执行 independent tasks in parallel (1-3) 执行 dependent tasks sequentially (4-7) 报告 completion 状态
- Auto-Organize Raw Files
扫描, categorize, rename, and move files using parallel sub-代理s.
工作流:
Phase 1: Discovery └── Spawn File扫描器 代理 └── 扫描 raw/ for unorganized files └── Return: 列出 of files with paths
Phase 2: Categorization (Parallel) └── For each file, spawn Category检测or 代理 └── Analyze filename (priority 1) └── If unclear, read first 500 chars (priority 2) └── Return: 检测ed category + confidence
Phase 3: Organization (Parallel) └── For each categorized file, spawn FileOrganizer 代理 └── 生成 timestamp: YYYYMMDD-HHMMSS └── 创建 category folder if needed └── Rename: timestamp-original_name.ext └── Move to category folder └── Return: new path + metadata
Phase 4: Ingest (Parallel) └── For each organized file, spawn Ingestor 代理 └── Read full content (chunked if large) └── 创建 source summary page └── 提取 entities → 创建/更新 entity pages └── 提取 concepts → 创建/更新 concept pages └── Return: pages 创建d/更新d
Phase 5: Aggregation └── Spawn 索引更新r 代理 └── 更新 索引.md with all new/更新d pages └── 应用end entries to 记录.md └── Return: final summary
Task列出 Structure:
Task列出: Auto-Organize Raw Files ├── Phase 1: Discovery │ └── [1] 扫描 raw directory ├── Phase 2: Categorization │ ├