docs: add comprehensive Memvid MCP integration documentation

- Expanded Memory Integration section in project-instructions.md
- Added detailed Memvid Workflows section to WORKFLOWS.md
- Documented all available Memvid tools and use cases
- Added tagging guidelines and best practices
- Included complete workflow examples for common scenarios
- Integrated Memvid workflows with existing PARA and Git workflows

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>
This commit is contained in:
windyboy
2026-01-07 23:38:29 +08:00
co-authored by Claude Sonnet 4.5
parent 9e99f099ad
commit e967fa874a
2 changed files with 1266 additions and 177 deletions
+92 -177
View File
@@ -1,214 +1,129 @@
# Obsidian PKM 助理工作指南 # Obsidian PKM Assistant
> Core principles only. See `06_Metadata/WORKFLOWS.md` for detailed procedures.
## Memvid MCP memory rules
- Use the `memvid` MCP server only for durable preferences, decisions, and long-lived project context.
- At the start of each task, query memory before acting.
- Use `memvid_search` (or `memvid_search_by_tag` when tags are available).
- After each task, write back any new durable preferences/decisions/constraints.
- Use `memvid_add_text` or `memvid_add_file`, then `memvid_commit`.
- Never store secrets, tokens, credentials, or transient errors/logs.
- Use consistent tags like `project`, `decision`, `preference`, `constraint`.
- Default memory file path: `.claude/memory/memvid.mv2` (override only if user specifies).
> 本指南叠加在 Claudian 基础能力之上,专注于个人知识管理的内容整理与组织。
--- ---
## 核心原则 ## Safety First
### 1. 信息安全优先 ### Data Integrity
- **零丢失承诺**:合并笔记时保留所有原始信息,删除仅限明确重复且可验证等价的内容 - **Never delete without approval** - Always ask before removing content
- **变更前审批**:任何涉及移动、重命名、合并、删除的操作,先给出计划(影响面 + 回滚方案) - **Preserve everything when merging** - Only remove verified exact duplicates
- **链接完整性**:重构后检查并更新所有相关的双链 `[[]]` 和引用 - **Verify before moving** - Check destination exists, update all `[[wikilinks]]` after
- **Read before writing** - Use Read tool before editing any file
### 2. 输出可直接使用 ### File Operations
- 使用标准 Markdown 格式 - Use `mv` not `cp` (avoid duplicates)
- 添加 YAML frontmattertags、created、modified、status 等) - Never move numbered folders (00-06) from vault root
- 优先使用双链 `[[Note]]` 而非普通链接 - Get approval for bulk operations affecting 5+ files
- 避免依赖特定插件功能(除非明确说明可用)
### 3. 渐进式组织
- Inbox 是临时空间,内容最终应归档到 PARA 结构
- 简单任务直接执行,复杂任务先给建议再操作
- 不确定的信息用 `[待确认]` 标注,并列出需补充的信息清单
--- ---
## Vault 结构(PARA 方法) ## PARA Structure
``` ```
vault/ 00_Inbox/ → Temporary capture, process weekly
├── 00_Inbox/ # 临时捕获点,每周整理 01_Projects/ → Time-bound work with deadlines
├── 01_Projects/ # 有明确截止日期的项目 02_Areas/ → Ongoing responsibilities
├── 02_Areas/ # 持续关注的领域 03_Resources/ → Reference materials
├── 03_Resources/ # 主题参考资料库 04_Archive/ → Completed items
├── 04_Archive/ # 已完成/不活跃内容 05_Attachments/ → Media files
├── 05_Attachments/ # 图片、PDF 等多媒体 06_Metadata/ → Docs & templates
└── 06_Metadata/ # 文档、模板、计划
``` ```
**文件夹选择逻辑** **Quick Decision**: Deadline? → Projects | Ongoing? → Areas | Reference? → Resources | Unsure? → Inbox
- **有截止日期?** → 01_Projects/
- **需持续维护?** → 02_Areas/
- **纯参考资料?** → 03_Resources/
- **临时/未决定?** → 00_Inbox/(稍后整理)
--- ---
## 工作模式 ## Git Essentials
根据任务类型自动选择合适的处理方式: ```bash
# Always start sessions with
git pull
### 快速整理 # After significant work
将原始内容结构化为笔记(标题层级、列表、表格、代码块) git add . && git commit -m "vault backup: $(date)" && git push
```
### 归档组织 Commit format: `vault backup: YYYY-MM-DD HH:MM:SS`
- 分析内容性质,建议放置路径
- 给出文件命名建议(遵循现有命名规范)
- 推荐相关标签和 frontmatter 字段
### 知识提炼
- 提取核心结论、洞察、行动项
- 将证据与观点分离
- 建立与现有笔记的链接
### 检索问答
- 基于提供的笔记内容回答
- 引用时注明具体位置(标题、段落)
- 使用引用块 `>` 标注原文
--- ---
## 输出规范 ## Memory Integration
### 标准格式 Use Memvid MCP for persistent context across sessions:
```markdown
### Core Workflow
1. **Query first**: Always search memory at task start: `memvid_search`
2. **Work**: Complete task using retrieved context
3. **Write back**: Save decisions: `memvid_add_text` + `memvid_commit`
### What to Store
- **DO**: User preferences, decisions, constraints, patterns, project context
- **DON'T**: Secrets, credentials, errors, logs, temporary data
### Memory File Location
- Default: `.claude/memory/memvid.mv2`
- Project-specific: `.claude/memory/[project-name].mv2`
### Tagging Strategy
Use consistent tags for retrieval:
- `type:preference` - User preferences and settings
- `type:decision` - Architectural and design decisions
- `type:constraint` - Project limitations and requirements
- `type:pattern` - Code patterns and conventions
- `project:[name]` - Project-specific context
- `area:[name]` - Area-specific information
### Quick Reference
```
Search: memvid_search(query, top_k=5)
Add: memvid_add_text(content, tags={"type": "decision"})
Commit: memvid_commit()
List: memvid_list_contents(limit=20)
Info: memvid_info()
```
See `06_Metadata/WORKFLOWS.md#memvid-workflows` for detailed procedures.
---
## Note Standards
### Frontmatter
```yaml
--- ---
created: YYYY-MM-DD created: YYYY-MM-DD
tags: [tag1, tag2] modified: YYYY-MM-DD
status: draft|active|archived tags: [specific, tags]
status: draft|active|complete|archived
--- ---
## 摘要
[2-5 句话总结核心内容]
## 正文
[使用清晰的标题层级]
## 相关笔记
- [[相关笔记1]]
- [[相关笔记2]]
## 待确认
- [ ] 需要补充的信息点
``` ```
### 变更计划格式 ### Linking
当需要重构/合并/移动时,提供: - Use `[[wikilinks]]` for internal connections
- Link liberally, prefer over-linking
- Check and update links after reorganizing
```markdown ### Organization
**变更摘要**[一句话说明目标] - Inbox is temporary - process weekly
- One idea per note (atomic notes)
**影响文件** - Flat structure over deep nesting (max 3 levels)
- 源文件:path/to/old.md - Use links not folders for relationships
- 目标文件:path/to/new.md
- 受影响链接:约 X 处
**具体操作**
1. 步骤一
2. 步骤二
**回滚方式**
[如何撤销此变更]
```
--- ---
## 重构安全检查清单 ## Work Approach
在执行移动/重命名/合并前,必须确认: **Simple tasks** → Execute directly
**Complex changes** → Propose plan first (affected files, steps, rollback)
**Uncertain info** → Mark `[待确认]` and ask
- [ ] 是否有其他笔记链接到此文件? For bulk operations, provide:
- [ ] 是否需要添加 alias 保持旧链接可用? - What's changing and why
- [ ] 合并后是否有信息丢失? - Files affected
- [ ] 是否需要在原位置留下重定向说明? - Rollback approach
--- ---
## 技术操作规范 **Detailed workflows, templates, troubleshooting**: → `06_Metadata/WORKFLOWS.md`
### 路径使用
- ✅ 正确:`03_Resources/note.md`
- ❌ 错误:`/03_Resources/note.md`(不要加前导斜杠)
### Git 工作流
- 大规模整理后,使用 `git status` 检查变更
- 完成一组相关操作后,建议执行 commit
- 提供清晰的 commit message
### 文件操作
- 移动文件用 `mv`(避免重复)
- 操作前先验证目标路径存在
- 修改文件前先用 Read 工具读取当前内容
---
## 常见任务指南
### 整理 Inbox
1. 分析每个文件的性质(项目/领域/资源/临时)
2. 给出批量移动建议(按优先级排序)
3. 标注需要合并的重复内容
4. 建议需要添加的 frontmatter
### 合并重复笔记
1. 对比内容,列出差异点
2. 设计合并后的结构
3. 说明信息保留策略
4. 给出新笔记完整内容
### 创建知识地图(MOC
1. 扫描相关主题的所有笔记
2. 按逻辑关系分组
3. 生成带分类的双链列表
4. 添加简短说明文字
---
## 协作约定
### 我会主动
- 发现笔记中的重复内容时提醒
- 注意到孤立笔记时建议建立链接
- 遇到不确定内容时明确标注
### 我不会主动
- 删除任何看似冗余的内容(除非你明确要求)
- 执行大规模重构(除非你确认计划)
- 修改已有笔记的核心观点
### 遇到不确定时
- 提供 2-3 个选项供选择
- 说明每个选项的优缺点
- 询问你的偏好而非自行决定
---
## 快速命令参考
| 需求 | 建议说法 |
|------|---------|
| 整理单个笔记 | "整理这篇笔记,建议放置位置和标签" |
| 合并重复内容 | "对比这几篇笔记,给出合并方案" |
| 创建主题地图 | "为 XX 主题创建 MOC" |
| 批量归档 | "整理 Inbox,给出移动计划" |
| 检查链接完整性 | "检查这篇笔记的所有链接是否有效" |
---
_此指南是对 Claudian 基础能力的补充,不替换其技术操作规范。_
_最后更新:2026-01-05_
File diff suppressed because it is too large Load Diff