NoteFlowOS

KO | EN

gitlite — search

#ai #typescript #llm #go #machine-learning #linux #docker #macos #java #windows #awesome #awesome-list #vue #mcp

(★ 10)

An AI-powered knowledge system for Obsidian that turns sources into projects and reusable knowledge.

NoteFlowOS 최신버젼 다운로드

최종 버전 다운로드 (.zip)

// repository documentation

Was this content helpful?

★ 0 (0 ratings)

# NoteFlowOS

> A knowledge operating system built on Obsidian. You capture and think — AI classifies, connects, and maintains.

[Quick Start](#quick-start) · [Operating Rhythm](#operating-rhythm) · [Core Loop](#core-loop) · [Design Philosophy](#design-philosophy)

---

## Design Motivation

The core tension in personal knowledge management: **the moment you record something is not the moment you should organize it.**

When you encounter valuable content, your attention is on the content itself. If you also have to decide "which folder?" or "what category?" at that moment, most people default to — save it now, sort it later. The result: notes pile up, but retrievable, reusable knowledge does not grow.

NoteFlowOS design principle: **decouple recording from organizing.**

When recording, write directly into the inbox with zero classification decisions. When ready to process, mark the status — AI automatically handles filing and linking. You own the thinking; the system owns the structure.

---

## Core Loop

```
Inbox → Mark "ready to process" → AI auto-files → Periodic review
```

Four steps, continuously cycling. You handle thinking and judgment; AI handles classification and maintenance.

> **What does "mark ready to process" mean?** Open the note's Properties panel at the top, change the `status` field to `processed`. Obsidian's property editor supports dropdown selection — no need to edit raw files.

> **What is "periodic review"?** Run `/lint` once a week. AI reports dead links, missing fields, and orphan notes. You decide what to act on. See "Operating Rhythm" below.

> **How is filing triggered?** Changing `status` to `processed` places the note in a processing queue. Next time you run `/hello-day`, AI will auto-file these notes (it calls `/compile` internally; running `/compile` manually has the same effect).

---

## Knowledge Flow

```
┌─────────────┐
│    Inbox    │  ← Everything enters here, no classification needed
└──────┬──────┘
       │  Mark status → Awaits next filing command
       ▼
┌─────────────────────────────────────┐
│          Sources                    │  ← Raw material, append-only
│   Feeds/ │  Research/  │  Imports/  │
└──────┬───┴────────────┴────────────┘
       │  AI extracts key information
       ▼
┌──────────────┐     ┌──────────────┐
│   Projects   │     │     Wiki     │  ← Long-term knowledge assets
│  Analyses    │     │   Articles   │
│  PRDs        │     │   Concepts   │
└──────────────┘     └──────────────┘
```

**Unidirectional flow, progressive refinement.** Raw material is never modified; knowledge accumulates through movement.

---

## System Responsibilities

| AI handles | You decide |
|---|---|
| Determine which directory a note belongs in | What is worth recording |
| Build links between notes | Which content is ready to process |
| Maintain indexes and directory structure | What topics to research, what reports to write |
| Detect dead links and orphan notes | Judgments and decisions during periodic review |
| Deduplicate (append to existing topics) | — |

---

## Quick Start

**Prerequisites:**
- [Obsidian](https://obsidian.md/) 1.4 or higher (requires the Properties panel)
- An AI coding assistant that supports AGENTS.md (recommended: [Claude Code](https://docs.anthropic.com/en/docs/claude-code); also works with Cursor, GitHub Copilot, etc.)

```bash
# 1. Clone the repo into your Obsidian vault directory
git clone https://github.com/OneMore07/NoteFlowOS.git

# 2. Open the folder as an Obsidian vault

# 3. In Obsidian: Settings → Core plugins → Enable "Templates" → Template options → Set folder to 99_System/模板/

# 4. Open your AI assistant, type /hello-day
```

> **How to launch with different tools:**
> - Claude Code: run `claude` in the repo directory, type `/hello-day`
> - Cursor / GitHub Copilot: open the Chat panel, type `/hello-day`
> - Other tools: paste AGENTS.md content into the system prompt, then type `/hello-day`

If everything works, AI will generate this week's todo document in `10_Journal/01_WEEK/`.

If AI doesn't recognize `/hello-day`, verify you launched from the repo root, or manually provide the AGENTS.md content.

---

## AI Tool Compatibility

The `AGENTS.md` file in the project root serves as the AI instruction file, defining all management rules. Major AI coding tools can read it:

| Tool | Integration |
|---|---|
| Claude Code | Auto-loaded when launched in repo directory |
| GitHub Copilot | Auto-loaded |
| OpenAI Codex | Auto-loaded |
| Cursor | Auto-loaded |
| Tongyi / Kimi / MiniMax | Paste AGENTS.md content into system prompt |

No additional Obsidian plugins required. To use `/import` for PDF/Word files, you need Python and a conversion tool: `pip install markitdown`

---

## Operating Rhythm

Enter these commands in your AI assistant's chat:

| When | You do | AI does |
|---|---|---|
| Every morning | Type `/hello-day` | Generate weekly todo, carry over incomplete tasks, file marked notes, flag inbox backlog |
| Anytime | Change note's status to `processed` in Properties | Note enters processing queue; auto-filed on next `/hello-day` or `/compile` |
| After a conversation | Type `/distill` | Extract insights worth keeping, save to Wiki |
| End of week | Type `/lint` | Health check: dead links, missing fields, orphan notes |

---

## Skill Reference

Beyond those mentioned in Operating Rhythm:

| Skill | Description |
|---|---|
| `/compile` | File queued notes to correct locations (`/hello-day` calls this automatically; can also run manually) |
| `/reindex` | Rebuild all directory index files |
| `/ai-curation` | Fetch AI news, generate structured digests |
| `/research` | Multi-round deep research, output structured reports |
| `/import` | Import PDF/Word/HTML files |

---

## Directory Structure

```
00_Inbox/           Inbox (just write, no classification needed)
10_Journal/         Journal
  ├── 00_YEAR/      Yearly reviews
  └── 01_WEEK/      Weekly todo + weekly report
20_Projects/        Active projects (reports, PRDs, analyses)
30_Wiki/            Knowledge base (long-term assets)
  ├── Articles/     Deep-dive topics
  └── Concepts/     Atomic concept cards
40_Sources/         Source material (append-only)
  └── Feeds/        News and newsletters
99_System/          System config (templates, rules, skill definitions)
```

---

## Design Philosophy

NoteFlowOS draws from three proven knowledge management methodologies, with an added AI automation layer:

**Three-layer isolation, unidirectional flow**
Raw input → processed material → refined output. Each layer has a single responsibility, no mixing. (From Zettelkasten / slip-box method.)

**Project and resource separation**
Active projects in Projects, reference material in Sources, distilled knowledge in Wiki. Each serves its purpose without contamination. (From Tiago Forte's PARA method.)

**Record first, organize later**
The inbox is deliberately broad: anything can enter. Organization is handled by the system's periodic rhythm, not personal willpower. (From Building a Second Brain's core principle.)

**Auto-sorting, zero decision cost**
No classification decisions at recording time. Once marked "ready to process," AI files according to Schema rules. Recording and organizing are fully decoupled — this is the core capability the AI layer brings.

---

## Acknowledgements

- **Tiago Forte** — PARA Method and Building a Second Brain
- **Niklas Luhmann** — Zettelkasten, the original linked note system
- **Obsidian** — The tool that makes this possible
- **Microsoft markitdown** — PDF/Word conversion capability

---

## License

[MIT](LICENSE)

NoteFlowOS

NoteFlowOS 최신버젼 다운로드

오늘의 추천 프로젝트

Was this content helpful?

Recent Feedback

NoteFlowOS

NoteFlowOS 최신버젼 다운로드

오늘의 추천 프로젝트

Was this content helpful?

Recent Feedback

Download README