ZarishNote

A free, open-source, ultra-lightweight WYSIWYG Markdown editor with a sandboxed private AI assistant.

ZarishNote is a desktop note-taking application that combines a rich Markdown editor (Milkdown/ProseMirror) with a sandboxed Wasmtime-based AI assistant — all running locally on your machine.

---

Repository Structure

Directory	Purpose
src/	Svelte 5 frontend
src-tauri/	Rust backend (Tauri v2)
ingestion/	Python document ingestion engine
docs/	Blueprint specifications, roadmaps, architecture docs

---

Tech Stack

Layer	Technology
Framework	Tauri v2 (Rust backend, native WebView)
Frontend	Svelte 5 + TypeScript 6 + Vite 8
Package manager	pnpm 11.x
Editor core	Milkdown 7.x (ProseMirror)
Sandbox	Wasmtime 46.x (Rust crate)
Vector store	In-memory HashMap + fastembed
Ingestion	Python CLI (MarkItDown + custom converters)
CI/CD	GitHub Actions

---

Getting Started

Prerequisites: Node.js ≥24.15, pnpm ≥11.5, Rust toolchain (rustup), Tauri system dependencies.

# Clone
git clone https://github.com/devopsariful/zs-note.git
cd zs-note

# Install frontend deps
pnpm install

# Run in dev mode
pnpm dev:tauri

# Or just run Vite dev server (without Tauri window)
pnpm dev

Status: Current version — v0.2.0
CI Status: See GitHub Actions for latest build status.

---

Project Map

zs-note/
├── src/                          # Svelte 5 frontend
│   ├── main.ts                   # Entry point
│   ├── app.css                   # Global styles / CSS variables
│   └── lib/
│       ├── components/           # Svelte components
│       ├── stores/               # Svelte 5 rune stores (.svelte.ts)
│       ├── commands/             # Tauri invoke wrappers
│       ├── milkdown/             # Milkdown editor setup
│       └── types.ts              # TypeScript interfaces
├── src-tauri/                    # Rust backend (Tauri v2)
│   ├── src/
│   │   ├── main.rs / lib.rs      # Entry + plugin registration
│   │   ├── commands/             # Tauri command handlers
│   │   ├── sandbox/              # Wasmtime sandbox engine
│   │   ├── ai/                   # AI provider implementations
│   │   ├── git/                  # Git engine
│   │   ├── mcp/                  # MCP client
│   │   ├── vector/               # Vector store
│   │   ├── config.rs / types.rs  # Configuration and types
│   │   └── capabilities/         # Tauri v2 capability permissions
│   ├── Cargo.toml
│   ├── tauri.conf.json
│   └── tests/                    # Integration tests
├── ingestion/                    # Python ingestion engine
│   ├── src/zarishnote_ingest/
│   │   ├── cli.py                # CLI entry point
│   │   ├── converters/           # Format converters
│   │   └── ...
│   └── pyproject.toml
├── docs/                         # Documentation
│   ├── README.md                 # Blueprint specification
│   ├── TODO.md                   # Implementation status
│   ├── ZARISHNOTE-COMPLETE-GUIDE.md
│   └── [subdirectories for specs]
├── scripts/
│   └── check-consistency.sh      # Repository consistency checker
├── package.json                  # Frontend package config
├── tsconfig.json                 # TypeScript config
├── pnpm-workspace.yaml           # pnpm workspace config
├── vite.config.ts                # Vite build config
└── svelte.config.js              # Svelte config

---

Build Status

See CI Status for latest results.

Check	Tool
Format	cargo fmt --check
Lint	cargo clippy
Tests	cargo test
TypeScript	pnpm typecheck
Python	ruff check src/

Features (Planned / In Progress)

See the blueprint TODO for detailed status breakdown.

Feature	Status
WYSIWYG Markdown editor	🏗 Scaffolded
Source + Split modes	🏗 Scaffolded
File tree + vault manager	🏗 Scaffolded
Git auto-commit + history	🏗 Scaffolded
Wasmtime sandbox	🏗 Scaffolded
AI chat (OpenAI, Claude, Gemini, Ollama)	🏗 Scaffolded
Document ingestion	🏗 Scaffolded
MCP tool integration	🏗 Scaffolded
Vector store / RAG	🏗 Scaffolded
Voice dictation	🏗 Scaffolded
Publish to GitHub	🏗 Scaffolded

---

Architecture

The application follows a three-layer architecture:

1. Rust Backend (src-tauri/) — Tauri v2 commands handle file I/O, Git operations, sandboxed WASM execution, AI provider communication, MCP tool routing, and vector store indexing.
2. Svelte Frontend (src/) — Reactive UI with Milkdown/ProseMirror for the editor, file tree sidebar, AI chat panel, settings, and modals.
3. Python Ingestion Engine (ingestion/) — Standalone CLI that converts documents (PDF, DOCX, PPTX, XLSX, EPUB, HTML, CSV, Jupyter, YouTube, Wikipedia, RSS, SERP) to Markdown.

All AI tools, MCP servers, and plugins execute inside a Wasmtime sandbox with configurable capabilities (filesystem scoping, network allow-list, memory limits, timeouts).

---

Development

Run consistency checks

# Check for inconsistencies
pnpm check:consistency

# Auto-fix common issues
pnpm check:consistency --fix

Run all tests

# Frontend
pnpm typecheck

# Backend
cd src-tauri
cargo test

# Python ingestion
cd ingestion
pytest -v

---

Contributing

1. Read the blueprint specifications to understand the design.
2. Check the TODO for open tasks.
3. File issues at github.com/devopsariful/zs-note/issues.
4. Run pnpm check:consistency before committing.

---

License

MIT — see LICENSE

---

Maintained by devopsariful