Zditor

Documentation - Official Website

中文版

---

Welcome to Zditor, an AI-powered markdown editor.

This repository is not for source code, it only contains Zditor-related documentation.

Table of Contents

• Getting Started
• Roadmap

Getting Started

Zditor is a WYSIWYG markdown editor that supports basic markdown syntax with extensions. Currently available for Windows, Macos, and Linux platforms. Download from the releases page.

video

zditor demo

Signature Issues

• Macos currently lacks code signing. After downloading, you'll need two additional steps:

  1. During installation
  2. When opening for the first time Go to Privacy & Security settings and select "Open Anyway", as shown below:

LLM Model Configuration

Uses OpenAI SDK's streaming output, supporting both API and local models. As shown below, enter the relevant parameters. For models, separate them with commas (English commas). The ollama API key can be left blank.

Using LLM Features

Supports text selection and dedicated chat tabs. Text selection enables single-turn conversations, while chat tabs support multi-turn conversations with file uploads, document selection as context, and drag-and-drop uploads.

License

Provides free licenses. Just enter your email, click send, and retrieve the license from your email.

Roadmap

• UI Optimization: Current UI needs improvement, especially in light mode. Considering using Claude for optimization.
• Bug Fixes: Many bugs exist, like mermaid sometimes failing to intercept errors causing layout issues. Prioritizing critical fixes.
• Infinite Canvas Inspiration: Inspired by Affine's infinite canvas, currently Zditor has Excalidraw but needs better integration with markdown:
  • Performance optimization (current performance issues may relate to webview)
  • Embed Excalidraw in markdown documents
  • AI conversion of markdown to Excalidraw
  • Fine-grained block-level embedding of Excalidraw in markdown
• Git Module Improvements:
  • Replaced slow open-source diff package with custom solution (handles 1M+ documents instantly), needs UI polish
  • Current git functionality (commit/push/pull) is too basic
• Multilingual Support: Currently only partial modules support multiple languages
• Performance Optimization:
  • While compute-intensive tasks are implemented in Rust, more optimization needed
  • Some ProseMirror plugins from open source need optimization
  • File opening performance still insufficient (though better than mainstream ProseMirror editors)
  • Goal: Achieve instant opening of 1M+ character documents (like Obsidian)
• Shortcuts: Many exist but need refinement

Thoughts 🤔

• Considered knowledge base features but rejected due to large app size (model embeddings) or requiring external services. Goal is to keep app size under 20MB.
• Global text selection plugin would require simulating mod c/v (with side effects) or deeper system APIs (though Swift experience exists, not ObjC). Douban's implementation seems good (likely uses simulated mod c/v 😂)
• Planning a state machine plugin - more advanced than mind maps. When AI analyzes code with if-else branches, it could generate Excalidraw-style hand-drawn mind maps. State machines would guide next steps based on conditions, handling loops better than mind maps.