Skip to content

Version 0.0.3

Markdown AI Editor 0.0.3 makes local document preparation more useful for agent-assisted development while keeping managed processing private by default.

New capabilities

  • A single Markdown AI Editor right-click submenu keeps transformations together.
  • Cross-platform keyboard shortcuts are available for all transformations.
  • Convert to AI Skill Format creates a reusable SKILL.md workflow.
  • Convert to AI Prompt Format creates a structured reusable prompt.
  • Convert to PRD creates a practical product requirements document.
  • The status bar shows the managed server's dynamic localhost port while it is running.

Why Qwen3

Managed mode now uses the pinned Qwen3-0.6B-Q8_0 model. It is a modern instruction-following model better suited to structured transformations than the prior small writing model, while keeping the model download near 640 MB. The extension also updates its pinned llama-server runtime so Qwen3 runs through the same private localhost workflow.

The model and runtime are downloaded only after your approval, over HTTPS, and verified against built-in SHA-256 checksums. The managed server binds to 127.0.0.1; selected text stays on your machine in managed mode.

Practical token hygiene

Use the extension to make rough notes concise and structured before sharing them with Copilot, Codex, Claude Code, or another coding agent. Removing repeated prose and organizing requirements can reduce repeated context and prompt tokens. This is preparation, not a token-count guarantee: external agents retain their own tokenization, billing, behavior, and privacy policies.

Prerequisites

  • VS Code 1.85 or later.
  • Windows x64, macOS Intel, macOS Apple Silicon, or Linux x64 for managed mode.
  • About 1 GB free disk space and 2 GB available RAM.
  • Internet access for the consented first-time download. After setup, managed editing works offline.

See First run, Managed local model, and Privacy and limitations before using a custom endpoint.