Prompt Optimizer Mini

A lightweight VS Code extension that transforms rough ideas into execution-ready prompts with a built-in model router and an optional on-demand local GGUF model.

Install

From VS Code Marketplace (Recommended)

Search "Prompt Optimizer Mini" in VS Code Extensions, or install via CLI:

code --install-extension fullstack1ape.prompt-optimizer-mini

From Source

1. Clone and open this repo in VS Code.
2. Install dependencies and compile:

npm install
npm run compile
python3 -B -m unittest scripts/test_promptopt_stability.py scripts/test_agent_promptopt_contract.py scripts/test_local_model_manager.py

3. Press F5 to launch Extension Development Host.

From Homebrew Tap

After the v1.0.0 release and tap are published:

brew tap Lab-Overflow/prompt-optimizer
brew install prompt-optimizer-mini
promptopt install-extension

Homebrew also exposes the longer prompt-optimizer-mini command.

Build and Install VSIX

npm run package
code --install-extension prompt-optimizer-mini-1.0.0.vsix

Features

• Zero Configuration — Install and start optimizing. If no remote model is available, a compact local model can be prepared on demand.
• Lightweight Package — Model weights are not bundled in the VSIX.
• Model Router GUI — Choose Auto, VS Code Chat, OpenAI-compatible API, Local Model, or Offline Template.
• Visible Optimization Trace — Chat and preview views show the optimized prompt plus an expandable before/after route panel.
• Bring Your Best Model — Configure your own base URL, API key, and model name. API keys are stored in VS Code SecretStorage.
• Local GGUF Model — Reuses Ollama or llama-cli, or downloads a llama.cpp runtime and the official Qwen/Qwen3-1.7B-GGUF Q8_0 model (about 1.83 GB).
• Few-Shot Templates — General, variable-driven, red-team review, and text-to-image prompt patterns are built in.
• Chat Participant — Type @promptopt in VS Code Chat to optimize prompts conversationally.
• Editor Command — Select text and run Prompt Optimizer Mini: Optimize Prompt from the command palette, or right-click for the context menu option.
• Manual Input — No selection needed. Run the command without selecting text and a prompt input box will appear.
• Multiple Output Options — Replace selection, insert at cursor, copy to clipboard, or open as a Markdown preview.
• Resilient Fallback — OpenAI-compatible API → VS Code Chat → Local GGUF model → Local Python formatter → Built-in template.
• Codex / Claude Code Adapters — Confirm once to add managed workspace adapters, then use @promptopt ... in reopened Codex or Claude Code chats.

How to Use

Method 1: Chat Participant

Open VS Code Chat and type:

@promptopt Optimize this into an execution-ready prompt: build a customer service chatbot

The extension uses the current chat model (Copilot, Codex, Claude, etc.) to optimize your prompt.

Method 2: Editor Command

1. With selection: Select rough prompt text in your editor -> Run Prompt Optimizer Mini: Optimize Prompt from command palette (Ctrl+Shift+P / Cmd+Shift+P), or right-click and select from the context menu.
2. Without selection: Run the same command → A prompt input box appears for you to type or paste your rough prompt.
3. Choose an output action:
   • Replace Selection — Overwrite the selected text with the optimized prompt
   • Insert At Cursor — Insert the optimized prompt at the current cursor position
   • Copy To Clipboard — Copy to clipboard for use anywhere
   • Open Preview — Open as a Markdown document for review

Typical Use Cases

• Turn product requirement notes into execution-ready prompts
• Turn coding task ideas into structured coding-agent prompts
• Turn writing outlines into prompts with output format and acceptance criteria
• Turn ambiguous requests into prompts with assumptions, constraints, and verification checks

Model Router

Use the command palette:

The default auto route uses:

The local model is the official Qwen/Qwen3-1.7B-GGUF Q8_0 file. It is compact enough for common 8 GB Apple Silicon Macs and laptops with 6 GB or more VRAM.

Settings

Requirements

• VS Code 1.105.0 or later
• Python 3.x for local model bootstrap and local formatter fallback
• Network access during the first local model setup

Links

• Source Code
• Report Issues
• License (Apache-2.0)