/ralph-wiggum

Overview

A Pi extension for long-running iterative development. Based on Geoffrey Huntley's Ralph Wiggum approach – a "flat" iteration loop without subagents where the model works through a checklist, reflecting periodically.

Features

• Self-starting – Ask Pi to set up and run loops automatically
• Multiple loops – Run parallel loops in the same repo
• Periodic reflection – Optional self-reflection to avoid grinding through wrong instructions
• Pauseable – Hit Esc to pause, resume later

Usage

Just ask Pi to set up a ralph-wiggum loop, or use commands directly:

# Start a new loop
/ralph start <name>

# Resume a paused loop
/ralph resume <name>

# Check status of all loops
/ralph status

# Stop current loop
/ralph stop

How it works

1. Pi sets up .ralph/<name>.md with goals and a checklist
2. You specify items per iteration and optional reflection frequency
3. Pi runs through the checklist, calling ralph_done after each iteration
4. Loop continues until complete, max iterations, or you pause

Installation

npm install @tmustier/pi-ralph-wiggum

Or via git:

pi install git:github.com/tmustier/pi-extensions

Why?

Long tasks dilute model attention. This keeps the model focused on a small chunk at a time, with clear checkpoints and the ability to pause and resume.