Ralph Wiggum
Long-running iterative development loops with optional self-reflection and parallel loop support.
What it does
Pi-ralph-wiggum implements Geoffrey Huntley's ralph-loop pattern for Pi. It runs a named, persistent development loop that works toward a goal across many iterations — reading the current state of the codebase, making progress, committing, and looping until the goal is reached or you stop it.
Multiple named loops can run in parallel in the same repo, each tracked independently. Optional self-reflection intervals let the loop pause and assess its own progress before continuing. Loops can run in-session or handed off to tmux for background execution. The .ralph directory stores loop state so progress is never lost if a session ends.
Why it's included
Some development tasks are too large for a single context window — refactors, migrations, feature implementations that span many files. Ralph Wiggum gives Pi a structured way to work on these over time: start a loop, check back later, resume if interrupted. It's the pattern for work that can't fit in one sitting.
Commands
| Command | What it does |
|---|---|
/ralph start <name|path> |
Start a new named loop |
/ralph resume <name> |
Resume a paused loop |
/ralph stop |
Pause the current loop |
/ralph-stop |
Stop active loop (idle state only) |
/ralph status |
Show all active and paused loops |
/ralph list --archived |
Show archived loops |
/ralph archive <name> |
Move a loop to archive |
/ralph clean [--all] |
Clean up completed loops |
/ralph cancel <name> |
Delete a loop entirely |
/ralph nuke [--yes] |
Delete all .ralph data |