Ralph Wiggum extension for Gemini CLI implements iterative, self-referential development loops. Named after Ralph Wiggum from The Simpsons, it embodies persistent iteration despite setbacks.

Ralph is a development methodology based on continuous AI agent loops. The technique creates a self-referential feedback loop where Gemini iteratively improves its work until completion.

Install the Ralph extension by running (requires Gemini CLI v0.4.0 or newer):

gemini extensions install https://github.com/AsyncFuncAI/ralph-wiggum-extension --auto-update

The --auto-update is optional: if specified, it will update to new versions as they are released.

/ralph-loop Build a REST API for todos --completion-promise 'DONE' --max-iterations 20

• --max-iterations <n> - Stop after N iterations (default: unlimited)
• --completion-promise '<text>' - Phrase that signals completion (quote multi-word phrases)

/cancel-ralph

# Basic loop with iteration limit
/ralph-loop Fix the auth bug --max-iterations 10

# Loop with completion promise
/ralph-loop --completion-promise 'ALL TESTS PASSING' Add unit tests

# Unlimited loop (use with caution)
/ralph-loop Refactor the cache layer

1. /ralph-loop creates a state file at .gemini/ralph-loop.local.md
2. Gemini works on the task iteratively
3. Each iteration, progress is made and tracked
4. Loop ends when:
   • Max iterations reached, OR
   • Completion promise is genuinely true and output

Monitor progress:

cat .gemini/ralph-loop.local.md

• Iteration > Perfection: Don't aim for perfect on first try
• Failures Are Data: Use them to improve
• Persistence Wins: Keep trying until success
• Trust the Process: Don't circumvent with false completion

• Original Ralph Technique
• Gemini CLI
• Gemini CLI Extensions

Apache License 2.0