whatisgithub

What is pi-grow-loop?

llblab/pi-grow-loop — explained in plain English

Analysis updated 2026-05-18

2TypeScriptAudience · developerComplexity · 3/5Setup · moderate

In one sentence

A small extension for the Pi coding assistant that makes it work through big tasks in small, visible steps, looping automatically while you stay in control and can interrupt or redirect at any time.

Mindmap

mindmap
  root((pi-grow-loop))
    What it does
      Loops in small steps
      User can interrupt
      No hidden queue
    How it works
      Worker picks one slice
      Continuation decides once
      Tool schedules next turn
    When to use
      Clear scoped tasks
      Multiple bounded pieces
      Not for single changes
    Audience
      Pi extension users
      Familiar with Pi model
    Checkpoints
      Reports changes
      Avoids repeats
      Stop proof option

Code map

Detail Auto

An interactive map of this repo's files and how they connect — its source is parsed live in your browser. Click Visualize to build it.

filefunction / class

What do people build with it?

USE CASE 1

Break a multi-part refactoring task into small steps with validation at each stage.

USE CASE 2

Work through a task with several bounded pieces where you want to review or redirect after each step.

USE CASE 3

Let an assistant loop through a scoped task automatically while keeping full control to interrupt anytime.

What is it built with?

TypeScript

How does it compare?

llblab/pi-grow-loop4ssh1/mine-sweeperachawla19/intuitcode-extension
Stars222
LanguageTypeScriptTypeScriptTypeScript
Last pushed2026-01-31
MaintenanceMaintained
Setup difficultymoderatemoderatemoderate
Complexity3/52/54/5
Audiencedevelopervibe coderdeveloper

Figures from each repo's GitHub metadata at analysis time.

How do you get it running?

Difficulty · moderate Time to first run · 30min

Requires familiarity with Pi's extension model, the README assumes you already understand how Pi extensions work.

So what is it?

pi-grow-loop is a small extension for Pi, a coding assistant. It adds a looping behavior so the assistant can keep working through a task in small, visible steps instead of doing everything at once or stopping after a single reply. The person using the assistant stays in charge and can interrupt or redirect at any point. The extension does not create a hidden queue or impose a fixed workflow. The assistant decides what matters, and the runtime only schedules the next turn. You use it when a task has a clear scope and can be broken into multiple useful pieces, each with its own validation. If the work is just one coherent change, the extension is not needed and ordinary one-shot execution is fine. The extension is for situations where an explicit outcome needs several bounded slices, each producing evidence and a natural point to stop or redirect. It is not triggered by keywords or task size alone. The system has three parts. A while-true skill acts as the worker. It looks at the current state, picks one actionable slice, executes it, validates the result, and hands off what it found. A grow-loop skill acts as the continuation layer. It locks the relevant scope, reads the worker's handoff, and decides exactly once whether another turn should run. A grow_loop tool handles the scheduling. It waits for the assistant to be idle, shows a three-second countdown, and sends a short prompt only if nothing else is pending. Each iteration has one checkpoint boundary. The worker reports what changed, what was proven, what remains, and any blockers. The continuation layer compares this checkpoint against the previous one to avoid repeated no-ops. If continuation is still safe and valuable, it schedules one more turn and ends. If not, it returns a stop proof. The runtime never blocks future calls, and the decision to continue belongs to the agent, not to a hidden state machine. The README is dense and written for people already familiar with Pi's extension model. It does not include much introductory material for newcomers. The full README is longer than what was shown.

Copy-paste prompts

Prompt 1
I have a Pi coding assistant and want it to loop through a large task in small steps. Show me how to set up the pi-grow-loop extension and configure the while-true worker and grow-loop continuation skills.
Prompt 2
Help me decide if pi-grow-loop is right for my task. The task involves breaking work into several pieces, each with its own validation. Walk me through when this extension helps and when ordinary one-shot execution is better.
Prompt 3
Explain how the pi-grow-loop checkpoint system works so I can understand what the worker reports at each step and how the continuation layer decides whether to schedule another turn or stop.
Prompt 4
I want to use pi-grow-loop for a scoped multi-step task. Help me structure my task into bounded slices where each slice produces evidence and a natural stopping point for review.

Frequently asked questions

What is pi-grow-loop?

A small extension for the Pi coding assistant that makes it work through big tasks in small, visible steps, looping automatically while you stay in control and can interrupt or redirect at any time.

What language is pi-grow-loop written in?

Mainly TypeScript. The stack also includes TypeScript.

How hard is pi-grow-loop to set up?

Setup difficulty is rated moderate, with roughly 30min to a first successful run.

Who is pi-grow-loop for?

Mainly developer.

Open on GitHub → Ask about another repo

This repo across BitVibe Labs

Verify against the repo before relying on details.