explaingit

smol-ai/developer

★ 12,186PythonAudience · vibe coderComplexity · 2/5Setup · moderate

TLDR

A Python tool that turns a plain-English description of what you want to build into a complete working codebase, describe your app, run the tool, and it writes all the files using GPT-4.

Mindmap

mindmap
  root((repo))
    What it does
      Text to codebase
      Full file generation
      Human in the loop
    Modes
      CLI Git Repo mode
      Python library mode
      HTTP API mode
    Tech
      Python
      GPT-4 backend
    Use cases
      App prototyping
      Chrome extensions
      Embedded codegen
    Audience
      Vibe coders
      No-code builders
Click or tap to explore — scroll the page freely

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

Things people build with this

USE CASE 1

Generate a complete starter codebase for a Chrome extension by describing what it should do in a short text prompt.

USE CASE 2

Iterate on generated code by pasting error messages back into your description and regenerating until the app runs.

USE CASE 3

Embed smol developer as a Python library into your own product to let users generate code from text descriptions.

Tech stack

PythonGPT-4

Getting it running

Difficulty · moderate Time to first run · 30min

Requires an OpenAI API key with GPT-4 access, which is a paid service.

In plain English

Smol developer is a Python tool that turns a written description of software into a working codebase. You describe what you want in plain text (for example, "a Tic Tac Toe game in HTML, JS, and CSS"), run the tool, and it uses a language model to generate all the files needed to build that thing. The intended workflow is human-in-the-loop. You write a prompt describing the app, the tool generates code, you run and review it, then paste any error messages or missing details back into the prompt and generate again. The README calls this "engineering with prompts, rather than prompt engineering." Once the output is close enough, you can take over the codebase yourself. The tool works in three modes. In Git Repo mode, you run it from the command line, passing your description as a string or from a markdown file. In Library mode, you install it as a Python package and call individual functions inside your own application: one function returns a coding plan, another returns a list of files to create, and a third generates the source code for each file. This lets you build smol developer into your own product. In API mode, the tool exposes an HTTP server that accepts tasks and executes them step by step following the Agent Protocol specification. By default the tool uses GPT-4 to generate code, though GPT-3.5-turbo is available as a faster and cheaper alternative. The README includes worked examples showing the tool generating a Chrome browser extension, a ChatGPT plugin, and a command-line Python application from short text prompts.

Copy-paste prompts

Prompt 1
Use smol developer to generate a simple TODO app in React from this description: a single-page app where I can add, check off, and delete tasks with data saved in localStorage.
Prompt 2
Run smol developer on my prompt file to generate a Python CLI tool that takes a folder path and resizes all images inside to a given pixel width.
Prompt 3
Set up smol developer in library mode and show me the Python code to call its functions to generate a file list and source code from a description string.
Open on GitHub → Explain another repo

← smol-ai on gitmyhub — every repo by this author, as a profile.

Verify against the repo before relying on details.