optimizer-gpt

OptimizerGPT

Feature: gpt

OptimizerGPT refactors messy or monolithic code files into clean, well-organized, modular structures. It reads a single large source file, identifies logical separation of concerns, and rewrites it as a set of focused modules following framework conventions.

What OptimizerGPT Solves

Agent-generated code (from BackendGPT or FrontendGPT) often starts as a single large file to satisfy the LLM context window. OptimizerGPT acts as a post-processor to enforce real-world code organization, splitting routes from handlers, models from services, and utilities from application logic.

How It Works

1. OptimizerGPT receives a file path pointing to a monolithic source file
2. It sends the file contents to the LLM with instructions to identify logical modules
3. The LLM returns a restructured file layout following standard conventions for the target stack
4. OptimizerGPT writes the new module files to the workspace, preserving the original

Supported Stacks

OptimizerGPT works with any language. It applies framework-idiomatic conventions when it recognizes the stack:

Stack	Resulting Structure
FastAPI (Python)	routes/, models/, services/, utils/
Axum (Rust)	routes.rs, models.rs, handlers.rs, state.rs
React (JS)	components/, hooks/, services/, utils/
Express (JS)	routes/, controllers/, middleware/, models/

SDK Usage

use autogpt::prelude::*;

#[tokio::main]
async fn main() {
    let persona = "Senior Software Engineer";
    let behavior = "Refactor workspace/backend/main.py into a clean FastAPI project structure.";

    let agent = OptimizerGPT::new(persona, behavior).await;

    AutoGPT::default()
        .with(agents![agent])
        .build()
        .expect("Failed to build AutoGPT")
        .run()
        .await
        .unwrap();
}