architect-gpt

ArchitectGPT

Feature: gpt CLI: autogpt arch

ArchitectGPT translates a high-level project description into a concrete system architecture. It determines the technologies, frameworks, and communication patterns needed, then generates an executable Python script using the diagrams library that renders a PNG architecture diagram.

What ArchitectGPT Solves

Software architecture decisions are typically made informally and documented inconsistently. ArchitectGPT externalizes this process: given a goal, it produces a versioned, reproducible visual diagram that serves as the authoritative architectural reference for all other agents.

How It Works

1. ArchitectGPT receives a project goal from ManagerGPT or directly from the user
2. It calls the configured LLM to generate a Python script using the diagrams library
3. The script is written to workspace/architect/diagram.py
4. A Python virtual environment is set up with diagrams installed in workspace/architect/.venv/
5. Running the script produces a PNG in that directory

CLI Usage

autogpt arch

Example session:

> Generate a Kubernetes architecture for a web app with Prometheus and Grafana monitoring.

After the agent completes:

# Render the diagram
./workspace/architect/.venv/bin/python ./workspace/architect/diagram.py
# ➜  simple_web_application_on_kubernetes.png created

SDK Usage

use autogpt::prelude::*;

#[tokio::main]
async fn main() {
    let persona = "Lead UX/UI Designer";
    let behavior = r#"Generate a diagram for a simple web application running on Kubernetes.
    It consists of a single Deployment with 2 replicas, a Service to expose the Deployment,
    and an Ingress to route external traffic. Also include a basic monitoring setup
    with Prometheus and Grafana."#;

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

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