Getting started

Prerequisites

• Node.js 22.13 or newer
• pnpm 11
• An OpenAI-compatible chat-completions endpoint and model that support formal function tools

The default configuration uses the OpenCode Zen gateway and the muse-spark-1.3-contributor-free model through a continuous agentic conversation. Set provider credentials or override the endpoint in .env.local when needed.

Install and run

pnpm install
cp .env.example .env.local
pnpm run dev

Open http://localhost:3000.

The values in .env.example are usable defaults. Change the provider URL, model, timeout, or optional key in .env.local if you operate a different compatible endpoint. See environment variables for the full reference.

Run the quality gate

pnpm run check

The gate runs linting, strict TypeScript checks, contract and utility tests, the production application build, and the VitePress documentation build.

Run the documentation site

pnpm run docs:dev

The documentation uses the GitHub Pages base path /requireganizer/. VitePress prints the exact local URL when the server starts.

First project

1. Create a project. Optionally give a starting intent to draft the product overview once; skip to write the overview yourself. The intent is kept as revision 0 provenance.
2. Review the product overview. Approve each item; generated output is draft until you do.
3. Continue through each stage in order. Request change or Approve each item; the left bar is the signature.
4. Approve Boundary Design, the Implementation Profile, and each formal contract bundle after review, the same way: item by item.
5. Review any downstream impact before applying a new approved-artifact revision.
6. Generate Project Setup and then generate each automated test from its structured case.

The model calls communicate when required scope is missing or contradictory. The application exposes that as a needs-input message. Answer the question by improving the project data, then run the stage again.