Plan

The first milestone is a demo app and an example cast, so a full run works on a laptop. Packaging and cloud adapters follow.

Milestones

1. Scaffold. Community files, CI checks and docs. Done.
2. Engine. The harness and reviewer, plus the results board. The dummyrun command offers run, review and board. The board is static HTML plus JSON. The engine language is chosen before the port, and it must support WebDriver BiDi for Chrome and Firefox.
3. Demo. A small open-source app with real roles, such as editor, author and reviewer. A cast of five or six personas runs against it and ends on a board with real findings.
4. Packaging. A Helm chart in deploy/helm/dummyrun and a compose file in deploy/compose, with a values file per cloud. Packages are published to the registries for the chosen language.
5. Clouds. Google Cloud (Vertex AI in europe-west2, Cloud Run jobs) and Azure (Azure AI Foundry in UK South, Container Apps). Every release runs the demo cast on both.
6. Targets. Adapters for plain URLs and GitHub Deployments, then Vercel, Netlify and Cloudflare Pages.

Words we use

A cast is a set of personas. Each persona has scenarios. A preview is a copy of an app built for one pull request. Findings land on the results board. The crash-test theme calls the cast “dummies” and gives us a mascot. The cautions are in NAMING.md.

Decisions