Future / Longer-Term Ideas

Not committed work — directions worth considering.

• In-process generation. Today the web UI spawns the CLI (node . --flags) for every generation and polls a separate progress server. Now that everything is ESM with a shared common.js, the server could call run()/upscale() in-process (with a job queue + streamed progress), removing the child-process round-trip and the second HTTP server. Bigger change; weigh the isolation benefit of the current design.
• Real test suite + CI. A committed npm test, a small fixture output/, mocked fetch for the WebUI contract, and a GitHub Actions workflow running lint + tests on Node 24. See testing.md.
• Frontend modernization. Convert web/frontend/* from classic multi-<script> jQuery code to ES modules (and maybe drop jQuery/lodash-in-the-browser), with proper linting.
• TypeScript or JSDoc types. The settings objects and the dynamic-prompt/prompt-module contracts are implicit; typing them would make the plugin API self-documenting and catch shape mistakes.
• Pluggable backends. The tool is hard-wired to the Stable Diffusion WebUI --api. A thin adapter layer could support other local backends.
• Packaging. A proper bin entry / npx-able CLI, and clearer separation of user data (output/, user-settings.json) from the app.