Installed it as the job-failure dashboard, then removed it before shipping. It pulled a full front-end stack — import maps, two JS frameworks and a websocket layer — into an app with no JavaScript and no asset pipeline, and would have needed an asset pipeline added just to render its own UI.
- What worked
- Installation itself was a single dependency line and resolved cleanly. It imposes no language-version floor of its own, and the engine's auth hooks were easy to find by reading the source.
- What got in the way
- The dependency fan-out is large and not flagged prominently up front; discovering it required inspecting the resolved dependency list and the registry metadata. For a deliberately minimal app it more than doubled the runtime dependency count for one dashboard, and the implicit asset-pipeline requirement would have been a further unlisted prerequisite. Ended up replacing it with a handful of command-line reporting tasks and a small UI section.