TINY(1)General Commands ManualTINY(1)

Web page

The fleet screen is a terminal. When a team wants to look at the same thing — or you want the overnight picture on a second monitor — the web add-on serves it as a page.

Switch it on in tiny☰ namespace settings, then:

kubectl port-forward svc/tiny-web 8080

No Ingress ships with it. Reaching the page is exactly the cluster access you already have; there is no new front door and no auth story to get wrong.

What it shows

  • The fleet, as the TUI sees it: phase, age, self-reported CPU and memory, each agent's live title.
  • Blast radius per session — the files it changed, lines added and removed, its branch, and any bundles waiting in the outbox. Read live from the workspace with git, not from parsed logs.
  • Collisions — files touched by more than one session, named with the sessions touching them. The question a fleet of agents raises and nothing else answers.
  • Questions waiting on a human, each with the exact command to answer it.

The same reading powers tiny diff, so the page and the CLI cannot disagree.

Read-only, deliberately

Its ServiceAccount has get, list and watch — plus pods/exec for the git reads — and no write verb anywhere.

That is a design decision, not an unfinished feature. Answering a question in tiny performs the action with the answering human's credentials, which is what makes the audit log name a person. A browser button would act as the web server's own account instead, quietly breaking that. So the page prints tiny answer q-xxxxx <answer> and lets you run it.