SvelteKit calling a separately deployed Rust service through a typed, authenticated server-side client — live status, real latency, contract validation and honest failure handling.
This page never talks to the service itself. Every button posts to SvelteKit, whose server calls the runtime-lab service — a separate Rust process in its own container — with a credential the browser never sees. Every figure below is measured on that call.
browser → +page.server.ts → client.ts → runtime-lab (Rust, separate container)
A target is a place the service runs, configured on the server and chosen here by name — never by URL. Nothing is called until you press Check.
The service in its own Podman container beside the dev server.
The same image on a hosted free tier, reached over HTTPS. Free hosts sleep when idle; the first call may take about a minute.
Send a preset order to the rules engine. The page sends only the preset's name; the server owns the request body. A refusal is the engine's answer, not an error.
Deliberate failures. Each drill changes one thing on our side and shows the service's real response — the service has no switch to fake a failure, on any host.
Calls made from this browser tab, newest first. Measured on the server for each call; nothing is stored and nothing is estimated. Times in milliseconds; a dash means not observed.
No calls yet.
Think this pattern could be better? Tell us how.
Leave feedback