Port printer monitoring to backend-agnostic printer.Printer interface

[kennhung]

Summary

• Make moonraker.Monitor implement the new internal/printer.Printer (and printer.Thumbnailer) interface, dropping its duplicate PrinterState/MonitorConfig types in favor of the shared ones.
• Switch main.go, internal/web, and internal/controller to depend on map[string]printer.Printer instead of the concrete *moonraker.Monitor, so a future second backend can be added without touching these layers.
• Rework the web API's Printer DTO to embed the neutral printer.Job/printer.PrinterState/printer.ErrorInfo types instead of Moonraker-raw structs; regenerate Swagger/OpenAPI + the frontend axios client accordingly.
• Update frontend/src/types.ts, utils.ts, PrinterCard.tsx, App.tsx to match the new DTO shape (drops the loadedFile/latestJob UUID-matching hack in favor of job.status === "in_progress").

Test plan

• go build ./... and go vet ./...
• npx tsc -b && vite build in frontend/
• npm run lint in frontend/ (only pre-existing generated-client warnings remain)
• Manual smoke test of /api/v1/printers and the dashboard against a live printer (blocked locally by another dev server already bound to :8080)

🤖 Generated with Claude Code

[kennhung]

Review

Overview: this PR extracts a backend-agnostic printer.Printer interface (plus optional Thumbnailer/RawReporter capability interfaces) into a new internal/printer package, and migrates moonraker.Monitor, main.go, internal/web, and internal/controller to depend on that interface instead of the concrete Moonraker type. The web DTO and frontend types are reworked to use the new neutral Job/PrinterState/ErrorInfo shapes. go build/go vet/gofmt are all clean, and there's no leftover Moonraker-specific leakage in web/controller/main.

Issues found

1. EstimatedRemaining (remaining_sec) is now computed incorrectly — regression (internal/moonraker/moonraker_monitor.go:141-151)

printDuration := printer.Seconds(m.printerObjects.PrintStats.GetPrintDuration().Seconds())
totalDuration := printer.Seconds(m.printerObjects.PrintStats.GetTotalDuration().Seconds())
...
remaining := totalDuration - printDuration

Klipper's print_stats.total_duration is wall-clock time since the job started including time spent paused, while print_duration excludes paused time. So total_duration - print_duration is the accumulated pause duration, not a completion ETA — for an unpaused print this is ~0 regardless of actual progress.

The previous code computed a real ETA from the slicer's estimated_time in gcode metadata, or extrapolated from virtualSD.progress: totalTime = printDuration / progress; estRemainSec = totalTime - printDuration. That logic isn't reproduced in Monitor.Job(), so the dashboard will show ~0s remaining for actively-printing jobs. Worth fixing before merge since it's a silent, user-facing regression rather than a pure refactor.

2. PrinterObjects(), LatestJob(), LoadedFile() are now dead code outside the package (internal/moonraker/moonraker_monitor.go:97-107)

Nothing outside internal/moonraker calls these anymore since web/controller now go through printer.Printer. Worth deleting if there's no plan to use them.

Minor / non-blocking

• moonraker_monitor.go:371-378: wire state "startup" used to map to a distinct KlippyStartup (shown as "info"/blue in the frontend); it now collapses into printer.Unknown (shown as "secondary"/gray). Likely an intentional simplification given the smaller printer.PrinterState enum, but it's a small user-visible change — worth confirming it's intended.
• types.ts:262: errorMessage: errorDetail?.message ?? printer.message — when a backend sets both Message() and ErrorDetail.Message, the frontend now prefers the specific error message over the generic status text. Worth double-checking this precedence matches intent.
• No unit test covers Monitor.Job()'s remaining-time logic — a test on the paused/non-paused/no-metadata cases would have caught issue Bump ejs from 3.1.9 to 3.1.10 in /frontend #1.
• PR description notes manual smoke testing was blocked locally (port conflict already in use) — worth doing before merge since Bump ejs from 3.1.9 to 3.1.10 in /frontend #1 produces plausible-but-wrong values rather than a crash, so it wouldn't surface via go vet/build.

No security concerns; import paths and internal/ layout match project conventions.

[kennhung]

Thanks for the review — pushed 1dc2b83.

Fixed:

1. EstimatedRemaining regression — restored the original two-path estimate from main's frontend logic: slicer's estimated_time * (1 - progress) when gcode metadata has it, else extrapolate print_duration/progress - print_duration. Kept TotalDuration as print_stats.total_duration (the separate display-only "elapsed since start incl. pauses" field) — that part wasn't actually wrong, only the ETA derivation from it was.
2. Removed the now-dead PrinterObjects()/LatestJob()/LoadedFile() — confirmed zero callers repo-wide.

Confirmed intentional (no change):

• Klippy startup → Unknown collapse — matches controller.Connector's pre-existing status mapping (startup already fell to the default/Unknown case there); the frontend color shift is an accepted side effect of unifying the enum.
• errorMessage: errorDetail?.message ?? printer.message — deliberately prefers the specific error detail over the generic status message.

Still haven't been able to smoke-test locally (port 8080 occupied by an unrelated already-running dev server), so this is verified by go build/go vet plus manually tracing the three Job() cases (slicer estimate, extrapolation fallback, neither available) rather than a live run.