Story 2 — Add rich interactive views to orium doctor --watch

[NoaLand]

Title

Add interactive runtime observability views to orium doctor --watch

Story

As a Scriptorium developer,
I want to observe runtime health and performance interactively over time,
so that I can understand performance trends, identify regressions, and locate bottlenecks across individual components.

Build on the diagnostic and telemetry capabilities introduced by orium doctor.

orium doctor --watch should provide a live TUI/dashboard rather than a one-shot diagnostic report.

Overview View

Provide an overall system view containing:

┌─ Scriptorium Doctor ───────────────────────────────────────┐
│ Overall  ● HEALTHY                         Uptime 01:42:17 │
├─ Processes ──────────────────┬─ IPC ───────────────────────┤
│ ● brush                      │ brush → inkstone   0.4 ms  │
│ ● inkstone                   │ inkstone → ink     0.3 ms  │
│ ● ink                        │                            │
│ ● dictionary                 │                            │
├─ Performance ────────────────┴────────────────────────────┤
│ avg     3.1 ms     p50 2.8 ms                            │
│ p95     5.4 ms     p99 8.7 ms                            │
├─ Recent Events ───────────────────────────────────────────┤
│ brush       composition changed                           │
│ inkstone    candidates generated                          │
│ ink         candidate window updated                      │
└───────────────────────────────────────────────────────────┘

Component Views

Allow individual runtime components to have dedicated views, for example:

orium doctor --watch inkstone
orium doctor --watch brush
orium doctor --watch dictionary

A component view may expose:

Requests       12,481
Average         1.42 ms
P50             1.17 ms
P95             2.83 ms
P99             5.21 ms
Errors             3

Historical Performance

Keep a bounded history of runtime telemetry so that performance can be visualised over time.

For example:

Latency

8ms │                  ╭╮
6ms │             ╭╮   ││
4ms │     ╭─╮    ╭╯╰───╯╰╮
2ms │─────╯ ╰────╯        ╰────
    └───────────────────────────
      -60s                  now

The historical view should make performance regressions and transient latency spikes easier to identify.

Views

The TUI may eventually provide views such as:

[1] Overview
[2] Latency
[3] IPC
[4] Errors
[5] Events

The exact interaction model is intentionally left open until the underlying telemetry capabilities are mature enough to drive the design.

Acceptance Criteria

• orium doctor --watch continuously updates runtime information.
• Overall system health can be observed without leaving the dashboard.
• Individual components can be inspected independently.
• Recent performance history can be visualised.
• Common latency statistics such as average, p50, p95 and p99 can be displayed.
• IPC latency can be distinguished from component processing latency where instrumentation permits.
• Recent errors/events can be correlated with performance behaviour.
• The dashboard consumes the diagnostic/telemetry model from Phase 1 rather than introducing a separate observability mechanism.
• The UI remains independent from the runtime components being observed.

Non-goals

This phase does not need to become a general-purpose monitoring platform.

The goal is specifically:

Make Scriptorium observable enough that runtime and performance problems can be investigated instead of guessed.