# Changelog — @zerotal/monitor All notable changes to this package are documented here. The format is based on [Keep a Changelog](https://keepachangelog.com/); this package follows the Zerotal monorepo's unified versioning. **Maturity: `stable`** ## [Unreleased] ## [1.10.0] — 2026-08-30 ### Fixed - **`MonitorStore` options no longer overwrite their own defaults with `undefined`.** The constructor applied `?? 100`-style defaults and then spread `...opts` **after** them, and object spread copies own properties even when their value is `undefined`. So `new MonitorStore({ retentionDays: cfg.retentionDays })` with an unset config put `undefined` straight back over the 7, and `prune()` computed a cutoff of `Date.now() - undefined * DAY_MS` — `NaN`, which prunes nothing and reports nothing. Spreading last is what made every `??` in that constructor decorative. The caller's options now go first and the defaults fill the gaps, which is the ordering `Socket`'s constructor already documents at length for the same reason. ## [1.9.0] — 2026-08-29 ### Added - **The snapshot shape is documented.** **Export JSON** hands you a `MonitorSnapshot` — the whole panel for a range, as data — and none of the forty-odd row types it is built from was named anywhere. `docs/monitor.md` now maps every field to its row type, which is what you need to post one somewhere, diff two of them, or drive an alerting integration off it rather than off the built-in thresholds. - **Section contributions can be typed.** `MonitorSection`, `MonitorSectionData`, `MonitorStat`, `MonitorTable`, `MonitorTableColumn` and `MonitorRow` are documented. Declaring the host's shape locally stays the recommended path — it is what keeps a contributing package free of any dependency on the monitor — but a large `resolve` is better served by a typo in a column key being a compile error than a blank cell. ### Changed - **INTERNAL: `MonitorLayout`, `MonitorPage` and `MonitorPanelHost` are marked `@internal`.** The panel's own shell. Still exported, still working; a contributor pushes a section into the host binding rather than constructing any of them. ## [1.6.0] — 2026-08-15 ### Fixed - **System info reported the runtime mode as the environment.** The panel showed `web` rather than `production` or `staging`, because it read `APP_ENV` after `setAppEnv()` had overwritten it. It reads `deployEnv()` now. ## [1.5.1] — 2026-08-15 ### Fixed - **Open-by-default access in development never applied.** The predicate read `APP_ENV`, which holds the runtime mode once the app has booted, so the panel demanded an explicit `auth` predicate even on a developer's machine. It asks `devSurfacesEnabled()` now — and still requires one anywhere that is not explicitly a development environment. ## [1.5.0] — 2026-08-15 ### Added - **Tests for the three riskiest untested modules**, chosen by consequence rather than by coverage percentage: - **`MonitorAuthMiddleware`** — the only thing between the public internet and a page showing recent requests, their users, their IPs and the SQL the app ran. Pinned against the two ways a gate fails open without anyone noticing: a forgotten `await` on an async predicate (a promise is always truthy, so a denying predicate would admit everyone) and a predicate that throws. Also pins that the _default_ predicate survives config merging as a callable — `deepMerge` cannot carry a function through, and the reattach only fires when the caller supplied one. - **`RingBuffer` / `TimeWindow` / `percentile`** — the bounded structures that keep monitoring memory from growing without limit. A buffer that quietly stops evicting is a memory leak in the observability layer, and an off-by-one percentile misreports latency in the direction nobody checks. - **`renderPrometheus`** — the one output a machine parses rather than a person reads, where a scraper rejects a malformed response without telling the application. Covers label escaping for route paths carrying a quote, a backslash or a newline, and metric-name sanitisation for display-string gauge labels — all of which come from application data, so none are hypothetical. Monitor's suite goes from 49 tests to 81. ### Changed - **Maturity is now `stable`** — the public API follows SemVer strictly for the rest of the 1.x line. Both gates are closed: `@zerotal/flow` and `@zerotal/flow-ui` (which the panel renders through) are themselves stable now, and the coverage gap is closed where it mattered. Worth stating plainly, since the surface looks larger than it is: of 685 lines of public API, 65 entries are type definitions describing the snapshot shape and only 11 are callable entry points — every one of which is documented. ## [1.1.0] — 2026-08-08 ### Fixed - **The reported framework version follows the release.** Two hardcoded version strings in `sources/live.ts` and `sources/system.ts` meant a monitored app reported whatever version was current when those lines were last edited. They now read the package's own manifest, which the lockstep release keeps correct. ## [1.0.3] — 2026-08-07 ### Changed - Re-released from a rebuilt repository so the build provenance resolves. The 1.0.2 attestation names a repository that was renamed away, which leaves the signature valid but the trace back to source dangling. No code changed. ## [1.0.2] — 2026-08-06 ### Fixed - Reported framework version was stale again; it now matches the release. ## [1.0.1] — 2026-08-06 ### Fixed - Reported framework version was pinned at the pre-release `1.1.0` in the health report and the system-meta fallback, so a monitored app advertised a version that was never published. ## [1.0.0] — 2026-08-05 _First public release._ ### Notes - Conforms to the Zerotal package conventions (provider in `src/provider/`, PascalCase config factory, `ZerotalError`-based errors, test coverage).