phase: 06_system_page_gpu_verdicts

All criteria verified and the project is fully green.

**Verified (final pass — all 3 phase-06 tasks were already in `complete/`)**
- `classify("gpu.busy_pct", 99)` / `gpu.mem_used_pct` → `None`; `THRESHOLDS` has no `gpu.*` row; `alerts.py`/`api.py` carry no GPU verdict; live browser: deskwork GPU at **91 %** and **98 %** renders `class="hero"`, `color=rgb(22,25,31)` (plain text), no bar, no `.sev-*`.
- GPU temp on the hero `.sub` from live agents: `Radeon 8060S Graphics · VRAM 165.1 MiB / 512.0 MiB · pool 78% · 48 °C`, R9700 `· 40 °C`, gateway prints none (no `n/a`, no `0 °C`).
- Charts collapsed by default: fresh load = 3 panels / 3 toggles / **0 open / 0 `<svg>`**; expand → 3 SVGs (510 px wide), repaints across polls (legend 3.3 %→5.6 %), collapsed siblings keep their own height (803 vs 358/339).
- `gpu.busy_pct@0` still answers from `/api/system/history` (345 points) — verdict gone, data intact.

**Tests / lint / coverage**
- `bash .agents/validate.sh` → **EXIT=0**; `uv run pytest --cov --cov-fail-under=94` → **550 passed, 8 skipped, 97.91 %**; `uv run ruff check src tests agent` → clean; `uv run pyright` → 0 errors; `node --check` all scripts clean; `uv pip list --outdated` empty.
- `uv run pytest -m live --live` → **8 passed** (real agents/engines/upsd); dev server restarted, `/api/health` all sources `ok`.
- New screenshots: `data/logs/phase06-final-collapsed-default.png`, `data/logs/phase06-final-expanded-gpu-busy-uncoloured.png`.

**Completion criteria:** classify→None ✅ · pytest+94 % gate ✅ · ruff+pyright ✅ · validate.sh ✅ · screenshots of the three behaviours on live data ✅ · AGENTS.md severity section claims no GPU rows ✅

**Docs:** fixed 3 items — (1) `AGENTS.md` pytest count stale (546 → 550); (2) `app.js` `severityFor` docstring still claimed "`page-ups.js` still calls it (phase 05 converts it)", contradicting the code/tests/AGENTS.md (comment-only); rest of README/AGENTS reviewed in full (env table vs `config.py`+`.env.example`, agent routes incl. `/schema`, alerts rules, page descriptions, contracts) — consistent.

**Notable:** fixed a real pre-existing defect in `scripts/dev-server.sh` — `stop` slept 1 s and deleted the pidfile without waiting for exit, so `restart` with a browser open overlapped two servers on one SQLite file, producing a `database is locked` store refusal that `/system` rendered as a red event. Now waits (bounded 20 s) then SIGKILLs; verified a restart with live polling logs no such error/event. No live infra config was changed.

**Next pending phase:** `.agents/phases/todo/07_llm_monitor_style_pages`

AGENTS.md

@@ -160,10 +160,14 @@ Responsibilities, one line each:
   `classify()` → `ok`/`warn`/`fail`/`None`, plus `attach()`/`with_severity()` to put the verdict
   beside a value. No measurement, no verdict: `None` in → `None` out, never a green pass for
   an unknown. The verdicts travel **additively** in payloads — `cpu.severity`,
-  `memory.severity`, `disks.severity`, `gpus[].severity`, each UPS card's `severity`
-  (charge/load/runtime), `severity` inside rated LLM cells (KV pressure, classified on a
+  `memory.severity`, `disks.severity`, each UPS card's `severity` (charge/load/runtime),
+  `severity` inside rated LLM cells (KV pressure, classified on a
   percent scale because the engines publish a 0-1 ratio), and `severity` per host/fleet/UPS
   row in `/api/summary`. Pages render the class; they never compare a number to a threshold.
+  There is **no `gpu.*` row and no `gpus[].severity`** (phase 06): these are LLM servers,
+  a busy GPU is the job at every level, so the GPU card and the fleet GPU column render
+  uncoloured — like `net.*`, whose colourlessness predates it. Removing the verdict is not
+  removing the data: `gpu.busy_pct@N` / `gpu.mem_used_pct@N` are still stored and charted.
 - `alerts.py` — `build(settings, job_states, host_payloads, engine_payloads, ups_payload,
   recent_events)` → the "what is wrong right now" list `/api/summary` serves as `alerts`
   (`{id, severity, source, message, href, at}`, fail first then warn, then by source). A
@@ -250,7 +254,7 @@ cp .env.dev.example .env.dev              # dev-only: ssh tunnels + live targets
 ./scripts/dev-server.sh status            # per-source ok/DOWN
 ./scripts/dev-server.sh logs
 
-uv run pytest                             # hermetic: 543 passed + 8 live-skipped, no network
+uv run pytest                             # hermetic: 550 passed + 8 live-skipped, no network
 uv run pytest --cov                       # coverage gate (97.9%; floor 94%)
 uv run pytest -m live --live              # opt-in: real fleet / engines / upsd
 uv run ruff check src tests agent && uv run pyright
@@ -287,8 +291,9 @@ branch `main`); there is no CI on it, so `bash .agents/validate.sh` is the only
   asserts it against the three captures (discrete GPU, APU, no GPU) and over HTTP, and
   `test_every_host_serves_the_hero_card_contract` asserts the same table against the live
   agents — so a rename fails a test, and a card that shows `n/a` cannot be "fixed" by
-  inventing a field. `network` carries no `severity` on purpose (throughput has no good/bad),
-  and the contract pins its absence. `tests/ups_contract.py` is the same device for the `/ups`
+  inventing a field. `network` carries no `severity` on purpose (throughput has no good/bad)
+  and neither does a GPU row (phase 06: a busy GPU is the job on an LLM server) — the
+  contract pins both absences. `tests/ups_contract.py` is the same device for the `/ups`
   hero row and battery glyph (`HERO_FIELDS`, `UPS_SHAPES`, `assert_ups_hero_contract`):
   `tests/test_ups_contract.py` drives it from `tests/fixtures/nut_<ups>.vars` — verbatim
   `LIST UPS`/`LIST VAR`/`LIST CMD` replies from the two real UPSes — and
@@ -414,10 +419,18 @@ Colour comes from the payload's `severity` dicts and nowhere else; a missing num
 verdict — gateway's CPU temp, any host's absent GPU — prints `n/a` **uncoloured**. NET is
 one slot with two lines (`.hero-lines`, `↓` in / `↑` out) and is deliberately colourless:
 throughput has no good/bad, so `network` carries no severity. The GPU slot holds one hero
-per card (model, VRAM used/total and the honest APU `pool` view in the `.sub`), and
-everything not glance-worthy — load averages, clock, disk read/write/busy, uptime, GPU
-power/temp, filesystems, processes — folds into a collapsed "More detail" `<details>`: a
-simpler surface, not a less complete page.
+per card whose `.sub` carries model, VRAM used/total, the honest APU `pool` view and the
+card's temperature (`gpus[].temp_c` — the agent's max over that card's hwmon labels, the
+same value stored as `gpu.temp_c@N`, so the hero and the history cannot disagree), and it is
+colourless for the same reason (phase 06): a 95 %-busy card on an LLM server is the job, so
+the hero carries no `severity` and no utilisation bar — a bar under a verdictless number
+would be a verdict by shape. A card whose driver exposes no temperature input prints **no**
+temperature in that footnote rather than `n/a` or `0 °C`: "where you can" is the operator's
+wording, and `n/a` is reserved for numbers the card is supposed to have. Everything not
+glance-worthy — load averages, clock, disk read/write/busy, uptime, GPU power and the
+per-label temp breakdown, filesystems, processes — folds into a collapsed "More detail"
+`<details>`: a simpler surface, not a less complete page (the table keeps its Temp column;
+this was a promotion, not a de-duplication).
 
 A `Δ<window>` label is a measurement, never a config value: `sources/host.py` sends the window
 the estimator actually used as the payload's `window_seconds`, and `/system`'s footnote prints
@@ -453,14 +466,28 @@ panel and the host/metric detail picker are deleted, along with the `#chart-*` /
 the `select` / `label.field` styling only those pickers used. Each panel stacks its own
 `.host-charts` — utilisation (CPU, memory, worst disk), network (in/out), then GPU busy — where
 the GPU slot is created and destroyed with `host.gpus`, so gateway shows neither a phantom card
-nor a hole. **One `/api/system/history` request per host per refresh**
+nor a hole. **That stack is collapsed behind a per-host `Charts` `<details>`** (phase 06: "only
+visible if expanded… there will be lots of hosts"): the default `/system` view is the hero row,
+the summary carries the window label the stack will draw (the chips themselves stay in the page
+head, one control per page), and `drawHostCharts()` is the single place that decides whether to
+paint — so nothing is built or drawn while the stack is shut (`LineChart` inside a
+`display:none` container measures 0 px, and rebuilding three charts per host every poll for
+plots nobody sees is the waste this removes). The **history request is deliberately unchanged**:
+one per host per refresh, naming every series the panel could draw, and its response is cached
+per host so the first expand paints from data in hand rather than waiting a poll; a cached
+response from another window is re-fetched instead of being drawn under the wrong label.
+Expanding one panel must not stretch its collapsed siblings into blank cards, so `#hosts`
+carries the existing `.grid.items-start`. **One `/api/system/history` request per host per refresh**
 (`hosts=<that host>&metrics=<that panel's series>&range=…`, pinned by
-`test_every_system_chart_is_drawn_inside_its_own_host_panel`): unknown metrics come back as
+`test_every_system_chart_is_drawn_inside_its_own_host_panel` and the collapse by
+`test_system_history_charts_are_collapsed_behind_a_toggle`): unknown metrics come back as
 empty series on purpose, so a panel asks for every series it could draw and asks once. A legend
 here names metrics (or one GPU's model), never hosts — the panel heading already says who this
 is, and a legend that grew a second host is the bug this replaced. A host that stops reporting
-loses its panel *and* its charts (`pruneHosts`), never a chart of nothing. Per-core visuals are
-gone from `/system` and `renderCores` is **deleted** from `charts.js` with its last caller.
+loses its panel, its `Charts` toggle *and* its charts (`pruneHosts` deletes the card plus the
+payload/chart/history maps), never a chart of nothing or an orphan control in the grid. Per-core
+visuals are gone from `/system` and `renderCores` is **deleted** from `charts.js` with its last
+caller.
 `drawHostChart()` shades a chart only when one series has points — two or three filled areas
 stack into one block and bury the quieter line, which is how a idle CPU would hide behind a busy
 memory plot; the `cpu.usage_pct@coreN` series is still stored and still queryable from