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run ruff check src tests agent` → clean; `uv run pyright` → 0 errors; `node --check` all scripts clean; `uv pip list --outdated` empty.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`

tests/card_contract.py

@@ -16,7 +16,7 @@ cpu %                        ``cpu.usage_pct`` (+ ``cpu.severity.usage_pct``)
 mem %                        ``memory.used_pct`` (+ ``memory.severity.used_pct``)
 cpu temp                     ``cpu.temp_c`` (+ ``cpu.severity.temp_c``)
 disk usage %                 ``disks.worst_used_pct`` + ``disks.worst_mount`` (+ verdict)
-graphics card                ``gpus[].model/busy_pct/vram_*/mem_used_pct`` (+ verdict)
+graphics card                ``gpus[].model/busy_pct/vram_*/mem_used_pct/temp_c`` (no verdict)
 network down/up              ``network.rx_bytes_per_s`` / ``tx_bytes_per_s`` (no verdict)
 ===========================  ==========================================================
 
@@ -27,7 +27,16 @@ into its fleet row, so the card and the overview cannot drift apart.
 Deliberately *not* card metrics (legibility, not missing data — the series stay in SQLite):
 per-core usage (``cpu.usage_pct@coreN``), swap, the process table, the raw sensor list, and
 per-interface rows. Network carries no ``severity`` on purpose; throughput has no good/bad,
-so ``severity.THRESHOLDS`` has no ``net.*`` row.
+so ``severity.THRESHOLDS`` has no ``net.*`` row. GPU rows carry no ``severity`` either
+(phase 06): these are LLM servers, a 95 %-busy card is the job and not a fault, so
+``THRESHOLDS`` has no ``gpu.*`` row and the card prints its numbers uncoloured at every
+utilisation — the *absence* of the dict is pinned below, the way ``network``'s is.
+
+``gpus[].temp_c`` is on the card because the number worth reading *while* a card is busy is
+how hot it is (phase 06, TODO.md L9). It is the agent's **max over that card's hwmon
+labels** — the same max the agent already computes and the same value stored as
+``gpu.temp_c@N`` — so the hero prints a measurement rather than a new field, and the value
+may honestly be ``None`` on a box whose driver exposes no temperature input.
 """
 from __future__ import annotations
 
@@ -54,8 +63,9 @@ GPU_METRICS: tuple[str, ...] = (
     "vram_used_pct",
     # the honest APU view: vram + GTT, because a Strix Halo keeps the model in GTT
     "mem_used_pct",
-    "severity.busy_pct",
-    "severity.mem_used_pct",
+    # the hero's footnote: the agent's max over this card's hwmon labels (None is legal —
+    # a driver with no temperature input must not be rendered as 0 °C)
+    "temp_c",
 )
 
 #: `network` must never grow a `severity` dict (see the module docstring).
@@ -95,6 +105,10 @@ def assert_card_contract(payload: Any) -> None:
     for gpu in gpus:
         for path in GPU_METRICS:
             dig(gpu, path)
+        assert "severity" not in gpu, (
+            "gpus[].severity would grade utilisation on a fleet built to be busy — "
+            "severity.THRESHOLDS has no gpu.* row and a busy GPU is the job, not a fault"
+        )
     for section in SEVERITY_FREE_SECTIONS:
         block = payload.get(section)
         assert isinstance(block, dict), f"{section} must be a dict"