fix(slm): dashboard status from SLMM's cached roster, not a device call

"No recent check-in" read a roster field (slm_last_check) that nothing
stamps, and the live-status fetch hit /measurement-state — which sends
Measure? to the DEVICE every refresh, competing with DOD polling.

Now read SLMM's /roster once: it carries each unit's cached NL43Status
(last_seen, measurement_state) — a cache read, no device call. is_recent is
derived from last_seen (advances only on a successful monitor poll, so
staleness == not being reached) within 5 min, for all non-retired units
(benched units can still be monitored). Net: fewer device calls AND the
dashboard reflects the live monitor.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

backend/routers/slm_dashboard.py

@@ -91,29 +91,41 @@ async def get_slm_units(
 
     one_hour_ago = datetime.utcnow() - timedelta(hours=1)
     for unit in units:
+        # Legacy default from the roster field; refined from SLMM's cached status below.
         unit.is_recent = bool(unit.slm_last_check and unit.slm_last_check > one_hour_ago)
+        unit.measurement_state = None
 
     if include_measurement:
-        async def fetch_measurement_state(client: httpx.AsyncClient, unit_id: str) -> str | None:
+        # SLMM's /roster carries each unit's CACHED status (last_seen,
-            try:
+        # measurement_state) from NL43Status — a DB read on SLMM's side, NOT a device
-                response = await client.get(f"{SLMM_BASE_URL}/api/nl43/{unit_id}/measurement-state")
+        # call. The live monitor refreshes that cache ~every 1.3s, so this reflects
-                if response.status_code == 200:
+        # real monitoring without sending Measure? to the device (which the old
-                    return response.json().get("measurement_state")
+        # /measurement-state did) and competing with DOD polling. One call covers all.
-            except Exception:
+        slmm_status = {}
-                return None
+        try:
-            return None
-
-        deployed_units = [unit for unit in units if unit.deployed and not unit.retired]
-        if deployed_units:
             async with httpx.AsyncClient(timeout=3.0) as client:
-                tasks = [fetch_measurement_state(client, unit.id) for unit in deployed_units]
+                r = await client.get(f"{SLMM_BASE_URL}/api/nl43/roster")
-                results = await asyncio.gather(*tasks, return_exceptions=True)
+                if r.status_code == 200:
+                    for dev in (r.json().get("devices") or []):
+                        slmm_status[dev.get("unit_id")] = dev.get("status") or {}
+        except Exception:
+            slmm_status = {}
 
-            for unit, state in zip(deployed_units, results):
+        # "Recent" = the monitor has a fresh successful read. last_seen only advances
-                if isinstance(state, Exception):
+        # on a successful poll, so staleness == the device isn't being reached.
-                    unit.measurement_state = None
+        recent_cutoff = datetime.utcnow() - timedelta(minutes=5)
-                else:
+        for unit in units:
-                    unit.measurement_state = state
+            st = slmm_status.get(unit.id)
+            if not st:
+                continue
+            unit.measurement_state = st.get("measurement_state")
+            last_seen = st.get("last_seen")
+            if last_seen:
+                try:
+                    ls = datetime.fromisoformat(last_seen.replace("Z", ""))
+                    unit.is_recent = ls > recent_cutoff
+                except Exception:
+                    pass
 
     return templates.TemplateResponse("partials/slm_device_list.html", {
         "request": request,