feat(db): table-health readout — per-table bloat metrics in admin card

You can't decide what to maintain without seeing what's bloating. Adds a
read-only health panel driven by Postgres' own statistics views.

- services/db_maintenance.py: get_table_health() queries pg_stat_user_tables +
  pg_total_relation_size + pg_database_size — per-table size, live/dead tuples,
  dead-tuple ratio (the bloat signal), and last (auto)vacuum/(auto)analyze.
- routes/admin.py: admin-only GET /api/admin/db-maintenance/health.
- SettingsView.vue: 'Table health' table in the maintenance card, all tables
  sorted by dead tuples, rows >=20% dead-ratio flagged; total DB size shown;
  refreshes after a Run-now so the dead-tuple drop is visible.
- Tests: health row/size shaping + null-timestamp passthrough; route + service
  surface.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

frontend/src/views/SettingsView.vue

@@ -160,6 +160,23 @@ const dbMaintLastRun = ref<DbMaintRun | null>(null);
 const savingDbMaint = ref(false);
 const dbMaintSaved = ref(false);
 const runningDbMaint = ref(false);
+interface DbTableHealth {
+  table: string; live: number; dead: number; dead_pct: number;
+  total_bytes: number; mod_since_analyze: number;
+  last_vacuum: string | null; last_analyze: string | null;
+}
+interface DbHealth { db_bytes: number; tables: DbTableHealth[] }
+const dbHealth = ref<DbHealth | null>(null);
+const loadingHealth = ref(false);
+const DEAD_PCT_WARN = 20; // dead-tuple ratio above this = autovacuum falling behind
+
+function formatBytes(n: number): string {
+  if (n < 1024) return `${n} B`;
+  const units = ["KB", "MB", "GB", "TB"];
+  let v = n / 1024, i = 0;
+  while (v >= 1024 && i < units.length - 1) { v /= 1024; i++; }
+  return `${v.toFixed(v >= 10 || i === 0 ? 0 : 1)} ${units[i]}`;
+}
 const pluginInstallCommands = computed(() => {
   const mkt = (pluginMarketplaceUrl.value || '').trim()
     || serverMarketplaceUrl.value
@@ -468,6 +485,7 @@ onMounted(async () => {
     } catch {
       // leave defaults
     }
+    await loadDbHealth();
   }
 
   // Load admin settings
@@ -737,6 +755,17 @@ async function saveDbMaintenance() {
   }
 }
 
+async function loadDbHealth() {
+  loadingHealth.value = true;
+  try {
+    dbHealth.value = await apiGet<DbHealth>("/api/admin/db-maintenance/health");
+  } catch {
+    // leave previous value
+  } finally {
+    loadingHealth.value = false;
+  }
+}
+
 async function runDbMaintenanceNow() {
   runningDbMaint.value = true;
   try {
@@ -747,6 +776,7 @@ async function runDbMaintenanceNow() {
       failed ? `Maintenance ran with ${failed} error(s)` : "Maintenance complete",
       failed ? "error" : "success",
     );
+    await loadDbHealth(); // reflect the dead-tuple drop
   } catch (e) {
     const body = (e as { body?: { error?: string } }).body;
     toastStore.show(body?.error || "Maintenance run failed", "error");
@@ -1825,6 +1855,41 @@ function formatUserDate(iso: string): string {
             </li>
           </ul>
         </div>
+
+        <div class="db-health">
+          <h3 class="subsection-label">
+            Table health
+            <span v-if="dbHealth" class="db-health-total">· database {{ formatBytes(dbHealth.db_bytes) }}</span>
+          </h3>
+          <p class="field-hint">
+            Dead-tuple ratio is bloat — rows left by updates/deletes not yet reclaimed.
+            Above {{ DEAD_PCT_WARN }}% on a large table means autovacuum is falling behind;
+            consider adding it to the maintenance set.
+          </p>
+          <p v-if="loadingHealth && !dbHealth" class="field-hint">Loading…</p>
+          <div v-else-if="dbHealth" class="db-health-scroll">
+            <table class="db-health-table">
+              <thead>
+                <tr>
+                  <th>Table</th><th class="num">Size</th><th class="num">Live</th>
+                  <th class="num">Dead</th><th class="num">Dead %</th>
+                  <th>Last vacuum</th><th>Last analyze</th>
+                </tr>
+              </thead>
+              <tbody>
+                <tr v-for="t in dbHealth.tables" :key="t.table" :class="{ 'dh-warn': t.dead_pct >= DEAD_PCT_WARN }">
+                  <td><code>{{ t.table }}</code></td>
+                  <td class="num">{{ formatBytes(t.total_bytes) }}</td>
+                  <td class="num">{{ t.live.toLocaleString() }}</td>
+                  <td class="num">{{ t.dead.toLocaleString() }}</td>
+                  <td class="num">{{ t.dead_pct }}%</td>
+                  <td>{{ t.last_vacuum ? new Date(t.last_vacuum).toLocaleString() : "—" }}</td>
+                  <td>{{ t.last_analyze ? new Date(t.last_analyze).toLocaleString() : "—" }}</td>
+                </tr>
+              </tbody>
+            </table>
+          </div>
+        </div>
       </section>
 
       <section class="settings-section full-width">
@@ -2481,6 +2546,32 @@ function formatUserDate(iso: string): string {
 }
 .db-maint-table-list .dm-status { color: var(--color-text-muted); }
 .db-maint-table-list li.dm-failed .dm-status { color: var(--color-danger); }
+
+/* DB table-health readout */
+.db-health { margin-top: 1.5rem; }
+.db-health-total { color: var(--color-text-muted); font-weight: 400; }
+.db-health-scroll { overflow-x: auto; margin-top: 0.5rem; }
+.db-health-table {
+  width: 100%;
+  border-collapse: collapse;
+  font-size: 0.82rem;
+}
+.db-health-table th, .db-health-table td {
+  text-align: left;
+  padding: 0.35rem 0.6rem;
+  border-bottom: 1px solid var(--color-border);
+  white-space: nowrap;
+}
+.db-health-table th {
+  font-size: 0.72rem;
+  text-transform: uppercase;
+  letter-spacing: 0.04em;
+  color: var(--color-text-muted);
+  font-weight: 500;
+}
+.db-health-table td.num, .db-health-table th.num { text-align: right; font-variant-numeric: tabular-nums; }
+.db-health-table tr.dh-warn td { color: var(--color-warning); }
+.db-health-table tr.dh-warn td:first-child code { color: var(--color-warning); }
 .btn-warn:hover:not(:disabled) {
   background: var(--color-warning);
   color: #fff;

src/scribe/routes/admin.py

@@ -225,6 +225,14 @@ async def get_db_maintenance():
     })
 
 
+@admin_bp.route("/db-maintenance/health", methods=["GET"])
+@admin_required
+async def get_db_maintenance_health():
+    """Read-only per-table bloat/health stats from Postgres + total DB size."""
+    from scribe.services.db_maintenance import get_table_health
+    return jsonify(await get_table_health())
+
+
 @admin_bp.route("/db-maintenance", methods=["PUT"])
 @admin_required
 async def update_db_maintenance():

src/scribe/services/db_maintenance.py

@@ -18,7 +18,9 @@ import logging
 import time
 from datetime import datetime, timezone
 
-from scribe.models import engine
+from sqlalchemy import text
+
+from scribe.models import async_session, engine
 from scribe.services.settings import get_admin_setting, set_admin_setting
 
 logger = logging.getLogger(__name__)
@@ -104,6 +106,63 @@ async def run_maintenance(tables: tuple[str, ...] | list[str] | None = None) ->
     return summary
 
 
+# Read-only table-health query against Postgres' own statistics views. The key
+# signal is the dead-tuple ratio (dead / (live + dead)) — that IS bloat; a high
+# ratio on a large table means autovacuum isn't keeping up. Timestamps confirm
+# maintenance is actually running on each table.
+_HEALTH_SQL = text("""
+    SELECT
+      relname AS table_name,
+      n_live_tup AS live,
+      n_dead_tup AS dead,
+      CASE WHEN n_live_tup + n_dead_tup > 0
+           THEN round(100.0 * n_dead_tup / (n_live_tup + n_dead_tup), 1)
+           ELSE 0 END AS dead_pct,
+      pg_total_relation_size(relid) AS total_bytes,
+      n_mod_since_analyze AS mod_since_analyze,
+      GREATEST(last_vacuum, last_autovacuum) AS last_vacuum,
+      GREATEST(last_analyze, last_autoanalyze) AS last_analyze
+    FROM pg_stat_user_tables
+    ORDER BY n_dead_tup DESC, total_bytes DESC
+""")
+
+
+def _iso(value) -> str | None:
+    return value.isoformat() if value is not None else None
+
+
+async def get_table_health() -> dict:
+    """Per-table health from Postgres statistics + the total database size.
+
+    Read-only (system catalogs only). Returns:
+        {"db_bytes": int, "tables": [
+            {"table": str, "live": int, "dead": int, "dead_pct": float,
+             "total_bytes": int, "mod_since_analyze": int,
+             "last_vacuum": iso|None, "last_analyze": iso|None}
+        ]}
+    """
+    async with async_session() as session:
+        db_bytes = (
+            await session.execute(text("SELECT pg_database_size(current_database())"))
+        ).scalar() or 0
+        rows = (await session.execute(_HEALTH_SQL)).mappings().all()
+
+    tables = [
+        {
+            "table": r["table_name"],
+            "live": int(r["live"] or 0),
+            "dead": int(r["dead"] or 0),
+            "dead_pct": float(r["dead_pct"] or 0),
+            "total_bytes": int(r["total_bytes"] or 0),
+            "mod_since_analyze": int(r["mod_since_analyze"] or 0),
+            "last_vacuum": _iso(r["last_vacuum"]),
+            "last_analyze": _iso(r["last_analyze"]),
+        }
+        for r in rows
+    ]
+    return {"db_bytes": int(db_bytes), "tables": tables}
+
+
 async def get_last_run() -> dict | None:
     """Return the most recent run summary, or None if it has never run."""
     raw = await get_admin_setting(_LAST_RUN_KEY, "")

tests/test_routes_db_maintenance.py

@@ -3,7 +3,12 @@
 
 def test_admin_handlers_callable():
     from scribe.routes import admin as admin_routes
-    for name in ("get_db_maintenance", "update_db_maintenance", "run_db_maintenance_now"):
+    for name in (
+        "get_db_maintenance",
+        "update_db_maintenance",
+        "run_db_maintenance_now",
+        "get_db_maintenance_health",
+    ):
         assert callable(getattr(admin_routes, name))
 
 
@@ -13,6 +18,7 @@ def test_db_maintenance_routes_registered():
     rules = {r.rule for r in app.url_map.iter_rules()}
     assert "/api/admin/db-maintenance" in rules
     assert "/api/admin/db-maintenance/run" in rules
+    assert "/api/admin/db-maintenance/health" in rules
 
 
 def test_scheduler_surface():
@@ -28,8 +34,13 @@ def test_scheduler_surface():
 
 
 def test_service_surface_and_allowlist():
-    from scribe.services.db_maintenance import MAINTENANCE_TABLES, get_last_run, run_maintenance
+    from scribe.services.db_maintenance import (
-    assert callable(run_maintenance) and callable(get_last_run)
+        MAINTENANCE_TABLES,
+        get_last_run,
+        get_table_health,
+        run_maintenance,
+    )
+    assert callable(run_maintenance) and callable(get_last_run) and callable(get_table_health)
     # The allowlist is a closed tuple covering exactly the high-churn tables.
     assert isinstance(MAINTENANCE_TABLES, tuple)
     assert set(MAINTENANCE_TABLES) == {

tests/test_services_db_maintenance.py

@@ -76,6 +76,41 @@ async def test_one_table_failure_does_not_abort_the_rest():
     assert all(r["ok"] for r in summary["tables"][1:])
 
 
+def _health_session(db_bytes, rows):
+    s = AsyncMock()
+    s.__aenter__ = AsyncMock(return_value=s)
+    s.__aexit__ = AsyncMock(return_value=False)
+    size_res = MagicMock()
+    size_res.scalar.return_value = db_bytes
+    rows_res = MagicMock()
+    rows_res.mappings.return_value.all.return_value = rows
+    s.execute = AsyncMock(side_effect=[size_res, rows_res])
+    return s
+
+
+@pytest.mark.asyncio
+async def test_table_health_shapes_rows_and_db_size():
+    from datetime import datetime, timezone
+
+    from scribe.services.db_maintenance import get_table_health
+    vac = datetime(2026, 6, 14, 4, 0, tzinfo=timezone.utc)
+    rows = [
+        {"table_name": "notes", "live": 1000, "dead": 300, "dead_pct": 23.1,
+         "total_bytes": 5_000_000, "mod_since_analyze": 50,
+         "last_vacuum": vac, "last_analyze": None},
+    ]
+    with patch("scribe.services.db_maintenance.async_session",
+               return_value=_health_session(42_000_000, rows)):
+        health = await get_table_health()
+
+    assert health["db_bytes"] == 42_000_000
+    t = health["tables"][0]
+    assert t["table"] == "notes"
+    assert t["dead_pct"] == 23.1
+    assert t["last_vacuum"] == vac.isoformat()
+    assert t["last_analyze"] is None  # null timestamp passes through as None
+
+
 @pytest.mark.asyncio
 async def test_summary_is_persisted_as_admin_setting():
     engine, _ = _mock_engine()