diff --git a/Plugin/unraid/css/varaverk.css b/Plugin/unraid/css/varaverk.css
index b6a5615..ed91489 100644
--- a/Plugin/unraid/css/varaverk.css
+++ b/Plugin/unraid/css/varaverk.css
@@ -237,29 +237,42 @@ body.vv-fullscreen #displaybox { padding-left: 1rem !important; padding-top: .5r
   align-items: start;
 }
 
-/* Conversations takes its height from the Assistant rather than setting its own. align-self
-   overrides the row's align-items:start for this one card, so it fills the track the Assistant
-   defines — level at rest, and level again after expanding, since the track grows with it.
+/* Explicit placement, two rows. The left column carries AI over Tokens; Conversations and the
+   Assistant span both rows, so the Assistant sets the height and the two stacked cards divide
+   what it sets. This has to be written as row+column rather than the spans that were inline on
+   the markup: a span says how wide, never which row, and auto-placement would drop Tokens into
+   the first gap it found rather than under AI. */
+#vv-ai-stats-card     { grid-column: 1 / span 1; grid-row: 1; }
+#vv-ai-tokens-card    { grid-column: 1 / span 1; grid-row: 2; }
+#vv-ai-chats-card     { grid-column: 2 / span 2; grid-row: 1 / span 2; }
+#vv-ai-assistant-card { grid-column: 4 / span 5; grid-row: 1 / span 2; }
 
-   It needs its own scroll to do that, which it did not need inside #vv-monitor: there it
-   inherited a cap from `.vv-card > div { flex:1; overflow-y:auto }`. Out here the list is
-   content-driven, so an install with forty saved chats would render forty rows and become the
-   tallest card itself — setting the row height instead of following it, which is the inverse of
-   what this rule is for. The h3 pins and the list takes what is left, the same shape the monitor
-   grid above uses. */
-#vv-monitor-ai #vv-ai-chats-card {
+/* Conversations and Tokens take their height from the Assistant rather than setting their own.
+   align-self overrides the row's align-items:start for these two, so they fill the track the
+   Assistant defines — level at rest, and level again after expanding, since the track grows
+   with it.
+
+   They need their own scroll to do that, which Conversations did not need inside #vv-monitor:
+   there it inherited a cap from `.vv-card > div { flex:1; overflow-y:auto }`. Out here both are
+   content-driven, so forty saved chats — or a ledger with both hosts and a footer — would make
+   the card the tallest thing in its column and set the height it is supposed to be following.
+   The h3 pins and the body takes what is left, the same shape the monitor grid above uses. */
+#vv-monitor-ai #vv-ai-chats-card,
+#vv-monitor-ai #vv-ai-tokens-card {
   align-self: stretch;
   display: flex;
   flex-direction: column;
   overflow: hidden;
 }
-#vv-monitor-ai #vv-ai-chats-card h3 { flex-shrink: 0; }
-/* height:0 with flex:1 is the part that makes this follow rather than lead. flex-basis is
-   already 0, so it changes nothing about the layout; what it changes is the height the list
+#vv-monitor-ai #vv-ai-chats-card h3,
+#vv-monitor-ai #vv-ai-tokens-card h3 { flex-shrink: 0; }
+/* height:0 with flex:1 is the part that makes these follow rather than lead. flex-basis is
+   already 0, so it changes nothing about the layout; what it changes is the height the body
    reports upward. Without it the card can still hand the auto-sized grid track a max-content
-   height of the whole list, and forty conversations would set the row that the Assistant is
-   supposed to set. It grows back to fill the track through flex, then scrolls. */
-#vv-monitor-ai #vv-ai-chats-card > div {
+   height of its whole contents, and the track it was meant to fit inside grows to fit it
+   instead. It grows back to fill the track through flex, then scrolls. */
+#vv-monitor-ai #vv-ai-chats-card > div,
+#vv-monitor-ai #vv-ai-tokens-card > div {
   flex: 1;
   height: 0;
   min-height: 0;
@@ -267,15 +280,57 @@ body.vv-fullscreen #displaybox { padding-left: 1rem !important; padding-top: .5r
   -ms-overflow-style: none;
   scrollbar-width: none;
 }
-#vv-monitor-ai #vv-ai-chats-card > div::-webkit-scrollbar { display: none; }
+#vv-monitor-ai #vv-ai-chats-card > div::-webkit-scrollbar,
+#vv-monitor-ai #vv-ai-tokens-card > div::-webkit-scrollbar { display: none; }
+
+/* ── AI token ledger — shared by the AI tab and the Monitor card ──────────────
+   Moved out of pages/ai.php so both surfaces can draw the same rows. A tab's inline <style> is
+   only present while that tab is, and the Monitor renders this ledger too. The AI panel's own
+   chrome (.vv-ai-tok, the two-column split) stayed behind — it is that panel's frame, not part
+   of the ledger. */
+.vv-ai-hostrow { display:flex; align-items:center; gap:8px; font-size:11px; padding:4px 6px;
+                 border-radius:4px; cursor:pointer; border:1px solid transparent; }
+.vv-ai-hostrow:hover  { background:#141414; }
+.vv-ai-hostrow.active { background:#14202c; border-color:#2d4a6a; }
+.vv-ai-hostrow-n { color:#8a8a8a; }
+.vv-ai-hostrow-h { color:#4a4a4a; font-family:monospace; font-size:10px; }
+.vv-ai-hostrow-v { margin-left:auto; color:#c8c8c8; font-family:monospace; font-size:11px; flex-shrink:0; }
+.vv-ai-hostrow-v.none { color:#3a3a3a; font-style:italic; font-family:inherit; font-size:10px; }
+.vv-ai-tok-grid  { display:grid; grid-template-columns:repeat(auto-fit,minmax(140px,1fr)); gap:12px; }
+.vv-ai-tok-l { font-size:9px; letter-spacing:.08em; text-transform:uppercase; color:#4a4a4a; }
+.vv-ai-tok-v { font-size:17px; font-weight:bold; color:#c8c8c8; font-family:monospace; line-height:1.3; }
+.vv-ai-tok-s { font-size:10px; color:#5a5a5a; }
+.vv-ai-tok-foot { margin-top:9px; padding-top:7px; border-top:1px solid #1a1a1a; font-size:10px;
+                  color:#4a4a4a; display:flex; gap:14px; flex-wrap:wrap; }
+
+/* The ledger in a 1-of-8 column. Every rule here is the same content given a sixth of the width
+   it has on the AI tab: the tiles stop sitting side by side, the host id drops under its name
+   instead of competing with the total for one line, and the footer stacks. */
+/* Name and total keep the first line, the host id takes the second on its own. Ordered rather
+   than reordered in the markup, because the AI tab wants all three on one line and this is the
+   same HTML. flex-basis:100% on the id is what forces the break; without the explicit order the
+   total would follow it onto a third line instead of staying beside the name. */
+#vv-ai-tokens-card .vv-ai-hostrow   { flex-wrap: wrap; gap: 1px 6px; padding: 4px 5px; }
+#vv-ai-tokens-card .vv-ai-hostrow-v { order: 1; }
+#vv-ai-tokens-card .vv-ai-hostrow-h { order: 2; flex-basis: 100%; }
+#vv-ai-tokens-card .vv-ai-tok-grid  { grid-template-columns: 1fr; gap: 9px; }
+#vv-ai-tokens-card .vv-ai-tok-v     { font-size: 15px; }
+#vv-ai-tokens-card .vv-ai-tok-foot  { flex-direction: column; gap: 3px; }
+/* Section headings inside the card. .vv-ai-tok-l is already the 9px uppercase label this wants,
+   so it is reused rather than restated — with room above it to separate the two halves. */
+#vv-ai-tokens-card .vv-ai-tok-head  { margin: 11px 0 5px; }
+#vv-ai-tokens-card .vv-ai-tok-head:first-child { margin-top: 0; }
 
 /* Medium width — 4 columns, mirroring #vv-monitor's own breakpoint. The 1/2/5 spans do not
    survive the narrower grid: 5 of 4 would silently overflow the track. */
 @media (max-width: 1024px) {
   #vv-monitor-ai { grid-template-columns: repeat(4, 1fr) !important; }
-  #vv-ai-stats-card     { grid-column: span 1 !important; }
-  #vv-ai-chats-card     { grid-column: span 3 !important; }
-  #vv-ai-assistant-card { grid-column: span 4 !important; }
+  #vv-ai-stats-card     { grid-column: 1 / span 1 !important; grid-row: 1 !important; }
+  #vv-ai-tokens-card    { grid-column: 1 / span 1 !important; grid-row: 2 !important; }
+  #vv-ai-chats-card     { grid-column: 2 / span 3 !important; grid-row: 1 / span 2 !important; }
+  /* The Assistant drops to its own full-width row here. Five columns of eight is a readable
+     transcript; three of four beside a list is not, and it is the card you sit and read. */
+  #vv-ai-assistant-card { grid-column: 1 / span 4 !important; grid-row: 3 !important; }
 }
 
 /* Phone — single column, natural heights, page scrolls. Matching heights is meaningless once
@@ -283,11 +338,16 @@ body.vv-fullscreen #displaybox { padding-left: 1rem !important; padding-top: .5r
    a scroll region inside it is a trap for a thumb. */
 @media (max-width: 480px) {
   #vv-monitor-ai { grid-template-columns: 1fr !important; }
-  #vv-monitor-ai > .vv-card { grid-column: 1 / -1 !important; }
-  #vv-monitor-ai #vv-ai-chats-card { display: block; overflow: visible; }
-  /* height:0 is only inert while this is a flex item — as a plain block it would collapse the
-     list to nothing, so it goes back with the flex context that justified it. */
-  #vv-monitor-ai #vv-ai-chats-card > div { height: auto; min-height: auto; overflow-y: visible; }
+  /* grid-row has to be released along with grid-column. The explicit rows above are the whole
+     reason Tokens sits under AI, and left in place on a single-column grid they would stack
+     cards into the same cell on top of each other rather than in sequence. */
+  #vv-monitor-ai > .vv-card { grid-column: 1 / -1 !important; grid-row: auto !important; }
+  #vv-monitor-ai #vv-ai-chats-card,
+  #vv-monitor-ai #vv-ai-tokens-card { display: block; overflow: visible; }
+  /* height:0 is only inert while these are flex items — as plain blocks it would collapse the
+     body to nothing, so it goes back with the flex context that justified it. */
+  #vv-monitor-ai #vv-ai-chats-card > div,
+  #vv-monitor-ai #vv-ai-tokens-card > div { height: auto; min-height: auto; overflow-y: visible; }
 }
 
 /* Monitor responsive — 4-column grid at medium width */
diff --git a/Plugin/unraid/include/ai.php b/Plugin/unraid/include/ai.php
index b4a795f..93569cc 100644
--- a/Plugin/unraid/include/ai.php
+++ b/Plugin/unraid/include/ai.php
@@ -511,15 +511,16 @@ function vv_ai_stats_cached(bool $live = false): array {
 // cache write, and — worse — the dashboard and the AI tab could disagree about whether the model
 // is resident, because they would have asked at different moments.
 //
-// Tokens are not part of vv_ai_stats(): that function's shape is the AI tab's banner contract,
-// and the ledger read is 15ms, so it is fetched here rather than widening the payload everything
-// else carries.
+// Tokens are deliberately absent. This block once carried today's total for a single line on the
+// AI card; that line is now a card of its own, which fetches the whole ledger from
+// api/ai.php?action=tokens because it shows the week and all-time beside today. Leaving the field
+// here would have run vv_ai_token_stats() on every cache write — once a minute, forever — and
+// discarded all but one number of it that nothing reads.
 function vv_ai_monitor_block(array $stats): array {
     return [
         'model'   => $stats['model']   ?? '',
         'runtime' => $stats['runtime'] ?? [],
         'index'   => $stats['index']   ?? [],
-        'tokens'  => vv_ai_token_stats()['today'] ?? null,
     ];
 }
 
diff --git a/Plugin/unraid/pages/ai.php b/Plugin/unraid/pages/ai.php
index e5d7f00..cd73867 100644
--- a/Plugin/unraid/pages/ai.php
+++ b/Plugin/unraid/pages/ai.php
@@ -127,20 +127,14 @@ if (is_dir('/var/log/varaverk')) {
 .vv-ai-tok { display:grid; grid-template-columns:var(--vv-tok-split,2fr 3fr); gap:1px; background:#1a1a1a;
              border:1px solid #262626; border-radius:6px; overflow:hidden; }
 @media (max-width:900px) { .vv-ai-tok { grid-template-columns:1fr; } }
-.vv-ai-hostrow { display:flex; align-items:center; gap:8px; font-size:11px; padding:4px 6px;
-                 border-radius:4px; cursor:pointer; border:1px solid transparent; }
-.vv-ai-hostrow:hover  { background:#141414; }
-.vv-ai-hostrow.active { background:#14202c; border-color:#2d4a6a; }
-.vv-ai-hostrow-n { color:#8a8a8a; }
-.vv-ai-hostrow-h { color:#4a4a4a; font-family:monospace; font-size:10px; }
-.vv-ai-hostrow-v { margin-left:auto; color:#c8c8c8; font-family:monospace; font-size:11px; flex-shrink:0; }
-.vv-ai-hostrow-v.none { color:#3a3a3a; font-style:italic; font-family:inherit; font-size:10px; }
-.vv-ai-tok-grid  { display:grid; grid-template-columns:repeat(auto-fit,minmax(140px,1fr)); gap:12px; }
-.vv-ai-tok-l { font-size:9px; letter-spacing:.08em; text-transform:uppercase; color:#4a4a4a; }
-.vv-ai-tok-v { font-size:17px; font-weight:bold; color:#c8c8c8; font-family:monospace; line-height:1.3; }
-.vv-ai-tok-s { font-size:10px; color:#5a5a5a; }
-.vv-ai-tok-foot { margin-top:9px; padding-top:7px; border-top:1px solid #1a1a1a; font-size:10px;
-                  color:#4a4a4a; display:flex; gap:14px; flex-wrap:wrap; }
+
+/* The rows and tiles this panel is built from live in css/varaverk.css, not here. The Monitor
+   tab renders the same ledger in a card of its own, and a style block on this page is only
+   loaded when this page is — which is why they had to leave. What stays above is this panel's
+   own chrome: the two-column split and the border around it, which the Monitor card does not
+   share because it is a .vv-card and already has one.
+     .vv-ai-hostrow*   host rows, including the active state the scope click sets
+     .vv-ai-tok-*      the Today / 7-day / All-time tiles and the footer line */
 
 /* ── Assistant-filed bug reports ─────────────────────────────────────────── */
 .vv-ai-bug   { border-left:2px solid #5a3a2a; background:#140f0c; border-radius:0 3px 3px 0;
diff --git a/Plugin/unraid/pages/monitor.php b/Plugin/unraid/pages/monitor.php
index 19d88f3..184c3b1 100644
--- a/Plugin/unraid/pages/monitor.php
+++ b/Plugin/unraid/pages/monitor.php
@@ -315,7 +315,12 @@ $_vv_doc_vars = array_merge(vv_conf_vars(), ['SCRIPTS_DIR' => SCRIPTS_DIR]);
          it is not here. api/ai.php refuses every action independently, so this is presentation
          rather than access control. -->
 
-    <div class="vv-card" id="vv-ai-stats-card" style="grid-column:span 1;">
+    <!-- Placement for this row lives in css/varaverk.css, not on these cards. It is two rows
+         deep — AI over Tokens in the first column — and a grid-column span here cannot say which
+         row a card belongs to, so an inline span would only fight the stylesheet for half the
+         answer. -->
+
+    <div class="vv-card" id="vv-ai-stats-card">
       <h3>
         <span style="display:flex;align-items:center;gap:5px;">
           <span class="vv-ico"><svg width="12" height="12" viewBox="0 0 12 12" fill="none" stroke="#aaa" stroke-width="1.1" stroke-linecap="round" stroke-linejoin="round"><rect x="2.5" y="2.5" width="7" height="7" rx="1.2"/><circle cx="4.7" cy="5" r="0.7" fill="#888" stroke="none"/><circle cx="7.3" cy="5" r="0.7" fill="#888" stroke="none"/><line x1="4.5" y1="7.3" x2="7.5" y2="7.3"/><line x1="6" y1="2.5" x2="6" y2="0.8"/><line x1="2.5" y1="6" x2="0.8" y2="6"/><line x1="9.5" y1="6" x2="11.2" y2="6"/></svg></span>
@@ -326,7 +331,22 @@ $_vv_doc_vars = array_merge(vv_conf_vars(), ['SCRIPTS_DIR' => SCRIPTS_DIR]);
       <div id="vv-ai-stats-body">Loading...</div>
     </div>
 
-    <div class="vv-card" id="vv-ai-chats-card" style="grid-column:span 2;">
+    <!-- The token ledger, the same one the AI tab draws, under the card whose subsystem spends
+         them. Its own card rather than three more lines on the AI card above: that card answers
+         "is the model healthy right now", and cost accumulated over a month is a different
+         question that was crowding it out one line at a time. -->
+    <div class="vv-card" id="vv-ai-tokens-card">
+      <h3>
+        <span style="display:flex;align-items:center;gap:5px;">
+          <span class="vv-ico"><svg width="12" height="12" viewBox="0 0 12 12" fill="none" stroke="#aaa" stroke-width="1.1" stroke-linecap="round"><line x1="2" y1="10.3" x2="2" y2="6.5"/><line x1="6" y1="10.3" x2="6" y2="2.4"/><line x1="10" y1="10.3" x2="10" y2="7.6"/></svg></span>
+          Tokens
+        </span>
+        <a href="/plugins/varaverk/Varaverk.page?tab=ai" class="vv-card-cog" title="AI tab">⚙</a>
+      </h3>
+      <div id="vv-ai-tokens-body">Loading...</div>
+    </div>
+
+    <div class="vv-card" id="vv-ai-chats-card">
       <h3>
         <span style="display:flex;align-items:center;gap:5px;">
           <span class="vv-ico"><svg width="13" height="12" viewBox="0 0 13 12" fill="none" stroke="#aaa" stroke-width="1.1" stroke-linecap="round" stroke-linejoin="round"><path d="M1 2.2C1 1.5 1.5 1 2.2 1H10.8C11.5 1 12 1.5 12 2.2V6.8C12 7.5 11.5 8 10.8 8H4.5L2 10.5V8H2.2C1.5 8 1 7.5 1 6.8Z"/></svg></span>
@@ -338,7 +358,7 @@ $_vv_doc_vars = array_merge(vv_conf_vars(), ['SCRIPTS_DIR' => SCRIPTS_DIR]);
 
     <!-- Named for what it is, not "dock" — the Scheduler tab's vv-ai-dock is a different
          component and vvAiDockOn() there tests for that id by name. -->
-    <div class="vv-card" id="vv-ai-assistant-card" style="grid-column:span 5;">
+    <div class="vv-card" id="vv-ai-assistant-card">
       <h3>
         <span style="display:flex;align-items:center;gap:5px;">
           <span class="vv-ico"><svg width="12" height="12" viewBox="0 0 12 12" fill="none" stroke="#aaa" stroke-width="1.1" stroke-linecap="round" stroke-linejoin="round"><circle cx="6" cy="6" r="5"/><path d="M4.4 4.6C4.4 3.7 5.1 3.1 6 3.1C6.9 3.1 7.6 3.7 7.6 4.5C7.6 5.9 6 5.7 6 7"/><circle cx="6" cy="8.8" r="0.6" fill="#888" stroke="none"/></svg></span>
@@ -1939,7 +1959,7 @@ function vvPollMonitor(live) {
 function vvRenderAi(ai) {
   const box = document.getElementById('vv-ai-stats-body');
   if (!box) return;
-  const rt = ai.runtime ?? {}, ix = ai.index ?? {}, tok = ai.tokens ?? null;
+  const rt = ai.runtime ?? {}, ix = ai.index ?? {};
 
   const line = (label, value, cls) =>
       `<div style="display:flex;justify-content:space-between;align-items:baseline;gap:8px;margin-bottom:6px;">`
@@ -1977,11 +1997,111 @@ function vvRenderAi(ai) {
     if (ix.stale) h += `<div style="font-size:10px;color:#8a6a3a;margin:-2px 0 7px;">source newer — reindex</div>`;
   }
 
-  if (tok && tok.turns) h += line('Today', tok.total.toLocaleString() + ' tok');
+  box.innerHTML = h;
+}
+
+// ── AI token ledger card ─────────────────────────────────────────────────────
+// Not part of the monitor payload. The cache writer runs once a minute and this is the one card
+// on the page whose numbers only move when a turn finishes, so it is fetched on its own from
+// api/ai.php?action=tokens — the same endpoint and the same shape the AI tab reads, so the two
+// surfaces cannot disagree about what has been spent.
+let vvTokData = null, vvTokScope = 'all';
+
+function vvTokAgo(ts) {
+  if (!ts) return '—';
+  const d = Math.floor(Date.now()/1000) - ts;
+  if (d < 60)    return 'just now';
+  if (d < 3600)  return Math.floor(d/60) + 'm';
+  if (d < 86400) return Math.floor(d/3600) + 'h';
+  return Math.floor(d/86400) + 'd';
+}
+
+function vvLoadTokens() {
+  return fetch('/plugins/varaverk/api/ai.php?action=tokens')
+    .then(r => r.json())
+    .then(d => { if (d.ok) { vvTokData = d.tokens; vvRenderTokens(); } })
+    .catch(() => {});
+}
+
+function vvRenderTokens() {
+  const box = document.getElementById('vv-ai-tokens-body');
+  if (!box || !vvTokData) return;
+  const num   = n => (n || 0).toLocaleString();
+  const hosts = vvTokData.hosts || {};
+
+  // A host with no rows reads "not collected here", never 0. Each host writes to its own data/
+  // and only ai_token_sync.sh moves a ledger between them, so a zero would claim the partner did
+  // no work when the truth is that this host cannot see its ledger at all.
+  let h = `<div class="vv-ai-tok-l vv-ai-tok-head">Nodes</div>`
+        + `<div class="vv-ai-hostrow${vvTokScope === 'all' ? ' active' : ''}" data-scope="all">`
+        + `<span class="vv-ai-hostrow-n">All hosts</span>`
+        + `<span class="vv-ai-hostrow-v">${num(vvTokData.all.total)}</span></div>`;
+
+  Object.keys(hosts).forEach(id => {
+    const x   = hosts[id];
+    const tag = x.self ? ' · this host' : x.synced ? ' · synced ' + vvTokAgo(x.synced) : '';
+    h += `<div class="vv-ai-hostrow${vvTokScope === id ? ' active' : ''}" data-scope="${vvEscAttr(id)}">`
+       + `<span class="vv-ai-hostrow-n">${vvEscHtml(x.name)}</span>`
+       + `<span class="vv-ai-hostrow-h">${vvEscHtml(id)}${vvEscHtml(tag)}</span>`
+       + (x.seen ? `<span class="vv-ai-hostrow-v">${num(x.all.total)}</span>`
+                 : `<span class="vv-ai-hostrow-v none">not collected here</span>`)
+       + `</div>`;
+  });
+
+  const sel   = vvTokScope === 'all' ? vvTokData : hosts[vvTokScope];
+  const scope = vvTokScope === 'all' ? 'all hosts'
+              : (hosts[vvTokScope] ? hosts[vvTokScope].name : vvTokScope);
+  h += `<div class="vv-ai-tok-l vv-ai-tok-head">Token usage — ${vvEscHtml(scope)}</div>`;
+
+  if (!sel || !sel.all.turns) {
+    h += vvTokScope === 'all'
+       ? `<div class="vv-ai-none">no turns recorded yet</div>`
+       : `<div class="vv-ai-none">no rows from this host in the local ledger</div>`;
+    box.innerHTML = h;
+    return;
+  }
+
+  const cell = (label, b) =>
+      `<div><div class="vv-ai-tok-l">${label}</div>`
+    + `<div class="vv-ai-tok-v">${num(b.total)}</div>`
+    + `<div class="vv-ai-tok-s">${num(b.turns)} turn${b.turns === 1 ? '' : 's'}`
+    + ` · ${num(b.prompt)} in / ${num(b.completion)} out</div></div>`;
+
+  h += `<div class="vv-ai-tok-grid">`
+     + cell('Today', sel.today) + cell('Last 7 days', sel.week) + cell('All time', sel.all)
+     + `</div>`;
+
+  // Profile and source splits describe the whole ledger, so they appear only under the all-hosts
+  // scope rather than sitting beneath a host heading they do not describe.
+  const foot = [];
+  if (vvTokData.first) foot.push(`since ${vvEscHtml(vvTokData.first)} · ${vvTokData.days} day${vvTokData.days === 1 ? '' : 's'}`);
+  if (vvTokData.best_tok_s) foot.push(`best ${vvEscHtml(vvTokData.best_tok_s)} tok/s`);
+  if (vvTokScope === 'all') {
+    const by = o => Object.keys(o || {}).map(k => `${vvEscHtml(k)} ${num(o[k])}`).join(' · ');
+    if (Object.keys(vvTokData.profiles || {}).length) foot.push('profile: ' + by(vvTokData.profiles));
+    if (Object.keys(vvTokData.sources  || {}).length) foot.push('source: '  + by(vvTokData.sources));
+  }
+  if (foot.length) h += `<div class="vv-ai-tok-foot">`
+                      + foot.map(f => `<span>${f}</span>`).join('') + `</div>`;
 
   box.innerHTML = h;
 }
 
+// Delegated on the container, which survives every render — the rows inside do not, and binding
+// per row would rebind the whole list on each poll to no purpose.
+if (document.getElementById('vv-ai-tokens-body')) {
+  document.getElementById('vv-ai-tokens-body').addEventListener('click', e => {
+    const row = e.target.closest('.vv-ai-hostrow');
+    if (!row) return;
+    vvTokScope = row.dataset.scope;
+    vvRenderTokens();   // scope is a view of data already held; no refetch
+  });
+  // 60s, not the 5s the dashboard polls at. The ledger only moves when a turn completes, and a
+  // turn completing here refreshes the card directly — this interval exists for turns taken on
+  // the AI tab and for the partner ledger that ai_token_sync.sh pulls every two hours.
+  vvPollRunner(vvLoadTokens, 60000);
+}
+
 // 5s against a payload the cache writer refreshes once a minute. Polling faster cannot make the
 // data newer — it only decides how soon the page notices the writer's update.
 vvPollRunner(vvPollMonitor, 5000);
@@ -2696,6 +2816,9 @@ if (document.getElementById('vv-mon-ai-chat')) {
     empty:   'Ask anything. Questions about this installation are handed to the Varaverk '
            + 'assistant automatically.',
     onChats: id => { if (vvMonChatList) vvMonChatList.setActive(id); },
+    // The turn that just finished is the one spend you watched happen. Waiting up to a minute
+    // for the poll to reflect it reads as the card being broken.
+    onTurn:  () => vvLoadTokens(),
   });
   vvMonChatList = VvAiChatList({ into: 'vv-mon-ai-chats', chat: vvMonChat });
   vvMonChatList.reload();