Pull partner token ledgers into a RAM cache so fleet totals are fleet-wide

AI/README-AI.md

@@ -250,11 +250,23 @@ straight from the shell, since it is just a delimited file:
 awk -F'|' -v d="$(date +%F)" '$1==d {p+=$6; c+=$7} END {print p+c}' data/ai_token_history.db
 ```
 
-**The host column is where the turn ran, not where the file is read.** Each host keeps its own
-`data/` and nothing syncs it, so a host only ever sees its own rows — the partner shows as
-"not collected here" on the tab, never as zero. Carrying a partner's totals would mean extending
-the partnership payload fetch; the column exists so that stays a display change rather than a
-migration.
+**The host column is where the turn ran, not where the file is read.** Each host writes only its
+own rows.
+
+`ai_token_sync.sh` pulls each partner's ledger into `/tmp/.cache/vv/ai/<slot>.tokens.db` — the
+same trick `conf_sync.sh` uses for partner confs, and it runs from
+`INTERMEDIATE_MAINTENANCE_SCRIPTS` every four hours. The tab then reads every ledger it can see,
+so a fleet total is a fleet total.
+
+Pull only, no push: nothing here is needed by anyone else, and a reader that fetches its own data
+controls its own freshness instead of depending on the partner's cron. A partner file may only
+contribute rows whose `host` column matches its filename — a ledger copied into the wrong slot
+would otherwise be double-counted against a total that still looked plausible.
+
+The cache is tmpfs with **no save/restore pair**, unlike the conf cache. Stale counters are worse
+than absent ones: absent renders as "not collected here", stale renders as fact. An unreachable
+partner leaves its file alone and logs at info, because a partner being down for weeks is a
+normal state, not an incident.
 
 Pruning is by row count (`AI_TOKEN_RETAIN_ROWS`, default 20000) and happens on write, but only
 once the file passes a size threshold — an ordinary turn costs a `stat()` and an append. The CLI

AI/ai_token_sync.sh

@@ -0,0 +1,191 @@
+#!/bin/bash
+# ==============================================================================================
+# ============================= AI Token Ledger Sync ===========================================
+# ==============================================================================================
+#
+# PURPOSE
+# ─────────────────────────────────────────────────────────────────────────────
+# Pulls each partner's AI token ledger into a RAM cache at /tmp/.cache/vv/ai/, so the AI tab
+# can report usage for the whole fleet instead of only the host the browser happens to be on.
+#
+# Each host records its own turns to data/ai_token_history.db and nothing syncs that file, so
+# without this a host can only ever see its own totals. The tab is careful to say "not collected
+# here" rather than 0 for a partner it cannot see; this script is what turns that into a number.
+#
+# Same trick as conf_sync.sh, and deliberately so — resolve the partner over Tailscale, scp one
+# small file into a tmpfs cache, let the reader treat a missing file as "unknown".
+#
+# ==============================================================================================
+# OPERATIONAL MODEL
+# ==============================================================================================
+#
+#   1. Gates — PARTNERSHIP_ENABLED, AI_ENABLED, AI_TOKEN_SYNC_ENABLED
+#   2. Per partner:
+#        a. Resolve their Tailscale IP
+#        b. Resolve their SCRIPTS_DIR from their varaverk.cfg (they may be in appdata mode)
+#        c. scp their data/ai_token_history.db → $AI_TOKEN_CACHE_DIR/<slot>.tokens.db
+#
+# Pull only, no push. conf_sync.sh pushes as well because a partner may be unable to reach us
+# and still needs our credentials; nothing here is needed by anyone else, and a reader that
+# fetches its own data controls its own freshness rather than depending on the partner's cron.
+#
+# ==============================================================================================
+# OPERATIONAL SAFEGUARDS
+# ==============================================================================================
+#
+#   An unreachable partner is not a failure.
+#       HOST2 is expected to be down for long stretches during onboarding. A warn every four
+#       hours would train the operator to ignore this script's output, and the AI diagnostic
+#       path treats every log WARN as actionable. Unresolvable partners are counted and
+#       reported once at info level; only a partner that resolves and then fails to transfer
+#       is treated as an error.
+#
+#   The cache is never written directly.
+#       scp lands on a .part file that is renamed into place, so a transfer interrupted halfway
+#       cannot leave the reader parsing half a ledger. A truncated final row would be skipped by
+#       the field-count check on the PHP side, but a torn file should not reach it at all.
+#
+#   Nothing is ever written back to the partner.
+#       This script only reads. A bug here cannot corrupt a partner's accounting.
+#
+#   The cache is tmpfs and deliberately not preserved.
+#       Unlike the conf cache there is no save/restore pair. Stale counters are worse than
+#       absent ones: absent reads as "not collected here", stale reads as fact.
+#
+# ==============================================================================================
+# CONFIGURATION
+# ==============================================================================================
+#
+# master.conf
+#
+#   AI_ENABLED               Whole AI subsystem gate
+#   AI_TOKEN_SYNC_ENABLED    This script's own toggle (default: true)
+#   PARTNERSHIP_ENABLED      Checked via require_partnership()
+#   SSH_KEY                  Key used for all partner ssh/scp operations
+#
+# load_config.sh
+#
+#   AI_TOKEN_CACHE_DIR       tmpfs directory the tab reads partner ledgers from
+#
+# host*.conf
+#
+#   HOST*  — hostnames used to build the partner list via detect_hosts()
+#
+# ==============================================================================================
+# RUNTIME MODES
+# ==============================================================================================
+#
+#   ai_token_sync.sh             Pull every reachable partner's ledger
+#   ai_token_sync.sh --dry-run   Report what would be pulled, transfer nothing
+#   ai_token_sync.sh --log       Verbose output
+#
+# ==============================================================================================
+
+SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
+source "$SCRIPT_DIR/../load_config.sh"
+
+parse_args "$@"
+
+if [[ "$EUID" -ne 0 ]]; then
+    error "Must be run as root"
+    exit 1
+fi
+
+acquire_lock
+
+detect_hosts
+require_partnership
+
+if [[ "${AI_ENABLED:-false}" != true ]]; then
+    log "AI_ENABLED=false — skipping"
+    exit 0
+fi
+
+if [[ "${AI_TOKEN_SYNC_ENABLED:-true}" == false ]]; then
+    log "AI_TOKEN_SYNC_ENABLED=false — skipping"
+    exit 0
+fi
+
+CACHE_DIR="$AI_TOKEN_CACHE_DIR"
+SSH_TIMEOUT=10
+
+# Mirrors conf_sync.sh — the remote may be in appdata storage mode, so its ledger is not
+# necessarily under /boot.
+_remote_scripts_dir() {
+    local ip="$1"
+    local cfg line sd
+    cfg=$(timeout "$SSH_TIMEOUT" ssh -i "$SSH_KEY" \
+        -o ConnectTimeout="$SSH_TIMEOUT" -o BatchMode=yes -o StrictHostKeyChecking=no \
+        "root@${ip}" "cat /boot/config/plugins/varaverk/varaverk.cfg 2>/dev/null" 2>/dev/null) || true
+    while IFS= read -r line; do
+        [[ "$line" == SCRIPTS_DIR=* ]] || continue
+        sd="${line#SCRIPTS_DIR=}"; sd="${sd//\"/}"; sd="${sd//\'/}"
+        echo "$sd"; return
+    done <<< "$cfg"
+    echo "/boot/config/plugins/varaverk"
+}
+
+[[ "$DRY_RUN" == true ]] && warn "DRY RUN — no changes will be made"
+
+if [[ "$DRY_RUN" == false ]]; then
+    mkdir -p "$CACHE_DIR" && chmod 755 "$CACHE_DIR"
+fi
+
+PULLED=0
+OFFLINE=0
+FAILED=0
+
+for host_var in $(compgen -v | grep -E '^HOST[0-9]+$' | sort); do
+    partner_host="${!host_var}"
+    [[ -z "$partner_host" ]] && continue
+    [[ "${host_var,,}" == "${MY_ID,,}" ]] && continue
+
+    partner_slot="${host_var,,}"
+    partner_ip=$(resolve_tailscale_ip "$partner_host" 2>/dev/null || true)
+
+    if [[ -z "$partner_ip" ]]; then
+        log "$partner_host — offline or unresolvable, leaving its ledger absent"
+        (( OFFLINE++ ))
+        continue
+    fi
+
+    remote_sd=$(_remote_scripts_dir "$partner_ip")
+    remote_db="${remote_sd}/data/ai_token_history.db"
+
+    if [[ "$DRY_RUN" == true ]]; then
+        warn "DRY RUN — would pull $partner_host:$remote_db → $CACHE_DIR/${partner_slot}.tokens.db"
+        continue
+    fi
+
+    # A partner with AI off has no ledger at all. That is not an error — it is the same
+    # "nothing collected" the tab already knows how to render.
+    if ! timeout "$SSH_TIMEOUT" ssh -i "$SSH_KEY" \
+        -o ConnectTimeout="$SSH_TIMEOUT" -o BatchMode=yes -o StrictHostKeyChecking=no \
+        "root@${partner_ip}" "[[ -f '$remote_db' ]]" 2>/dev/null; then
+        log "$partner_host — no ledger on that host yet"
+        rm -f "$CACHE_DIR/${partner_slot}.tokens.db"
+        (( OFFLINE++ ))
+        continue
+    fi
+
+    if timeout "$SSH_TIMEOUT" scp -i "$SSH_KEY" \
+        -o ConnectTimeout="$SSH_TIMEOUT" -o BatchMode=yes -o StrictHostKeyChecking=no \
+        "root@${partner_ip}:${remote_db}" \
+        "$CACHE_DIR/${partner_slot}.tokens.db.part" 2>/dev/null \
+       && mv -f "$CACHE_DIR/${partner_slot}.tokens.db.part" "$CACHE_DIR/${partner_slot}.tokens.db"; then
+        chmod 644 "$CACHE_DIR/${partner_slot}.tokens.db" 2>/dev/null
+        echo "Pulled ${partner_slot} ledger from $partner_host ✅"
+        (( PULLED++ ))
+    else
+        rm -f "$CACHE_DIR/${partner_slot}.tokens.db.part"
+        warn "Could not pull ${partner_slot} ledger from $partner_host"
+        (( FAILED++ ))
+    fi
+done
+
+info "AI token sync complete — pulled $PULLED${OFFLINE:+, $OFFLINE unavailable}${FAILED:+, $FAILED failed}"
+
+# Only a partner that answered and then failed the transfer is worth an exit code. An absent
+# partner is the normal state while HOST2 is being rebuilt.
+[[ "$FAILED" -gt 0 ]] && exit 1
+exit 0

Deployment/master.conf.template

@@ -380,6 +380,7 @@
         # to 1 day). 30min was true "always fresh" but had no consumer that needed it that often.
         "Arrs_Stack/arr_cache_prefill.sh ARR_PREFILL_WAIT_MINUTES=1"  # keep the shared arr tracked-data cache fresh
         "Arrs_Stack/arrs_failed_stalled_recovery.sh"  # blocklist + re-search failed/stalled arr queue items
+        "AI/ai_token_sync.sh"                         # pull partner AI token ledgers into the tmpfs cache
     )
     # arr_sync.sh runs as a fixed first step in intermediate_sync_maintenance.sh — not listed here.
     # It is controlled by ARR_SYNC_ENABLED (see Arr Sync section above).
@@ -1661,6 +1662,13 @@
     AI_TOKEN_DB="$DATA_DIR/ai_token_history.db"
     AI_TOKEN_RETAIN_ROWS=20000      # oldest rows dropped past this — years of ordinary use
 
+# AI/ai_token_sync.sh pulls each partner's ledger into the tmpfs cache the tab reads, so the
+# fleet total is a fleet total. Same trick conf_sync.sh uses for partner confs, minus the push:
+# nothing here is needed by anyone else, so the reader fetches its own data and controls its
+# own freshness. An unreachable partner is a quiet skip, not a warning — a partner is expected
+# to be down for long stretches, and a four-hourly warn trains you to ignore the script.
+    AI_TOKEN_SYNC_ENABLED=true
+
 # ━━━ AI Feature Toggles ━━━
 # Tier 1 is narration — it cannot change a decision. Tier 2 adds context to a decision a script
 # already made. Tier 3 assists a human. Enable in that order, and give each one weeks.

Plugin/unraid/include/ai.php

@@ -611,6 +611,14 @@ function vv_ai_token_prune(string $db): void {
     @rename($tmp, $db);
 }
 
+// Partner ledgers pulled by AI/ai_token_sync.sh into tmpfs. Same trick conf_sync.sh uses for
+// partner confs: the reader treats a missing file as "unknown", never as zero.
+//
+// Mirrors AI_TOKEN_CACHE_DIR in load_config.sh. Hardcoded rather than read from conf because it
+// is a derived path constant on the shell side too — neither end reads it from a conf file, so
+// there is no single value to drift away from.
+const VV_AI_TOKEN_CACHE_DIR = '/tmp/.cache/vv/ai';
+
 // Aggregates the file into today / last 7 days / all time, both overall and per host.
 //
 // Every detected host is present in 'hosts' whether or not it has rows, with 'seen' telling the
@@ -629,31 +637,49 @@ function vv_ai_token_stats(): array {
         'first'  => null, 'last' => null, 'best_tok_s' => null,
     ];
 
+    $me = vv_detect_host();
     foreach (vv_known_hosts() as $id => $name) {
-        $out['hosts'][$id] = ['name' => $name, 'seen' => false, 'self' => $id === vv_detect_host(),
-                              'today' => $blank, 'week' => $blank, 'all' => $blank];
+        $out['hosts'][$id] = ['name' => $name, 'seen' => false, 'self' => $id === $me,
+                              'synced' => null, 'today' => $blank, 'week' => $blank, 'all' => $blank];
     }
 
-    $fh = @fopen($db, 'r');
-    if (!$fh) return $out;
+    // Ledgers to read: our own, plus whatever AI/ai_token_sync.sh has pulled from partners.
+    //
+    // Each entry carries the host slot it is allowed to contribute rows for. A partner file may
+    // only add rows whose host column matches its filename — a ledger copied into the wrong slot,
+    // or a partner that somehow cached ours, would otherwise be counted twice against a total
+    // that still looked plausible. Our own file is trusted for any host, because it is the only
+    // one written here and its host column is written by vv_detect_host().
+    $ledgers = [[$db, null]];
+    foreach ((array)@glob(VV_AI_TOKEN_CACHE_DIR . '/host*.tokens.db') as $partnerDb) {
+        if (!preg_match('/(host\d+)\.tokens\.db$/', $partnerDb, $m)) continue;
+        if ($m[1] === $me) continue;                       // our own ledger is already read above
+        $ledgers[] = [$partnerDb, $m[1]];
+        if (isset($out['hosts'][$m[1]])) $out['hosts'][$m[1]]['synced'] = @filemtime($partnerDb) ?: null;
+    }
 
     $add = function (array $b, int $p, int $c): array {
         $b['turns']++; $b['prompt'] += $p; $b['completion'] += $c; $b['total'] += $p + $c;
         return $b;
     };
 
+    foreach ($ledgers as [$path, $onlyHost]) {
+        $fh = @fopen($path, 'r');
+        if (!$fh) continue;
+
         while (($line = fgets($fh)) !== false) {
             $f = explode('|', rtrim($line, "\r\n"));
             if (count($f) < 8) continue;                   // partial write, or a format change
             [$date, , $host, $profile, $source, $p, $c, $tokS] = $f;
             if (!preg_match('/^\d{4}-\d{2}-\d{2}$/', $date)) continue;
+            if ($onlyHost !== null && $host !== $onlyHost) continue;
             $p = (int)$p; $c = (int)$c;
 
             if (!isset($out['hosts'][$host])) {
-            // A host in the file that conf no longer lists — renamed, or a row copied in. Shown
-            // rather than dropped, so the totals always reconcile with the per-host rows.
+                // A host in the file that conf no longer lists — renamed, or a row copied in.
+                // Shown rather than dropped, so the totals reconcile with the per-host rows.
                 $out['hosts'][$host] = ['name' => $host, 'seen' => false, 'self' => false,
-                                    'today' => $blank, 'week' => $blank, 'all' => $blank];
+                                        'synced' => null, 'today' => $blank, 'week' => $blank, 'all' => $blank];
             }
             $out['hosts'][$host]['seen'] = true;
 
@@ -666,13 +692,14 @@ function vv_ai_token_stats(): array {
             $out['profiles'][$profile] = ($out['profiles'][$profile] ?? 0) + $p + $c;
             $out['sources'][$source]   = ($out['sources'][$source]   ?? 0) + $p + $c;
             if ($tokS !== '' && (float)$tokS > (float)($out['best_tok_s'] ?? 0)) $out['best_tok_s'] = (float)$tokS;
-        // Min/max rather than first-row/last-row. Appends are chronological in practice, but a
-        // file that has been hand-edited, merged or restored out of order would otherwise report
-        // a span that reads backwards — and the row cap means the oldest row is not permanent.
+            // Min/max rather than first-row/last-row. Appends are chronological in practice, but
+            // a file hand-edited, merged or restored out of order would otherwise report a span
+            // that reads backwards — and the row cap means the oldest row is not permanent.
             if ($out['first'] === null || $date < $out['first']) $out['first'] = $date;
             if ($out['last']  === null || $date > $out['last'])  $out['last']  = $date;
         }
         fclose($fh);
+    }
 
     arsort($out['profiles']);
     arsort($out['sources']);

Plugin/unraid/pages/ai.php

@@ -473,9 +473,15 @@ if (is_dir('/var/log/varaverk')) {
            + `<span class="vv-ai-hostrow-v">${num(tokData.all.total)}</span></div>`;
     Object.keys(hosts).forEach(id => {
       const x = hosts[id];
+      // Three states, not two. "synced" is a partner whose ledger ai_token_sync.sh pulled;
+      // "not collected here" is a partner we have never seen a ledger for. Collapsing the
+      // second into a zero would report the partner as idle when the truth is we cannot see it.
+      const tag = x.self ? ' · this host'
+                : x.synced ? ' · synced ' + ago(x.synced)
+                : '';
       hh += `<div class="vv-ai-hostrow${tokScope === id ? ' active' : ''}" data-scope="${esc(id)}">`
           + `<span class="vv-ai-hostrow-n">${esc(x.name)}</span>`
-          + `<span class="vv-ai-hostrow-h">${esc(id)}${x.self ? ' · this host' : ''}</span>`
+          + `<span class="vv-ai-hostrow-h">${esc(id)}${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>`;

load_config.sh

@@ -198,7 +198,13 @@
     VV_CACHE_DIR="/tmp/vv_cache"
     CONF_RAM_CACHE_DIR="/tmp/.cache/vv/d"   # tmpfs — cleared every reboot, repopulated by conf_sync.sh
     ARR_CACHE_DIR="/tmp/arr_cache"           # tmpfs — cleared every reboot, restored from DATA_DIR backup by arr_cache_age_seconds()
-    export CONF_DIR LOG_DIR VV_CACHE_DIR CONF_RAM_CACHE_DIR ARR_CACHE_DIR
+    # Partner AI ledgers pulled by AI/ai_token_sync.sh. A sibling of the conf cache rather than
+    # the same directory: that one holds credentials at chmod 700 and is snapshotted to /boot by
+    # conf_cache_save.sh. These are non-secret counters that are worthless when stale, so they
+    # must not be preserved across a reboot — losing them just means "not collected here" until
+    # the next sync, which is the honest answer anyway.
+    AI_TOKEN_CACHE_DIR="/tmp/.cache/vv/ai"   # tmpfs — cleared every reboot, repopulated by ai_token_sync.sh
+    export CONF_DIR LOG_DIR VV_CACHE_DIR CONF_RAM_CACHE_DIR ARR_CACHE_DIR AI_TOKEN_CACHE_DIR
 
 # ━━━ Cleanup ━━━
     unset _conf _host_confs_loaded _adapter LOAD_CONFIG_DIR