add pictures to statcounter

README.md

@@ -6,7 +6,10 @@ das komplette serverjonas search
 - `launcher.py` – startet/überwacht den Crawler und rollt Updates aus (siehe unten)
 - `search/` – Suchmaschine darüber (CLI + HTTP-API + Weboberfläche, Port 8083),
   siehe [search/README.md](search/README.md)
-- `web/` – Statistik-Webseite (Port 8081, `jonassearch-stats.service`)
+- `web/` – Statistik-Webseite (Port 8081, `jonassearch-stats.service`).
+  Seiten werden alle 30 Min. gezählt; Bilder alle 2 h exakt, dazwischen nur
+  geschätzt. Alle Diagramme zeigen ausschließlich die exakten Messpunkte.
+- `web/tui.py` – Terminal-Dashboard für dieselben Statistiken (siehe unten)
 - `priorityweb/` – Vorschlagsverfeinerer (Port 8082, `jonassearch-priority.service`)
 - `search/jonassearch-engine.service` – systemd-Unit für die Suchmaschine
   (Port 8083, `jonassearch-engine.service`)
@@ -16,6 +19,27 @@ das komplette serverjonas search
 - `add_pages_pictures.sql` – einmalig laufen lassen: legt `pages.pictures`
   (`text[]`) für die Bild-URLs der gecrawlten Seiten an (idempotent)
 
+## Terminal-Dashboard (`web/tui.py`)
+
+Zeigt Seiten- und Bildanzahl, Wachstum, einen Sparkline-Verlauf und die
+letzten Messwerte direkt im Terminal. Liest dieselbe
+`web/page_counts.json` wie der Statserver – kein Webserver, keine
+zusätzlichen Pakete (nur `curses` aus der Standardlib).
+
+```bash
+python3 web/tui.py                 # interaktiv, Aktualisierung alle 2 s
+python3 web/tui.py --window 7d     # 24h (Default) | 7d | 30d | all
+python3 web/tui.py --once          # Momentaufnahme, z. B. für Logs/Skripte
+python3 web/tui.py --once --width 100 --window all
+```
+
+Tasten im interaktiven Modus: `q` beendet, `r` lädt die Datei sofort neu,
+`t` wechselt das Zeitfenster. `--ascii` ersetzt die Unicode-Blöcke der
+Sparkline. Läuft die Ausgabe nicht in einem Terminal (Pipe, Cron), wird
+automatisch die `--once`-Ausgabe verwendet. Fehlt die Datei, zeigt das
+Dashboard „No samples yet“; ist sie defekt, erscheint die Fehlermeldung
+im Dashboard (bei `--once` auf stderr, Exit-Code 1).
+
 ## Sehr große Seiten / tsvector-Limit
 
 PostgreSQL begrenzt einen einzelnen `tsvector`-Wert hart auf 1.048.575 Bytes

web/app.py

@@ -145,7 +145,7 @@ HTML = r"""
             50% { opacity: 0.35; }
         }
 
-        .ranges { display: flex; gap: 8px; margin-bottom: 18px; }
+        .ranges { display: flex; flex-wrap: wrap; gap: 8px; margin-bottom: 18px; }
 
         .range-btn {
             background: var(--card);
@@ -302,6 +302,7 @@ HTML = r"""
         <button class="range-btn active" data-range="all">All</button>
         <button class="range-btn" data-range="7d">Last 7 days</button>
         <button class="range-btn" data-range="24h">Last 24 h</button>
+        <button class="range-btn" data-range="6h">Last 6 h</button>
     </div>
 
     <div class="grid">
@@ -315,6 +316,11 @@ HTML = r"""
             <div class="value" id="stat-per-day">–</div>
             <div class="hint">Average over full history</div>
         </div>
+        <div class="card">
+            <div class="label">Growth 6 h</div>
+            <div class="value" id="stat-delta-6h">–</div>
+            <div class="hint" id="stat-delta-6h-hint">vs. measurement ~6 h ago</div>
+        </div>
         <div class="card">
             <div class="label">Growth 24 h</div>
             <div class="value" id="stat-delta-24h">–</div>
@@ -400,6 +406,55 @@ HTML = r"""
         <div id="heatmap"></div>
     </div>
 
+    <div class="section-title">Images – all graphs</div>
+    <div class="chart-sub">Only exact measurements (counted every 2 h) are plotted – the estimates in between are excluded.</div>
+    <div class="chart-sub" id="images-empty" style="display:none">
+        No image counts yet – they appear after the next collector run with the image migration in place.
+    </div>
+
+    <div class="chart-container">
+        <div class="chart-title">Images over time</div>
+        <div class="chart-sub">Image URLs found while crawling · exact count every 2 h · only exact measurements are plotted (no estimates) · dashed line = trend &amp; 24 h forecast</div>
+        <div class="chart-box"><canvas id="imagesMainChart"></canvas></div>
+    </div>
+
+    <div class="chart-container">
+        <div class="chart-title">Change per 5 minutes</div>
+        <div class="chart-sub">Normalized to a 5-minute interval · spans between two exact counts (2 h) · grey = raw · colored = smoothed (moving average over 5 measurements)</div>
+        <div class="chart-box small"><canvas id="imagesDeltaChart"></canvas></div>
+    </div>
+
+    <div class="chart-container">
+        <div class="chart-title">Cumulative net growth</div>
+        <div class="chart-sub">Total images gained/lost since the start of the selected range · linear time axis</div>
+        <div class="chart-box small"><canvas id="imagesCumChart"></canvas></div>
+    </div>
+
+    <div class="chart-container">
+        <div class="chart-title">Images per hour (last 48 h)</div>
+        <div class="chart-sub">Sum of changes per hour bucket · green = growth, red = loss</div>
+        <div class="chart-box small"><canvas id="imagesHourlyChart"></canvas></div>
+    </div>
+
+    <div class="chart-row">
+        <div class="chart-container">
+            <div class="chart-title">Image growth by hour of day</div>
+            <div class="chart-sub">Average gain per 5 min, grouped by hour of day (whole history)</div>
+            <div class="chart-box small"><canvas id="imagesHodChart"></canvas></div>
+        </div>
+        <div class="chart-container">
+            <div class="chart-title">Records – biggest image changes</div>
+            <div class="chart-sub">Largest gains and losses between two consecutive measurements (selected range)</div>
+            <div id="records-images"></div>
+        </div>
+    </div>
+
+    <div class="section-title">Activity heatmap – image gain per hour of day</div>
+    <div class="chart-container heatmap">
+        <div class="chart-sub">Rows = days · columns = hours (0–23) · green = growth, red = loss, grey = no data</div>
+        <div id="heatmap-images"></div>
+    </div>
+
     <div class="footer" id="footer"></div>
 </div>
 
@@ -559,10 +614,26 @@ function parsePoints(values) {
         .sort(function (a, b) { return a.x - b.x; });
 }
 
-function filterRange(points) {
+// Nur exakte Bildzählungen (alle 2 h). Die dazwischen geschätzten Werte
+// werden in keinem Diagramm dargestellt, damit alle Graphen ausschließlich
+// echte Messpunkte zeigen.
+function toImagePoints(points) {
+    return (points || [])
+        .filter(function (p) { return p.images !== null && p.images !== undefined; })
+        .filter(function (p) { return p.imagesExact; })
+        .map(function (p) { return { x: p.x, y: p.images }; });
+}
+
+const RANGE_SPANS = {
+    '6h': 6 * 3600000,
+    '24h': 24 * 3600000,
+    '7d': 7 * 24 * 3600000
+};
+
+function filterRange(points, anchorX) {
     if (range === 'all' || !points.length) return points;
-    const last = points[points.length - 1].x;
+    const last = anchorX === undefined ? points[points.length - 1].x : anchorX;
-    const span = range === '24h' ? 24 * 3600000 : 7 * 24 * 3600000;
+    const span = RANGE_SPANS[range] || RANGE_SPANS['24h'];
     return points.filter(function (p) { return p.x >= last - span; });
 }
 
@@ -675,6 +746,12 @@ function baseScales(yFmt) {
 
 function tooltipTitle(items) { return fmtDateFull(items[0].parsed.x); }
 
+// Pages keep the existing camelCase ids (mainChart); other metrics get a prefix
+// (imagesMainChart).
+function chartId(prefix, suffix) {
+    return prefix ? prefix + suffix : suffix.charAt(0).toLowerCase() + suffix.slice(1);
+}
+
 function gradientFill(colorTop, colorBottom) {
     return function (context) {
         const chart = context.chart;
@@ -689,8 +766,14 @@ function gradientFill(colorTop, colorBottom) {
 
 /* ---------- charts ---------- */
 
-function renderCharts(points, deltas, smoothedDeltas) {
+function renderCharts(cfg) {
-    destroyCharts();
+    const prefix = cfg.prefix;
+    const points = cfg.points;
+    const deltas = cfg.deltas;
+    const smoothedDeltas = cfg.smoothedDeltas;
+    const unit = cfg.unit;
+    const accent = cfg.accent;
+    const accentBar = cfg.accentBar;
     const grid = 'rgba(128,128,128,0.15)';
 
     /* 1. main chart + trend */
@@ -710,21 +793,21 @@ function renderCharts(points, deltas, smoothedDeltas) {
         ];
     }
 
-    charts.main = new Chart(document.getElementById('mainChart'), {
+    charts[prefix + 'main'] = new Chart(document.getElementById(chartId(prefix, 'MainChart')), {
         type: 'line',
         data: {
             datasets: [
                 {
-                    label: 'Pages',
+                    label: unit,
                     data: points,
                     parsing: false,
-                    borderColor: ACCENT,
+                    borderColor: accent,
                     borderWidth: 2,
                     pointRadius: 2.5,
                     pointHoverRadius: 5,
                     tension: 0.25,
                     fill: true,
-                    backgroundColor: gradientFill('rgba(79,140,255,0.28)', 'rgba(79,140,255,0.02)'),
+                    backgroundColor: gradientFill(cfg.gradTop, cfg.gradBottom),
                     spanGaps: false
                 },
                 {
@@ -753,8 +836,8 @@ function renderCharts(points, deltas, smoothedDeltas) {
                         title: tooltipTitle,
                         label: function (item) {
                             return item.datasetIndex === 0
-                                ? (fmt(item.parsed.y) + ' Pages')
+                                ? (fmt(item.parsed.y) + ' ' + unit)
-                                : ('Trend: ' + fmt(item.parsed.y) + ' Pages');
+                                : ('Trend: ' + fmt(item.parsed.y) + ' ' + unit);
                         }
                     }
                 }
@@ -763,7 +846,7 @@ function renderCharts(points, deltas, smoothedDeltas) {
     });
 
     /* 2. delta chart */
-    charts.delta = new Chart(document.getElementById('deltaChart'), {
+    charts[prefix + 'delta'] = new Chart(document.getElementById(chartId(prefix, 'DeltaChart')), {
         type: 'line',
         data: {
             datasets: [
@@ -781,7 +864,7 @@ function renderCharts(points, deltas, smoothedDeltas) {
                     label: 'Change per 5 min',
                     data: smoothedDeltas,
                     parsing: false,
-                    borderColor: ACCENT,
+                    borderColor: accent,
                     borderWidth: 2.5,
                     pointRadius: 0,
                     pointHoverRadius: 4,
@@ -824,7 +907,7 @@ function renderCharts(points, deltas, smoothedDeltas) {
                         label: function (item) {
                             const raw = item.raw;
                             return [
-                                'Smoothed: ' + fmtSigned(raw.y) + ' Pages per 5 min',
+                                'Smoothed: ' + fmtSigned(raw.y) + ' ' + unit + ' per 5 min',
                                 'Raw: ' + fmtSigned(raw.raw) + ' per 5 min (change ' + fmtSigned(raw.diff) + ' in ' + fmtDur(raw.dt) + ')'
                             ];
                         }
@@ -839,7 +922,7 @@ function renderCharts(points, deltas, smoothedDeltas) {
         return { x: p.x, y: p.y - points[0].y };
     });
 
-    charts.cum = new Chart(document.getElementById('cumChart'), {
+    charts[prefix + 'cum'] = new Chart(document.getElementById(chartId(prefix, 'CumChart')), {
         type: 'line',
         data: {
             datasets: [{
@@ -874,7 +957,7 @@ function renderCharts(points, deltas, smoothedDeltas) {
                 tooltip: {
                     callbacks: {
                         title: tooltipTitle,
-                        label: function (item) { return fmtSigned(item.parsed.y) + ' Pages since range start'; }
+                        label: function (item) { return fmtSigned(item.parsed.y) + ' ' + unit + ' since range start'; }
                     }
                 }
             }
@@ -891,7 +974,7 @@ function renderCharts(points, deltas, smoothedDeltas) {
     const hourlyLabels = hours.map(function (h) { return fmtDate(h); });
     const hourlyValues = hours.map(function (h) { return Math.round(buckets[h]); });
 
-    charts.hourly = new Chart(document.getElementById('hourlyChart'), {
+    charts[prefix + 'hourly'] = new Chart(document.getElementById(chartId(prefix, 'HourlyChart')), {
         type: 'bar',
         data: {
             labels: hourlyLabels,
@@ -917,7 +1000,7 @@ function renderCharts(points, deltas, smoothedDeltas) {
             },
             plugins: {
                 legend: { labels: { usePointStyle: true, boxWidth: 8, color: mutedColor() } },
-                tooltip: { callbacks: { label: function (item) { return fmtSigned(item.parsed.y) + ' Pages in this hour'; } } }
+                tooltip: { callbacks: { label: function (item) { return fmtSigned(item.parsed.y) + ' ' + unit + ' in this hour'; } } }
             }
         }
     });
@@ -934,14 +1017,14 @@ function renderCharts(points, deltas, smoothedDeltas) {
         return hodCount[i] ? Math.round(s / hodCount[i] * 10) / 10 : 0;
     });
 
-    charts.hod = new Chart(document.getElementById('hodChart'), {
+    charts[prefix + 'hod'] = new Chart(document.getElementById(chartId(prefix, 'HodChart')), {
         type: 'bar',
         data: {
             labels: hodValues.map(function (_, i) { return (i < 10 ? '0' : '') + i + ':00'; }),
             datasets: [{
                 label: 'Avg gain per 5 min',
                 data: hodValues,
-                backgroundColor: hodValues.map(function (v) { return v >= 0 ? 'rgba(79,140,255,0.65)' : 'rgba(255,107,107,0.65)'; }),
+                backgroundColor: hodValues.map(function (v) { return v >= 0 ? accentBar : 'rgba(255,107,107,0.65)'; }),
                 borderRadius: 4
             }]
         },
@@ -959,7 +1042,7 @@ function renderCharts(points, deltas, smoothedDeltas) {
             },
             plugins: {
                 legend: { labels: { usePointStyle: true, boxWidth: 8, color: mutedColor() } },
-                tooltip: { callbacks: { label: function (item) { return fmtSigned(item.parsed.y) + ' Pages per 5 min (avg of ' + hodCount[item.dataIndex] + ' measurements)'; } } }
+                tooltip: { callbacks: { label: function (item) { return fmtSigned(item.parsed.y) + ' ' + unit + ' per 5 min (avg of ' + hodCount[item.dataIndex] + ' measurements)'; } } }
             }
         }
     });
@@ -967,8 +1050,8 @@ function renderCharts(points, deltas, smoothedDeltas) {
 
 /* ---------- records table ---------- */
 
-function renderRecords(deltas) {
+function renderRecords(deltas, id, unit) {
-    const el = document.getElementById('records');
+    const el = document.getElementById(id);
     if (deltas.length < 2) {
         el.innerHTML = '<div class="chart-sub">Not enough data yet.</div>';
         return;
@@ -985,7 +1068,7 @@ function renderRecords(deltas) {
 
     el.innerHTML =
         '<table class="records">' +
-        '<tr><th>Time</th><th style="text-align:right">Change</th><th style="text-align:right">Span</th></tr>' +
+        '<tr><th>Time</th><th style="text-align:right">Change (' + unit + ')</th><th style="text-align:right">Span</th></tr>' +
         rows(gains, 'pos') +
         rows(losses, 'neg') +
         '</table>';
@@ -993,8 +1076,8 @@ function renderRecords(deltas) {
 
 /* ---------- heatmap ---------- */
 
-function renderHeatmap(points) {
+function renderHeatmap(points, id, unit) {
-    const el = document.getElementById('heatmap');
+    const el = document.getElementById(id);
     if (!points.length) { el.innerHTML = ''; return; }
 
     function localDayKey(d) {
@@ -1044,7 +1127,7 @@ function renderHeatmap(points) {
                 const op = maxAbs ? Math.min(0.85, 0.15 + 0.7 * Math.abs(dv) / maxAbs) : 0.3;
                 const color = dv >= 0 ? '46,204,113' : '255,107,107';
                 style = ' style="background:rgba(' + color + ',' + op.toFixed(2) + ')"';
-                title = fmtDay(new Date(day + 'T12:00:00').getTime()) + ' ' + h + ':00 – ' + fmtSigned(dv) + ' Pages (' + fmt(val) + ' total)';
+                title = fmtDay(new Date(day + 'T12:00:00').getTime()) + ' ' + h + ':00 – ' + fmtSigned(dv) + ' ' + unit + ' (' + fmt(val) + ' total)';
             }
             html += '<td><div class="cell"' + style + ' title="' + title + '"></div></td>';
         }
@@ -1123,6 +1206,12 @@ function renderAll() {
         setSigned('stat-per-day', perHour * 24);
     }
 
+    const span6h = closestBefore(all, last.x - 6 * 3600000, Math.max(3 * intervalMs, 3600000));
+    setDeltaCard('stat-delta-6h', 'stat-delta-6h-hint',
+        span6h ? last.y - span6h.y : null,
+        span6h ? last.x - span6h.x : 0,
+        span6h ? (last.y - span6h.y) * (3600000 / Math.max(1, last.x - span6h.x)) * (5 / 60) : 0);
+
     const span24h = closestBefore(all, last.x - 24 * 3600000, Math.max(6 * intervalMs, 3 * 3600000));
     setDeltaCard('stat-delta-24h', 'stat-delta-24h-hint',
         span24h ? last.y - span24h.y : null,
@@ -1217,9 +1306,55 @@ function renderAll() {
         document.getElementById('footer').textContent = 'Forecast available once there are at least 2 measurements.';
     }
 
-    renderCharts(fpoints, deltas, smoothedDeltas);
+    /* images: same graphs, but only exact measurements (every 2 h) */
-    renderRecords(deltas);
+    const imageAll = toImagePoints(all);
-    renderHeatmap(all);
+    const imageFpoints = filterRange(imageAll, last.x);
+    const imageDeltas = computeDeltas(imageFpoints).deltas;
+    const imageSmoothY = movingAverage(imageDeltas.map(function (d) { return d.y; }), DELTA_SMOOTH_WINDOW);
+    const imageSmoothedDeltas = imageDeltas.map(function (d, i) {
+        return { x: d.x, y: imageSmoothY[i], raw: d.y, dt: d.dt, diff: d.diff };
+    });
+    const imagesEmpty = document.getElementById('images-empty');
+    const hasImageValues = all.some(function (p) { return p.images !== null; });
+    if (!imageAll.length) {
+        imagesEmpty.textContent = hasImageValues
+            ? 'Only estimated image counts so far – the graphs show exact counts only (every 2 h).'
+            : 'No image counts yet – they appear after the next collector run with the image migration in place.';
+        imagesEmpty.style.display = '';
+    } else if (!imageFpoints.length) {
+        imagesEmpty.textContent = 'No exact image counts in the selected range (counted every 2 h).';
+        imagesEmpty.style.display = '';
+    } else {
+        imagesEmpty.style.display = 'none';
+    }
+
+    destroyCharts();
+    renderCharts({
+        prefix: '',
+        points: fpoints,
+        deltas: deltas,
+        smoothedDeltas: smoothedDeltas,
+        unit: 'Pages',
+        accent: ACCENT,
+        accentBar: 'rgba(79,140,255,0.65)',
+        gradTop: 'rgba(79,140,255,0.28)',
+        gradBottom: 'rgba(79,140,255,0.02)'
+    });
+    renderCharts({
+        prefix: 'images',
+        points: imageFpoints,
+        deltas: imageDeltas,
+        smoothedDeltas: imageSmoothedDeltas,
+        unit: 'Images',
+        accent: ACCENT2,
+        accentBar: 'rgba(180,122,255,0.65)',
+        gradTop: 'rgba(180,122,255,0.28)',
+        gradBottom: 'rgba(180,122,255,0.02)'
+    });
+    renderRecords(deltas, 'records', 'Pages');
+    renderRecords(imageDeltas, 'records-images', 'Images');
+    renderHeatmap(all, 'heatmap', 'Pages');
+    renderHeatmap(imageAll, 'heatmap-images', 'Images');
 }
 
 async function update() {

web/test_stats.py

@@ -129,6 +129,59 @@ class StatsPageTest(unittest.TestCase):
         self.assertIn("Estimated · exact every 2 h", html)
         self.assertIn("imagesExact", html)
 
+    def test_page_contains_6h_range_and_card(self):
+        client = stats_app.app.test_client()
+        response = client.get("/")
+        self.assertEqual(response.status_code, 200)
+        html = response.get_data(as_text=True)
+        self.assertIn('data-range="6h"', html)
+        self.assertIn('id="stat-delta-6h"', html)
+        self.assertIn('id="stat-delta-6h-hint"', html)
+        self.assertIn("'6h': 6 * 3600000", html)
+        self.assertIn("closestBefore(all, last.x - 6 * 3600000", html)
+
+    def test_page_contains_all_image_graphs(self):
+        client = stats_app.app.test_client()
+        response = client.get("/")
+        self.assertEqual(response.status_code, 200)
+        html = response.get_data(as_text=True)
+
+        self.assertIn("Images – all graphs", html)
+        for canvas in (
+            "imagesMainChart",
+            "imagesDeltaChart",
+            "imagesCumChart",
+            "imagesHourlyChart",
+            "imagesHodChart",
+        ):
+            self.assertIn('id="' + canvas + '"', html)
+        self.assertIn('id="records-images"', html)
+        self.assertIn('id="heatmap-images"', html)
+        self.assertIn('id="images-empty"', html)
+
+        # the image series skips entries without an image count (old data)
+        # and only plots exact measurements - no estimates between the 2 h counts
+        self.assertIn("function toImagePoints(points)", html)
+        self.assertIn("p.images !== null && p.images !== undefined", html)
+        self.assertIn(".filter(function (p) { return p.imagesExact; })", html)
+        self.assertIn("only exact measurements are plotted (no estimates)", html)
+        self.assertIn("Only exact measurements (counted every 2 h) are plotted", html)
+        self.assertIn("Only estimated image counts so far", html)
+        self.assertIn("No exact image counts in the selected range", html)
+        # existing page chart ids must stay lowercase (mainChart, not MainChart)
+        self.assertIn("function chartId(prefix, suffix)", html)
+        self.assertIn("chartId(prefix, 'MainChart')", html)
+        # both series use the same time window (anchored to the latest page sample)
+        self.assertIn("filterRange(imageAll, last.x)", html)
+
+        # both metrics go through the same renderers, with the right units
+        self.assertIn("unit: 'Images'", html)
+        self.assertIn("unit: 'Pages'", html)
+        self.assertIn("renderRecords(deltas, 'records', 'Pages')", html)
+        self.assertIn("renderRecords(imageDeltas, 'records-images', 'Images')", html)
+        self.assertIn("renderHeatmap(all, 'heatmap', 'Pages')", html)
+        self.assertIn("renderHeatmap(imageAll, 'heatmap-images', 'Images')", html)
+
 
 if __name__ == "__main__":
     unittest.main(verbosity=2)

web/test_tui.py

@@ -0,0 +1,341 @@
+#!/usr/bin/env python3
+"""
+Tests für das Terminal-Dashboard (web/tui.py), ohne Terminal und ohne DB.
+
+Geprüft werden Datei-Lesen, Zeitfenster, Sparkline, die Textdarstellung
+und der Aufruf über die Kommandozeile (`--once`).
+"""
+
+import importlib.util
+import json
+import os
+import subprocess
+import sys
+import tempfile
+import unittest
+from datetime import datetime, timedelta
+
+HERE = os.path.dirname(os.path.abspath(__file__))
+ROOT = os.path.dirname(HERE)
+
+_spec = importlib.util.spec_from_file_location(
+    "jonassearch_stats_tui", os.path.join(HERE, "tui.py")
+)
+tui = importlib.util.module_from_spec(_spec)
+_spec.loader.exec_module(tui)
+
+try:
+    _stats_spec = importlib.util.spec_from_file_location(
+        "jonassearch_stats_app_for_tui", os.path.join(HERE, "app.py")
+    )
+    stats_app = importlib.util.module_from_spec(_stats_spec)
+    _stats_spec.loader.exec_module(stats_app)
+except Exception:  # z. B. Flask nicht installiert
+    stats_app = None
+
+NOW = datetime(2026, 10, 17, 12, 0, 0)
+
+
+def entry(timestamp, count, images=None, exact=None):
+    data = {"timestamp": timestamp.isoformat(), "count": count}
+    if images is not None:
+        data["images"] = images
+    if exact is not None:
+        data["images_exact"] = exact
+    return data
+
+
+def sample_entries():
+    return [
+        entry(NOW - timedelta(hours=2), 980_000, images=100, exact=True),
+        entry(NOW - timedelta(hours=1), 981_000),
+        entry(NOW - timedelta(minutes=30), 981_950, images=120, exact=False),
+    ]
+
+
+def normalized(entries):
+    """Rohdaten so aufbereiten, wie load_entries() sie liefert."""
+    result = []
+    for item in entries:
+        copy = dict(item)
+        copy["timestamp"] = tui.parse_timestamp(copy["timestamp"])
+        result.append(copy)
+    return result
+
+
+class FormatTest(unittest.TestCase):
+
+    def test_fmt_int_uses_dots(self):
+        self.assertEqual(tui.fmt_int(987832), "987.832")
+        self.assertEqual(tui.fmt_int(None), "-")
+
+    def test_fmt_delta(self):
+        self.assertEqual(tui.fmt_delta(950), "+950")
+        self.assertEqual(tui.fmt_delta(-12), "-12")
+        self.assertEqual(tui.fmt_delta(0), "0")
+        self.assertEqual(tui.fmt_delta(None), "-")
+
+    def test_fmt_duration(self):
+        self.assertEqual(tui.fmt_duration(30), "30 s")
+        self.assertEqual(tui.fmt_duration(90), "1 min")
+        self.assertEqual(tui.fmt_duration(5400), "1 h 30 min")
+        self.assertEqual(tui.fmt_duration(90000), "1 d 1 h")
+
+    def test_samples_label(self):
+        self.assertEqual(tui._samples_label(1), "1 sample")
+        self.assertEqual(tui._samples_label(2), "2 samples")
+
+    def test_relative_time(self):
+        self.assertEqual(tui.relative_time(NOW, NOW), "just now")
+        self.assertEqual(
+            tui.relative_time(NOW - timedelta(minutes=3), NOW), "3 min ago"
+        )
+
+
+class ParseTimestampTest(unittest.TestCase):
+
+    def test_iso_and_z(self):
+        self.assertEqual(
+            tui.parse_timestamp("2026-10-17T20:44:20.187761"),
+            datetime(2026, 10, 17, 20, 44, 20, 187761),
+        )
+        parsed = tui.parse_timestamp("2026-10-17T20:44:20Z")
+        self.assertIsNone(parsed.tzinfo)
+        self.assertEqual(
+            parsed,
+            datetime.fromisoformat("2026-10-17T20:44:20+00:00")
+            .astimezone().replace(tzinfo=None),
+        )
+
+    def test_invalid(self):
+        self.assertIsNone(tui.parse_timestamp("kaputt"))
+        self.assertIsNone(tui.parse_timestamp(""))
+        self.assertIsNone(tui.parse_timestamp(None))
+
+
+class LoadEntriesTest(unittest.TestCase):
+
+    def setUp(self):
+        self.tmp = tempfile.TemporaryDirectory()
+        self.path = os.path.join(self.tmp.name, "page_counts.json")
+
+    def tearDown(self):
+        self.tmp.cleanup()
+
+    def _write(self, data):
+        with open(self.path, "w", encoding="utf-8") as handle:
+            if isinstance(data, str):
+                handle.write(data)
+            else:
+                json.dump(data, handle)
+
+    def test_missing_file_is_empty(self):
+        self.assertEqual(tui.load_entries(self.path), [])
+
+    def test_sorts_and_skips_bad_rows(self):
+        self._write([
+            entry(NOW, 20),
+            "kaputt",
+            {"timestamp": "kaputt", "count": 1},
+            {"timestamp": NOW.isoformat(), "count": "x"},
+            entry(NOW - timedelta(minutes=30), 10),
+        ])
+        entries = tui.load_entries(self.path)
+        self.assertEqual([e["count"] for e in entries], [10, 20])
+        self.assertIsInstance(entries[0]["timestamp"], datetime)
+
+    def test_api_format_is_accepted(self):
+        self._write({"values": [entry(NOW, 5, images=7, exact=True)]})
+        entries = tui.load_entries(self.path)
+        self.assertEqual(len(entries), 1)
+        self.assertEqual(entries[0]["images"], 7)
+        self.assertTrue(entries[0]["images_exact"])
+
+    def test_invalid_images_becomes_none(self):
+        self._write([{"timestamp": NOW.isoformat(), "count": 1,
+                      "images": "viele"}])
+        self.assertIsNone(tui.load_entries(self.path)[0]["images"])
+
+    def test_corrupt_json_raises(self):
+        self._write("{not json")
+        with self.assertRaises(tui.StatsDataError):
+            tui.load_entries(self.path)
+
+    def test_wrong_toplevel_type_raises(self):
+        self._write(42)
+        with self.assertRaises(tui.StatsDataError):
+            tui.load_entries(self.path)
+
+
+class SharedScheduleTest(unittest.TestCase):
+
+    def test_exact_interval_matches_statserver(self):
+        if stats_app is None:
+            self.skipTest("web/app.py nicht importierbar (Flask fehlt)")
+        self.assertEqual(
+            tui.EXACT_IMAGE_INTERVAL, stats_app.EXACT_IMAGE_INTERVAL
+        )
+
+
+class WindowTest(unittest.TestCase):
+
+    def test_window_filters_old_entries(self):
+        entries = normalized([
+            entry(NOW - timedelta(days=3), 1),
+            entry(NOW - timedelta(hours=1), 2),
+        ])
+        self.assertEqual(
+            [e["count"] for e in tui.select_window(entries, "24h", NOW)], [2]
+        )
+        self.assertEqual(len(tui.select_window(entries, "all", NOW)), 2)
+
+    def test_latest_images_falls_back(self):
+        entries = normalized([
+            entry(NOW - timedelta(hours=1), 1, images=10, exact=True),
+            entry(NOW, 2),
+        ])
+        images, exact, timestamp = tui.latest_images(entries)
+        self.assertEqual(images, 10)
+        self.assertTrue(exact)
+        self.assertEqual(timestamp, entries[0]["timestamp"])
+
+    def test_next_exact_text(self):
+        recent = normalized([
+            entry(NOW - timedelta(minutes=30), 1, images=1, exact=True)
+        ])
+        self.assertEqual(
+            tui.next_exact_text(recent, NOW), "next exact in 1 h 30 min"
+        )
+        overdue = normalized([
+            entry(NOW - timedelta(hours=3), 1, images=1, exact=True)
+        ])
+        self.assertEqual(
+            tui.next_exact_text(overdue, NOW), "exact count due now"
+        )
+        self.assertEqual(
+            tui.next_exact_text(normalized([entry(NOW, 1)]), NOW),
+            "no exact count yet",
+        )
+
+
+class SparklineTest(unittest.TestCase):
+
+    def test_width_and_padding(self):
+        line = tui.sparkline([1, 2, 3], 10)
+        self.assertEqual(len(line), 10)
+        self.assertTrue(line.startswith("       "))
+
+    def test_single_sample(self):
+        self.assertEqual(tui.sparkline([5], 1), tui.SPARK_BLOCKS[4])
+
+    def test_increasing_values_use_increasing_blocks(self):
+        line = tui.sparkline([1, 2, 3, 4], 4)
+        indices = [tui.SPARK_BLOCKS.index(char) for char in line]
+        self.assertEqual(indices, sorted(indices))
+
+    def test_constant_values_do_not_crash(self):
+        self.assertEqual(tui.sparkline([7, 7, 7], 3), tui.SPARK_BLOCKS[4] * 3)
+
+    def test_empty(self):
+        self.assertEqual(tui.sparkline([], 10), "")
+
+
+class RenderTest(unittest.TestCase):
+
+    def test_shows_numbers_and_hints(self):
+        lines = tui.render_dashboard(
+            normalized(sample_entries()), now=NOW, width=100
+        )
+        text = "\n".join(lines)
+        self.assertIn("Pages      981.950", text)
+        self.assertIn("Images     120", text)
+        self.assertIn("estimated", text)
+        self.assertIn("exact count due now", text)
+        self.assertIn("+950 since previous", text)
+        self.assertIn("q quit", text)
+        self.assertIn("t window (24h)", text)
+        self.assertIn("Recent samples", text)
+
+    def test_lines_fit_width(self):
+        for width in (32, 48, 60, 80):
+            lines = tui.render_dashboard(
+                normalized(sample_entries()), now=NOW, width=width, rows=24
+            )
+            for line in lines:
+                self.assertLessEqual(len(line), width, repr(line))
+
+    def test_row_budget_keeps_footer(self):
+        lines = tui.render_dashboard(
+            normalized(sample_entries()), now=NOW, width=80, rows=12
+        )
+        self.assertLessEqual(len(lines), 12)
+        self.assertIn("q quit", lines[-1])
+
+    def test_unreadable_file_error(self):
+        temp = tempfile.TemporaryDirectory()
+        self.addCleanup(temp.cleanup)
+        path = os.path.join(temp.name, "page_counts.json")
+        with open(path, "w", encoding="utf-8") as handle:
+            handle.write("{kaputt")
+
+        try:
+            tui.load_entries(path)
+        except tui.StatsDataError as exc:
+            lines = tui.render_dashboard([], now=NOW, width=90, error=str(exc))
+        else:
+            self.fail("StatsDataError erwartet")
+
+        self.assertIn("invalid JSON", "\n".join(lines))
+        self.assertIn("q quit", "\n".join(lines))
+
+    def test_empty_data(self):
+        text = "\n".join(tui.render_dashboard([], now=NOW, width=80))
+        self.assertIn("No samples yet", text)
+
+
+class OnceModeTest(unittest.TestCase):
+
+    def setUp(self):
+        self.tmp = tempfile.TemporaryDirectory()
+        self.addCleanup(self.tmp.cleanup)
+        self.path = os.path.join(self.tmp.name, "page_counts.json")
+        with open(self.path, "w", encoding="utf-8") as handle:
+            json.dump(sample_entries(), handle)
+
+    def _run(self, *extra):
+        return subprocess.run(
+            [sys.executable, os.path.join(HERE, "tui.py"), "--once",
+             "--file", self.path, "--width", "100"] + list(extra),
+            capture_output=True,
+            text=True,
+            cwd=ROOT,
+        )
+
+    def test_once_prints_dashboard(self):
+        result = self._run()
+        self.assertEqual(result.returncode, 0, result.stderr)
+        self.assertIn("Pages", result.stdout)
+        self.assertIn("981.950", result.stdout)
+        self.assertIn("q quit", result.stdout)
+
+    def test_once_window_flag(self):
+        result = self._run("--window", "all", "--ascii")
+        self.assertEqual(result.returncode, 0, result.stderr)
+        self.assertIn("all data", result.stdout)
+
+    def test_once_corrupt_file_exits_nonzero(self):
+        with open(self.path, "w", encoding="utf-8") as handle:
+            handle.write("{kaputt")
+        result = self._run()
+        self.assertEqual(result.returncode, 1)
+        self.assertIn("error:", result.stderr)
+
+    def test_once_missing_file_is_empty_state(self):
+        os.unlink(self.path)
+        result = self._run()
+        self.assertEqual(result.returncode, 0, result.stderr)
+        self.assertIn("No samples yet", result.stdout)
+
+
+if __name__ == "__main__":
+    unittest.main()

web/tui.py

@@ -0,0 +1,640 @@
+#!/usr/bin/env python3
+"""
+Terminal-Dashboard für den Statserver.
+
+Liest dieselbe Datei wie `web/app.py` (`page_counts.json`) und zeigt
+Seitenzahl, Bildanzahl, Wachstum, einen Sparkline-Verlauf und die
+letzten Messwerte direkt im Terminal. Startet keinen Webserver und
+benötigt keine zusätzlichen Pakete (nur curses aus der Standardlib).
+
+Aufruf:
+    python3 web/tui.py                 # interaktiv (q beendet)
+    python3 web/tui.py --once          # Momentaufnahme für Skripte/Logs
+    python3 web/tui.py --window 7d     # 24h (Default) | 7d | 30d | all
+"""
+
+from __future__ import annotations
+
+import argparse
+import curses
+import json
+import locale
+import shutil
+import sys
+import time
+from datetime import datetime, timedelta
+from pathlib import Path
+
+DB_DATA = Path(__file__).parent / "page_counts.json"
+
+# Muss zu EXACT_IMAGE_INTERVAL in web/app.py passen (dort per Test
+# abgesichert): Bilder werden nur alle 2 Stunden exakt gezählt.
+EXACT_IMAGE_INTERVAL = 2 * 60 * 60  # Sekunden
+
+REFRESH_SECONDS = 2.0
+DEFAULT_WINDOW = "24h"
+WINDOWS = {
+    "24h": timedelta(hours=24),
+    "7d": timedelta(days=7),
+    "30d": timedelta(days=30),
+    "all": None,
+}
+WINDOW_ORDER = ("24h", "7d", "30d", "all")
+WINDOW_LABELS = {
+    "24h": "last 24 h",
+    "7d": "last 7 days",
+    "30d": "last 30 days",
+    "all": "all data",
+}
+
+SPARK_BLOCKS = "▁▂▃▄▅▆▇█"
+ASCII_BLOCKS = ".:-=+*#%"
+
+TABLE_ROWS_ONCE = 15
+
+
+class StatsDataError(Exception):
+    """Die Statistikdatei fehlt nicht, ist aber nicht lesbar/parsebar."""
+
+
+# --------------------------------------------------------------------------
+# Daten
+# --------------------------------------------------------------------------
+
+def parse_timestamp(value):
+    """ISO-Zeitstempel -> naive lokale datetime (oder None)."""
+    if not isinstance(value, str) or not value.strip():
+        return None
+
+    text = value.strip()
+    if text.endswith("Z"):
+        text = text[:-1] + "+00:00"
+
+    try:
+        timestamp = datetime.fromisoformat(text)
+    except ValueError:
+        return None
+
+    if timestamp.tzinfo is not None:
+        timestamp = timestamp.astimezone().replace(tzinfo=None)
+
+    return timestamp
+
+
+def load_entries(path=DB_DATA):
+    """
+    Messwerte laden, sortiert nach Zeitstempel.
+
+    Akzeptiert sowohl die Liste aus `page_counts.json` als auch das
+    Format der API (`{"values": [...]}`). Einträge ohne gültigen
+    Zeitstempel oder ohne Seitenzahl werden übersprungen.
+    """
+    path = Path(path)
+
+    try:
+        raw = json.loads(path.read_text(encoding="utf-8"))
+    except FileNotFoundError:
+        return []
+    except OSError as exc:
+        raise StatsDataError(f"{path}: {exc}") from exc
+    except json.JSONDecodeError as exc:
+        raise StatsDataError(
+            f"{path}: invalid JSON (line {exc.lineno}: {exc.msg})"
+        ) from exc
+
+    if isinstance(raw, dict):
+        raw = raw.get("values", [])
+
+    if not isinstance(raw, list):
+        raise StatsDataError(f"{path}: expected a JSON list of samples")
+
+    entries = []
+
+    for item in raw:
+        if not isinstance(item, dict):
+            continue
+
+        timestamp = parse_timestamp(item.get("timestamp"))
+        if timestamp is None:
+            continue
+
+        try:
+            count = int(item.get("count"))
+        except (TypeError, ValueError):
+            continue
+
+        entry = dict(item)
+        entry["timestamp"] = timestamp
+        entry["count"] = count
+
+        images = item.get("images")
+        if images is None:
+            entry["images"] = None
+        else:
+            try:
+                entry["images"] = int(images)
+            except (TypeError, ValueError):
+                entry["images"] = None
+
+        entries.append(entry)
+
+    entries.sort(key=lambda entry: entry["timestamp"])
+    return entries
+
+
+def select_window(entries, window, now=None):
+    """Nur Messwerte innerhalb des gewählten Zeitfensters."""
+    delta = WINDOWS.get(window, WINDOWS[DEFAULT_WINDOW])
+    if delta is None:
+        return list(entries)
+
+    now = now or datetime.now()
+    cutoff = now - delta
+    return [entry for entry in entries if entry["timestamp"] >= cutoff]
+
+
+def latest_images(entries):
+    """
+    (images, images_exact, timestamp) des neuesten Eintrags mit
+    Bildanzahl. Ältere Dateien haben kein `images`-Feld -> None.
+    """
+    for entry in reversed(entries):
+        if entry.get("images") is not None:
+            return (
+                entry["images"],
+                entry.get("images_exact") is True,
+                entry["timestamp"],
+            )
+    return None, False, None
+
+
+def next_exact_text(entries, now):
+    """Wann die nächste exakte Bildzählung fällig ist."""
+    last_exact = None
+
+    for entry in reversed(entries):
+        if entry.get("images_exact") is True:
+            last_exact = entry["timestamp"]
+            break
+
+    if last_exact is None:
+        return "no exact count yet"
+
+    remaining = EXACT_IMAGE_INTERVAL - (now - last_exact).total_seconds()
+    if remaining <= 0:
+        return "exact count due now"
+    return f"next exact in {fmt_duration(remaining)}"
+
+
+# --------------------------------------------------------------------------
+# Formatierung
+# --------------------------------------------------------------------------
+
+def fmt_int(value):
+    if value is None:
+        return "-"
+    return f"{int(value):,}".replace(",", ".")
+
+
+def fmt_delta(value):
+    if value is None:
+        return "-"
+    value = int(value)
+    if value == 0:
+        return "0"
+    if value > 0:
+        return "+" + fmt_int(value)
+    return "-" + fmt_int(abs(value))
+
+
+def fmt_duration(seconds):
+    seconds = max(0, int(round(seconds)))
+
+    if seconds < 60:
+        return f"{seconds} s"
+
+    minutes = seconds // 60
+    if minutes < 60:
+        return f"{minutes} min"
+
+    hours, minutes = divmod(minutes, 60)
+    if hours < 24:
+        return f"{hours} h {minutes} min"
+
+    days, hours = divmod(hours, 24)
+    return f"{days} d {hours} h"
+
+
+def relative_time(timestamp, now):
+    seconds = (now - timestamp).total_seconds()
+    if seconds < 5:
+        return "just now"
+    return f"{fmt_duration(seconds)} ago"
+
+
+def sparkline(values, width, blocks=SPARK_BLOCKS):
+    """Sparkline aus Zahlen; links mit Leerzeichen auf `width` aufgefüllt."""
+    if width <= 0 or not values:
+        return ""
+
+    values = list(values)[-width:]
+    low, high = min(values), max(values)
+
+    if high == low:
+        chars = [blocks[len(blocks) // 2]] * len(values)
+    else:
+        span = high - low
+        last = len(blocks) - 1
+        chars = [
+            blocks[min(last, int((value - low) / span * last + 0.5))]
+            for value in values
+        ]
+
+    return " " * (width - len(values)) + "".join(chars)
+
+
+def _fit(text, width):
+    return text[:max(0, width)]
+
+
+def _samples_label(count):
+    return f"{count} sample" if count == 1 else f"{count} samples"
+
+
+def _rate(window_entries):
+    """Wachstum in Seiten pro Stunde (None bei < 2 Messwerten)."""
+    if len(window_entries) < 2:
+        return None
+
+    first, last = window_entries[0], window_entries[-1]
+    hours = (last["timestamp"] - first["timestamp"]).total_seconds() / 3600
+
+    if hours <= 0:
+        return None
+
+    return (last["count"] - first["count"]) / hours
+
+
+def _deltas(entries):
+    """Änderung der Seitenzahl gegenüber dem vorherigen Messwert."""
+    result = {}
+    previous = None
+
+    for entry in entries:
+        result[id(entry)] = (
+            None if previous is None else entry["count"] - previous
+        )
+        previous = entry["count"]
+
+    return result
+
+
+def _time_label(timestamp, now, compact):
+    if compact:
+        return timestamp.strftime("%H:%M")
+    if timestamp.date() == now.date():
+        return timestamp.strftime("%H:%M")
+    return timestamp.strftime("%d.%m. %H:%M")
+
+
+def _table_line(time_text, pages, delta, images, flag, compact):
+    if compact:
+        return (
+            f" {time_text:<6} {pages:>10} {delta:>8}"
+            f" {images:>6} {flag}"
+        )
+    return (
+        f" {time_text:<13} {pages:>12} {delta:>9}"
+        f" {images:>10}  {flag}"
+    )
+
+
+def _build_table(window_entries, delta_by_id, now, width, limit):
+    """Neueste Messwerte zuerst; Kopfzeile als erste Zeile."""
+    compact = width < 62
+    header = _table_line(
+        "Time",
+        "Pages",
+        "Chg" if compact else "Change",
+        "Images",
+        "",
+        compact,
+    )
+
+    rows = []
+    newest_first = list(reversed(window_entries))
+
+    for entry in newest_first[:limit]:
+        flag = ""
+        if entry.get("images") is not None:
+            flag = "exact" if entry.get("images_exact") is True else "est."
+
+        rows.append(_table_line(
+            _time_label(entry["timestamp"], now, compact),
+            fmt_int(entry["count"]),
+            fmt_delta(delta_by_id.get(id(entry))),
+            fmt_int(entry.get("images")),
+            flag,
+            compact,
+        ))
+
+    return header, rows
+
+
+# --------------------------------------------------------------------------
+# Darstellung
+# --------------------------------------------------------------------------
+
+def render_dashboard(entries, window=DEFAULT_WINDOW, now=None, width=80,
+                     rows=None, source=None, error=None,
+                     blocks=SPARK_BLOCKS):
+    """
+    Dashboard als Liste von Textzeilen.
+
+    Reine Funktion ohne curses: dadurch testbar und auch für `--once`
+    bzw. Umleitungen nutzbar. `rows` begrenzt die Höhe (Terminal);
+    ohne Angabe werden bis zu 15 Messwerte angezeigt.
+    """
+    now = now or datetime.now()
+    width = max(24, int(width))
+    source = str(source or DB_DATA)
+
+    lines = [_fit(" JonasSearch stats", width)]
+
+    if error:
+        lines.append("")
+        lines.append(_fit(f" ! {error}", width))
+        lines.append("")
+        lines.append(_fit(" q quit | r retry", width))
+        return lines[:rows] if rows else lines
+
+    if not entries:
+        lines.append("")
+        lines.append(_fit(" No samples yet.", width))
+        lines.append(_fit(f" Waiting for the collector to write {source}", width))
+        lines.append("")
+        lines.append(_fit(" q quit | r reload", width))
+        return lines[:rows] if rows else lines
+
+    latest = entries[-1]
+    previous = entries[-2] if len(entries) > 1 else None
+    delta = None if previous is None else latest["count"] - previous["count"]
+
+    window_entries = select_window(entries, window, now)
+    window_note = ""
+    if not window_entries:
+        window_entries = [latest]
+        window_note = " (no recent samples)"
+
+    rate = _rate(window_entries)
+    images, images_exact, images_ts = latest_images(entries)
+
+    lines.append(_fit(
+        f" updated {latest['timestamp']:%d.%m.%Y %H:%M:%S}"
+        f" ({relative_time(latest['timestamp'], now)})",
+        width,
+    ))
+    lines.append("")
+
+    lines.append(_fit(f" Pages      {fmt_int(latest['count'])}", width))
+
+    growth = [fmt_delta(delta) + " since previous" if delta is not None
+              else "first sample"]
+    if rate is not None:
+        growth.append(
+            f"{fmt_delta(round(rate))}/h over {WINDOW_LABELS[window]}"
+            f" ({_samples_label(len(window_entries))})"
+        )
+    lines.append(_fit("            " + " | ".join(growth), width))
+    lines.append("")
+
+    lines.append(_fit(f" Images     {fmt_int(images)}", width))
+
+    if images is None:
+        images_detail = "not counted yet"
+    else:
+        kind = "exact" if images_exact else "estimated"
+        images_detail = (
+            f"{kind} | counted {images_ts:%H:%M} | exact every 2 h"
+            f" | {next_exact_text(entries, now)}"
+        )
+    lines.append(_fit("            " + images_detail, width))
+    lines.append("")
+
+    values = [entry["count"] for entry in window_entries]
+    low, high = min(values), max(values)
+    lines.append(_fit(
+        f" Pages | {WINDOW_LABELS[window]}{window_note}"
+        f" | {_samples_label(len(window_entries))}",
+        width,
+    ))
+
+    range_text = f"  ({fmt_int(low)} - {fmt_int(high)})"
+    spark_width = max(8, min(64, width - len(range_text) - 2))
+    spark = sparkline(values, spark_width, blocks)
+    lines.append(_fit(" " + spark + range_text, width))
+
+    delta_by_id = _deltas(entries)
+    footer = ["", _fit(
+        f" q quit | r reload | t window ({window}) | file: {source}",
+        width,
+    )]
+
+    if rows is None:
+        limit = TABLE_ROWS_ONCE
+    else:
+        # 3 Zeilen Overhead: Leerzeile, "Recent samples", Tabellenkopf
+        limit = max(0, rows - len(lines) - len(footer) - 3)
+
+    if limit > 0:
+        header, table = _build_table(
+            window_entries, delta_by_id, now, width, limit
+        )
+        lines += ["", _fit(" Recent samples", width), _fit(header, width)]
+        lines += [_fit(row, width) for row in table]
+
+    lines += footer
+
+    if rows is not None and len(lines) > rows:
+        # Sehr kleines Terminal: Kerninfo kürzen, Footer unten halten.
+        lines = lines[:max(0, rows - 1)] + [lines[-1]]
+
+    return lines
+
+
+# --------------------------------------------------------------------------
+# Interaktiv / einmalig
+# --------------------------------------------------------------------------
+
+def _blocks_for(args):
+    if args.ascii:
+        return ASCII_BLOCKS
+
+    encoding = (getattr(sys.stdout, "encoding", None) or "").lower()
+    return SPARK_BLOCKS if "utf" in encoding else ASCII_BLOCKS
+
+
+def _render(args, entries, now, width, rows=None):
+    return render_dashboard(
+        entries,
+        window=args.window,
+        now=now,
+        width=width,
+        rows=rows,
+        source=args.file,
+        blocks=_blocks_for(args),
+    )
+
+
+def emit_once(args, stream=None, errors=None):
+    """Momentaufnahme ausgeben (kein curses). Rückgabe: Exit-Code."""
+    stream = stream or sys.stdout
+    errors = errors or sys.stderr
+
+    try:
+        entries = load_entries(args.file)
+    except StatsDataError as exc:
+        errors.write(f"error: {exc}\n")
+        return 1
+
+    width = args.width or shutil.get_terminal_size((80, 24)).columns
+    lines = _render(args, entries, datetime.now(), width)
+    stream.write("\n".join(lines) + "\n")
+    return 0
+
+
+def next_window(window):
+    try:
+        index = WINDOW_ORDER.index(window)
+    except ValueError:
+        index = 0
+    return WINDOW_ORDER[(index + 1) % len(WINDOW_ORDER)]
+
+
+def run_interactive(args):
+    try:
+        curses.wrapper(_interactive_loop, args)
+    except KeyboardInterrupt:
+        pass
+    except curses.error as exc:
+        print(
+            f"Terminal-Dashboard nicht möglich ({exc}); "
+            f"gebe Momentaufnahme aus.",
+            file=sys.stderr,
+        )
+        return emit_once(args)
+
+    return 0
+
+
+def _interactive_loop(stdscr, args):
+    try:
+        curses.curs_set(0)
+    except curses.error:
+        pass
+
+    stdscr.timeout(max(50, int(args.interval * 1000)))
+
+    entries = []
+    error = None
+    loaded_at = 0.0
+    window = args.window
+
+    while True:
+        monotonic_now = time.monotonic()
+
+        if monotonic_now - loaded_at >= args.interval:
+            try:
+                entries = load_entries(args.file)
+                error = None
+            except StatsDataError as exc:
+                error = str(exc)
+            loaded_at = monotonic_now
+
+        height, width = stdscr.getmaxyx()
+        lines = render_dashboard(
+            entries,
+            window=window,
+            now=datetime.now(),
+            width=width,
+            rows=height,
+            source=args.file,
+            error=error,
+            blocks=_blocks_for(args),
+        )
+
+        stdscr.erase()
+        for index, line in enumerate(lines[:height]):
+            attributes = curses.A_BOLD if index == 0 else 0
+            try:
+                stdscr.addnstr(index, 0, line, max(0, width - 1), attributes)
+            except curses.error:
+                pass
+        stdscr.refresh()
+
+        key = stdscr.getch()
+
+        if key in (ord("q"), 27):
+            return
+        if key == ord("r"):
+            loaded_at = 0.0
+        elif key == ord("t"):
+            window = next_window(window)
+
+
+def main(argv=None):
+    parser = argparse.ArgumentParser(
+        description=(
+            "Terminal-Dashboard für den Statserver "
+            "(liest web/page_counts.json)."
+        ),
+    )
+    parser.add_argument(
+        "--file",
+        default=str(DB_DATA),
+        help=f"Statistikdatei (Default: {DB_DATA})",
+    )
+    parser.add_argument(
+        "--window",
+        choices=WINDOW_ORDER,
+        default=DEFAULT_WINDOW,
+        help="Zeitfenster für Verlauf und Tabelle (Default: 24h)",
+    )
+    parser.add_argument(
+        "--once",
+        action="store_true",
+        help="nur einmal ausgeben und beenden",
+    )
+    parser.add_argument(
+        "--width",
+        type=int,
+        default=None,
+        help="Breite für --once (Default: Terminalbreite)",
+    )
+    parser.add_argument(
+        "--interval",
+        type=float,
+        default=REFRESH_SECONDS,
+        help=f"Reload-Intervall in Sekunden (Default: {REFRESH_SECONDS})",
+    )
+    parser.add_argument(
+        "--ascii",
+        action="store_true",
+        help="ASCII-Sparkline statt Unicode-Blöcke",
+    )
+
+    args = parser.parse_args(argv)
+
+    try:
+        locale.setlocale(locale.LC_ALL, "")
+    except locale.Error:
+        pass
+
+    if args.once or not (sys.stdin.isatty() and sys.stdout.isatty()):
+        return emit_once(args)
+
+    return run_interactive(args)
+
+
+if __name__ == "__main__":
+    sys.exit(main())