{# Settings → Benchmark: most recent runs and a link to the CLI.
   Read-only in v1; the runner is CLI-driven. #}
<section id="benchmark" aria-labelledby="benchmark-heading">
<h2 id="benchmark-heading">Benchmark</h2>
<p class="muted">Run the built-in benchmark corpus through the pipeline
   to measure transcription accuracy (WER) and throughput. Compare results
   across configuration changes or hardware upgrades.</p>

{% if benchmark_runs %}
<table class="runs-table" role="grid">
  <thead>
    <tr>
      <th>Tag</th>
      <th>Status</th>
      <th>WER</th>
      <th>Time</th>
      <th>Date</th>
    </tr>
  </thead>
  <tbody>
  {% for run in benchmark_runs %}
    <tr>
      <td>{{ run.tag or '—' }}</td>
      <td>{{ run.status }}</td>
      <td>
        {% if run.summary and run.summary.pooled_wer is not none %}
          {{ "%.1f%%"|format(run.summary.pooled_wer * 100) }}
        {% else %}
          —
        {% endif %}
      </td>
      <td>
        {% if run.summary and run.summary.total_time_s is not none %}
          {{ "%.0f"|format(run.summary.total_time_s) }}s
        {% else %}
          —
        {% endif %}
      </td>
      <td>{{ run.created_at.strftime("%Y-%m-%d %H:%M") if run.created_at else '—' }}</td>
    </tr>
  {% endfor %}
  </tbody>
</table>
{% else %}
<p>No benchmark runs yet.</p>
{% endif %}

<p class="muted">Run from the command line:
   <code>voxint benchmark run</code> (full corpus) or
   <code>voxint benchmark run --quick</code> (speech files only).
   Compare two runs: <code>voxint benchmark compare ID1 ID2</code>.</p>
</section>