{% extends "base.html" %}
{% block title %}Samples · FreeWeight{% endblock %}
{% block content %}
{% if error %}
<h2>Samples</h2>
<div class="error-state" role="alert">
  <p><strong>These samples could not be shown.</strong></p>
  <p>{{ error }}</p>
  <p class="muted"><a href="/runs">Back to all runs</a></p>
</div>
{% else %}
<h2>{{ test.test_name }} <span class="muted">{{ test.test_key }}</span></h2>
<p class="muted">
  Run <a href="/runs/{{ detail.run.id }}"><code>{{ detail.run.id[:10] }}…</code></a> ·
  <span class="status status-{{ test.status }}">{{ test.status }}</span> ·
  {{ test.completed_cases }} / {{ test.total_cases }} cases × {{ test.repetitions }} repetitions
</p>
{% if samples %}
<table>
  <caption class="muted">
    {{ samples | length }} sample(s). A sample that could not be scored shows — for its score and
    keeps its row: it is excluded from the aggregate, not from the count. A skipped sample was
    never sent — its error column says why.
  </caption>
  <thead>
    <tr>
      <th scope="col">Case</th>
      <th scope="col">Rep</th>
      <th scope="col">Status</th>
      <th scope="col">Prompt</th>
      <th scope="col">Score</th>
      <th scope="col">Method</th>
      <th scope="col">Output tokens</th>
      <th scope="col">Output chars</th>
      <th scope="col">TTFT (ms)</th>
      <th scope="col">Wall (ms)</th>
      <th scope="col">Finish</th>
      <th scope="col">Error</th>
      <th scope="col">Inspect</th>
    </tr>
  </thead>
  <tbody>
    {% for sample in samples %}
    <tr>
      <td><code>{{ sample.case_id }}</code></td>
      <td class="numeric">{{ sample.repetition }}</td>
      <td><span class="status status-{{ sample.status }}">{{ sample.status | replace("_", " ") }}</span></td>
      <td>
        {% if sample.prompt_id %}
        <code>{{ sample.prompt_id }}</code> <span class="muted">{{ sample.prompt_version }}</span>
        {% else %}<span class="muted" title="This suite's cases carry literal text.">—</span>{% endif %}
      </td>
      <td class="numeric">
        {% if sample.score is none %}
        <span class="muted" title="{{ sample.error_code or 'not scored' }}">—</span>
        {% else %}{{ "%.3f" | format(sample.score) }}{% endif %}
      </td>
      <td>{{ sample.score_method or "—" }}</td>
      <td class="numeric">{{ sample.output_tokens if sample.output_tokens is not none else "—" }}</td>
      <td class="numeric">{{ sample.output_chars if sample.output_chars is not none else "—" }}</td>
      <td class="numeric">
        {% if sample.client_ttft_ms is none %}
        <span class="muted" title="Only a streamed call has a first-token moment to observe.">—</span>
        {% else %}{{ "%.1f" | format(sample.client_ttft_ms) }}{% endif %}
      </td>
      <td class="numeric">{{ "%.1f" | format(sample.client_wall_ms) if sample.client_wall_ms is not none else "—" }}</td>
      <td>{{ sample.finish_reason or "—" }}</td>
      <td>{{ sample.error_code or "—" }}</td>
      {# The second of the two interactions UI standards §13 allows between a headline metric and
         its raw record: the case inspector shows this sample's prompt identity, its response, its
         tool calls, its per-criterion scoring and the telemetry observed while it ran. #}
      <td><a href="/results/samples/{{ sample.id }}">inspect</a></td>
    </tr>
    {% endfor %}
  </tbody>
</table>
{% else %}
<div class="empty-state">
  <p><strong>No samples yet.</strong></p>
  <p>This test has not produced a sample. It is real and empty, not missing.</p>
</div>
{% endif %}
{% endif %}
{% endblock %}