{% extends "base.html" %}
{% block title %}Grade run — FreeWeight{% endblock %}
{% block content %}
<h2>Grade run <code>{{ run_id[:8] }}</code></h2>

{% if error %}
<div class="error-state" role="alert"><p>{{ error }}</p></div>
{% endif %}

{% if view %}
<p class="muted">
  <strong>{{ view.goal_name }}</strong> declares {{ view.criteria | length }}
  {{ 'criterion' if view.criteria | length == 1 else 'criteria' }} graded by a person rather than a rule or a jury.
  Blinded and shuffled: which model wrote these is not shown and is not fetched, and the order is
  not the order they were produced in. Every grade is saved the moment you submit it; the run's
  composite and its capability evidence are refreshed with it.
</p>

<dl class="card-grid">
  <div class="card">
    <dt>Graded</dt>
    <dd>{{ view.recorded }} / {{ view.expected }}</dd>
    <p class="card-note">{{ view.samples | length }} samples × {{ view.criteria | length }} human criteria.</p>
  </div>
  <div class="card">
    <dt>Complete</dt>
    <dd style="font-size: 18px;">{{ 'yes' if view.complete else 'not yet' }}</dd>
    <p class="card-note">
      {% if view.complete %}Every sample measures the whole rubric. <a href="/evidence?capability=user.{{ view.goal_slug }}">See the evidence</a>.
      {% else %}Ungraded criteria are excluded from the composite and the applied weight says so.{% endif %}
    </p>
  </div>
</dl>

{% if view.samples %}
{% for sample in view.samples %}
<h3 id="sample-{{ loop.index }}">Sample {{ loop.index }} of {{ view.samples | length }} <span class="unit">task {{ sample.case_id }}</span></h3>
<pre>{{ sample.response_text }}</pre>
<div class="table-scroll">
  <table>
    <caption>Grade this sample on each human criterion. Notes stay beside the grade.</caption>
    <thead>
      <tr>
        <th scope="col">Criterion</th>
        <th scope="col">Your grade</th>
      </tr>
    </thead>
    <tbody>
      {% for criterion in view.criteria %}
      {% set existing = sample.grades.get(criterion.key) %}
      <tr>
        <td>
          <strong>{{ criterion.name }}</strong> <span class="unit">weight {{ '%.2f' % criterion.weight }}</span>
          {% if criterion.descriptors %}
          <details>
            <summary class="unit">what the points mean</summary>
            <dl>
              {% for point, text in criterion.descriptors | dictsort(reverse=true) %}
              <dt>{{ point }}</dt><dd>{{ text }}</dd>
              {% endfor %}
            </dl>
          </details>
          {% endif %}
        </td>
        <td>
          <form method="post" action="/runs/{{ run_id }}/grade">
            <input type="hidden" name="sample_id" value="{{ sample.sample_id }}">
            <input type="hidden" name="criterion" value="{{ criterion.key }}">
            <fieldset style="border: 0; padding: 0;">
              <legend class="unit">Grade</legend>
              {% for point in range(1, criterion.scale_points + 1) %}
              <label for="g-{{ sample.sample_id }}-{{ criterion.key }}-{{ point }}">
                <input type="radio" id="g-{{ sample.sample_id }}-{{ criterion.key }}-{{ point }}"
                       name="grade" value="{{ point }}"
                       {% if existing and existing.grade == point %}checked{% endif %}>
                {{ point }}
              </label>
              {% endfor %}
            </fieldset>
            <div class="field">
              <label for="n-{{ sample.sample_id }}-{{ criterion.key }}">Note</label>
              <input id="n-{{ sample.sample_id }}-{{ criterion.key }}" name="note"
                     value="{{ existing.note if existing else '' }}">
            </div>
            <button type="submit">{{ 'Update grade' if existing else 'Save grade' }}</button>
            {% if existing %}
            <span class="status status-completed">graded {{ existing.grade }}</span>
            {% endif %}
          </form>
        </td>
      </tr>
      {% endfor %}
    </tbody>
  </table>
</div>
{% endfor %}
{% else %}
<div class="empty-state">
  <h3>Nothing to grade</h3>
  <p>This run stored no completed samples with their text. A goal run stores every response by
    default, so a run with none either failed every sample or was cancelled.</p>
</div>
{% endif %}
{% elif not error %}
<div class="empty-state"><p>No run to grade.</p></div>
{% endif %}
{% endblock %}