YG_FT/frontend/src/views/eval/EvalDetailView.vue

<script setup lang="ts">
import { computed, onMounted, onUnmounted, ref } from 'vue'
import { useRoute } from 'vue-router'
import PageCard from '@/components/PageCard.vue'
import ModelStatusTag from '@/components/ModelStatusTag.vue'
import { getEvalDetail } from '@/api/modules/eval'
import { usePolling } from '@/composables/usePolling'
import type { EvalSampleResult, EvalTaskDetail } from '@/types'

const route = useRoute()
const taskId = route.params.id as string

const loading = ref(true)
const loadError = ref('')
const detail = ref<EvalTaskDetail | null>(null)
const keyword = ref('')
const judgementFilter = ref('')
const currentPage = ref(1)
const pageSize = ref(10)
const ACTIVE_STATUSES = new Set(['pending', 'queued', 'running'])

const filteredSamples = computed(() => {
  const normalizedKeyword = keyword.value.trim().toLowerCase()
  return (detail.value?.samples || []).filter((sample) => {
    const matchesJudgement = !judgementFilter.value || sample.judgement === judgementFilter.value
    const searchableText = [
      sample.input,
      sample.reference_answer,
      sample.model_output,
      sample.evaluation_reason,
      sample.error_type,
    ].filter(Boolean).join(' ').toLowerCase()
    return matchesJudgement && (!normalizedKeyword || searchableText.includes(normalizedKeyword))
  })
})

const paginatedSamples = computed(() => {
  const start = (currentPage.value - 1) * pageSize.value
  return filteredSamples.value.slice(start, start + pageSize.value)
})

const completionRate = computed(() => {
  const total = detail.value?.sample_count || 0
  return total ? Math.round(((detail.value?.completed_count || 0) / total) * 100) : 0
})

const passRate = computed(() => {
  const completed = detail.value?.completed_count || 0
  return completed ? Math.round(((detail.value?.passed_count || 0) / completed) * 100) : 0
})

const overallScore = computed(() => formatScore(detail.value?.overall_score, detail.value?.overall_score_max))

function formatDateTime(value?: string) {
  if (!value) return '-'
  const date = new Date(value)
  return Number.isNaN(date.getTime())
    ? value
    : date.toLocaleString('zh-CN', { hour12: false })
}

function formatScore(score?: number | null, maxScore?: number) {
  if (score == null) return '-'
  const decimals = Number.isInteger(score) ? 0 : 2
  return `${score.toFixed(decimals)}${maxScore != null ? ` / ${maxScore}` : ''}`
}

function judgementTagType(judgement?: EvalSampleResult['judgement']) {
  if (judgement === '正确') return 'success'
  if (judgement === '部分正确') return 'warning'
  if (judgement === '错误') return 'danger'
  return 'info'
}

function resetPage() {
  currentPage.value = 1
}

async function loadDetail(options: { silent?: boolean } = {}) {
  if (!options.silent) loading.value = true
  loadError.value = ''
  try {
    detail.value = await getEvalDetail(taskId)
  } catch {
    detail.value = null
    loadError.value = '评测详情加载失败，请稍后重试。'
  } finally {
    if (!options.silent) loading.value = false
  }
}

const { start: startPolling, stop: stopPolling } = usePolling(
  async () => {
    await loadDetail({ silent: true })
    if (!ACTIVE_STATUSES.has(String(detail.value?.status || ''))) {
      stopPolling()
    }
  },
  5000,
  { immediate: false },
)

onMounted(async () => {
  await loadDetail()
  if (ACTIVE_STATUSES.has(String(detail.value?.status || ''))) {
    startPolling()
  }
})

onUnmounted(stopPolling)
</script>

<template>
  <PageCard class="eval-detail-page" v-loading="loading">
    <template #header>
      <div class="page-heading">
        <div class="heading-copy">
          <div class="heading-row">
            <h1>{{ detail?.eval_task_name || '评测任务详情' }}</h1>
            <ModelStatusTag v-if="detail" :status="detail.status" />
          </div>
          <dl class="task-meta">
            <div><dt>任务 ID</dt><dd>{{ detail?.id || taskId }}</dd></div>
            <div><dt>评测模型</dt><dd>{{ detail?.model_name || '-' }}</dd></div>
            <div><dt>测试集</dt><dd>{{ detail?.dataset || '-' }}</dd></div>
            <div><dt>评测指标</dt><dd>{{ detail?.metric || '-' }}</dd></div>
          </dl>
        </div>
      </div>
    </template>

    <div v-if="loadError" class="load-state is-error" role="alert">
      <i class="fa fa-exclamation-circle" aria-hidden="true" />
      <h2>无法加载评测详情</h2>
      <p>{{ loadError }}</p>
      <el-button type="primary" @click="() => loadDetail()">重新加载</el-button>
    </div>

    <template v-else-if="detail">
      <section class="overview-grid" aria-label="评测概览">
        <div class="overview-item score-hero">
          <span>综合得分</span>
          <strong>{{ overallScore }}</strong>
          <small>大模型综合评分</small>
        </div>
        <div class="overview-item">
          <span>样本通过率</span>
          <strong>{{ detail.completed_count ? `${passRate}%` : '-' }}</strong>
          <small>通过 {{ detail.passed_count || 0 }} / 已评测 {{ detail.completed_count }}</small>
        </div>
        <div class="overview-item">
          <span>评测进度</span>
          <strong>{{ detail.completed_count }} / {{ detail.sample_count }}</strong>
          <el-progress :percentage="completionRate" :show-text="false" :stroke-width="5" />
        </div>
        <div class="overview-item overview-time">
          <span>{{ detail.completed_time ? '完成时间' : '创建时间' }}</span>
          <strong>{{ formatDateTime(detail.completed_time || detail.create_time) }}</strong>
          <small>{{ detail.completed_time ? '评测任务已完成' : `当前进度 ${completionRate}%` }}</small>
        </div>
      </section>

      <section class="overall-review" aria-labelledby="overall-review-title">
        <div class="review-copy">
          <div class="section-heading">
            <div>
              <h2 id="overall-review-title">大模型综合评价</h2>
              <p>基于全部已评测样本生成的总体结论</p>
            </div>
            <el-tag v-if="detail.evaluator_model" type="primary" size="small">
              评审模型：{{ detail.evaluator_model }}
            </el-tag>
          </div>
          <p class="review-text">
            {{ detail.overall_evaluation || (detail.status === 'running' ? '评测仍在进行，综合评价将在样本评分完成后生成。' : '暂无综合评价。') }}
          </p>

          <div class="suggestion-block">
            <h3>改进建议</h3>
            <ul v-if="detail.improvement_suggestions?.length" class="improvement-list">
              <li v-for="suggestion in detail.improvement_suggestions" :key="suggestion">
                {{ suggestion }}
              </li>
            </ul>
            <p v-else class="empty-copy">暂无改进建议。</p>
          </div>
        </div>
      </section>

      <section v-if="detail.dimension_summary?.length" class="dimension-summary" aria-labelledby="dimension-title">
        <div class="section-heading compact-heading">
          <div>
            <h2 id="dimension-title">维度表现</h2>
            <p>查看各评测维度的得分与样本通过率</p>
          </div>
        </div>
        <div class="dimension-grid">
          <div v-for="dimension in detail.dimension_summary" :key="dimension.name" class="dimension-item">
            <div class="dimension-label">
              <strong>{{ dimension.name }}</strong>
              <span>{{ formatScore(dimension.score, dimension.max_score) }}</span>
            </div>
            <el-progress :percentage="dimension.pass_rate" :stroke-width="7" />
          </div>
        </div>
      </section>

      <section class="sample-section" aria-labelledby="sample-results-title">
        <div class="sample-toolbar">
          <div>
            <h2 id="sample-results-title">测试集样本评测结果</h2>
            <p>共 {{ filteredSamples.length }} 条结果，展开行可查看评分依据与子维度分数</p>
          </div>
          <div class="sample-filters" aria-label="样本筛选">
            <el-input
              v-model="keyword"
              clearable
              placeholder="搜索问题、回答或评价"
              aria-label="搜索样本"
              @input="resetPage"
            >
              <template #prefix><i class="fa fa-search" aria-hidden="true" /></template>
            </el-input>
            <el-select
              v-model="judgementFilter"
              clearable
              placeholder="全部判定"
              aria-label="按判定筛选"
              @change="resetPage"
            >
              <el-option label="正确" value="正确" />
              <el-option label="部分正确" value="部分正确" />
              <el-option label="错误" value="错误" />
            </el-select>
          </div>
        </div>

        <el-table
          v-if="filteredSamples.length"
          class="sample-results-table"
          :data="paginatedSamples"
          row-key="id"
          table-layout="fixed"
        >
          <el-table-column type="expand" width="48">
            <template #default="{ row }">
              <div class="sample-detail-grid">
                <div class="evaluation-reason">
                  <span>大模型评分依据</span>
                  <p>{{ row.evaluation_reason || '暂无评分依据。' }}</p>
                </div>
                <div v-if="row.error_type" class="error-type">
                  <span>错误类型</span>
                  <el-tag type="danger" size="small">{{ row.error_type }}</el-tag>
                </div>
                <div v-if="row.dimension_scores?.length" class="sample-dimensions">
                  <span>子维度评分</span>
                  <div class="dimension-score-list">
                    <span v-for="item in row.dimension_scores" :key="item.name">
                      {{ item.name }} <strong>{{ formatScore(item.score, item.max_score) }}</strong>
                    </span>
                  </div>
                </div>
              </div>
            </template>
          </el-table-column>
          <el-table-column label="#" prop="index" width="56" align="center" />
          <el-table-column label="问题 / 输入" min-width="180">
            <template #default="{ row }"><p class="cell-copy">{{ row.input }}</p></template>
          </el-table-column>
          <el-table-column label="标准答案" min-width="170">
            <template #default="{ row }"><p class="cell-copy">{{ row.reference_answer || '-' }}</p></template>
          </el-table-column>
          <el-table-column label="模型回答" min-width="190">
            <template #default="{ row }"><p class="cell-copy">{{ row.model_output || '等待生成' }}</p></template>
          </el-table-column>
          <el-table-column label="得分" width="90" align="center">
            <template #default="{ row }">
              <span class="sample-score">{{ formatScore(row.score, row.max_score) }}</span>
            </template>
          </el-table-column>
          <el-table-column label="判定" width="96" align="center">
            <template #default="{ row }">
              <el-tag :type="judgementTagType(row.judgement)" size="small">
                {{ row.judgement || (row.status === 'pending' ? '待评测' : '-') }}
              </el-tag>
            </template>
          </el-table-column>
        </el-table>

        <div v-else class="load-state sample-empty">
          <i class="fa fa-list-alt" aria-hidden="true" />
          <h3>{{ detail.samples.length ? '没有符合筛选条件的样本' : '暂无样本评测结果' }}</h3>
          <p>{{ detail.status === 'running' ? '任务正在运行，结果生成后会显示在这里。' : '请调整筛选条件或稍后重试。' }}</p>
        </div>

        <el-pagination
          v-if="filteredSamples.length > pageSize"
          v-model:current-page="currentPage"
          v-model:page-size="pageSize"
          background
          layout="total, sizes, prev, pager, next"
          :page-sizes="[10, 20, 50]"
          :total="filteredSamples.length"
          aria-label="样本结果分页"
        />
      </section>
    </template>
  </PageCard>
</template>

<style scoped lang="scss">
.eval-detail-page {
  min-width: 0;
}

.page-heading,
.heading-row,
.section-heading,
.sample-toolbar,
.sample-filters,
.dimension-label {
  display: flex;
  align-items: center;
}

.heading-copy {
  width: 100%;
  min-width: 0;
}

.heading-row {
  gap: 10px;
  flex-wrap: wrap;
}

.heading-row h1 {
  margin: 0;
  color: #303133;
  font-size: 18px;
  font-weight: 600;
  line-height: 1.4;
}

.task-meta {
  display: flex;
  flex-wrap: wrap;
  gap: 8px 24px;
  margin: 10px 0 0;
}

.task-meta > div {
  display: flex;
  gap: 6px;
  min-width: 0;
}

.task-meta dt,
.task-meta dd {
  margin: 0;
  font-size: 13px;
}

.task-meta dt {
  color: #909399;
}

.task-meta dd {
  color: #606266;
  font-weight: 500;
}

.overview-grid {
  display: grid;
  grid-template-columns: repeat(4, minmax(0, 1fr));
  border: 1px solid #e4e7ed;
  border-radius: 8px;
  overflow: hidden;
  background: #fff;
}

.overview-item {
  min-width: 0;
  min-height: 92px;
  padding: 16px 20px;
  border-right: 1px solid #e4e7ed;
}

.overview-item:last-child {
  border-right: 0;
}

.overview-item > span,
.evaluation-reason > span,
.error-type > span,
.sample-dimensions > span {
  display: block;
  margin-bottom: 8px;
  color: #606266;
  font-size: 13px;
  font-weight: 500;
}

.overview-item > strong {
  display: block;
  color: #303133;
  font-size: 22px;
  font-weight: 600;
  line-height: 1.3;
  font-variant-numeric: tabular-nums;
}

.score-hero > strong {
  color: var(--primary-color);
}

.overview-item small {
  display: block;
  margin-top: 5px;
  color: #909399;
  font-size: 12px;
}

.overview-item :deep(.el-progress) {
  margin-top: 8px;
}

.overview-time > strong {
  font-size: 14px;
  font-weight: 500;
  line-height: 1.6;
}

.overall-review {
  margin-top: 16px;
  border: 1px solid #e4e7ed;
  border-radius: 8px;
  background: #fff;
}

.review-copy {
  padding: 18px 20px;
}

.section-heading,
.sample-toolbar {
  justify-content: space-between;
  gap: 16px;
}

.section-heading h2,
.sample-toolbar h2 {
  margin: 0;
  color: #303133;
  font-size: 16px;
  font-weight: 600;
}

.section-heading p,
.sample-toolbar p {
  margin: 5px 0 0;
  color: #909399;
  font-size: 13px;
}

.review-text {
  max-width: 1100px;
  margin: 16px 0;
  color: #606266;
  font-size: 14px;
  line-height: 1.75;
}

.suggestion-block {
  padding: 14px 16px;
  border: 1px solid #ebeef5;
  border-radius: 6px;
  background: #f8f9fb;
}

.suggestion-block h3 {
  margin: 0 0 8px;
  color: #303133;
  font-size: 14px;
  font-weight: 600;
}

.improvement-list {
  display: grid;
  grid-template-columns: repeat(2, minmax(0, 1fr));
  gap: 6px 24px;
  margin: 0;
  padding-left: 18px;
  color: #606266;
  font-size: 13px;
  line-height: 1.6;
}

.dimension-summary,
.sample-section {
  margin-top: 16px;
  padding: 18px 20px;
  border: 1px solid #e4e7ed;
  border-radius: 8px;
  background: #fff;
}

.compact-heading {
  align-items: center;
}

.dimension-grid {
  display: grid;
  grid-template-columns: repeat(4, minmax(0, 1fr));
  gap: 0;
  margin-top: 16px;
  border: 1px solid #ebeef5;
  border-radius: 6px;
  overflow: hidden;
}

.dimension-item {
  min-width: 0;
  padding: 14px 16px;
  border-right: 1px solid #ebeef5;
  background: #fafafa;
}

.dimension-item:last-child {
  border-right: 0;
}

.dimension-label {
  justify-content: space-between;
  margin-bottom: 8px;
  gap: 8px;
  color: #606266;
  font-size: 13px;
}

.dimension-label strong {
  overflow: hidden;
  text-overflow: ellipsis;
  white-space: nowrap;
}

.dimension-label span {
  flex: 0 0 auto;
  color: #303133;
  font-weight: 600;
  font-variant-numeric: tabular-nums;
}

.sample-toolbar {
  align-items: center;
  margin-bottom: 14px;
}

.sample-filters {
  flex: 0 1 450px;
  justify-content: flex-end;
  gap: 10px;
}

.sample-filters :deep(.el-input) {
  width: min(260px, 100%);
}

.sample-filters :deep(.el-select) {
  width: 130px;
}

.sample-results-table {
  width: 100%;
  border: 1px solid #ebeef5;
  border-radius: 6px;
}

.cell-copy {
  display: -webkit-box;
  overflow: hidden;
  margin: 0;
  color: #606266;
  font-size: 13px;
  line-height: 1.55;
  -webkit-box-orient: vertical;
  -webkit-line-clamp: 2;
}

.sample-score {
  color: #303133;
  font-weight: 600;
  font-variant-numeric: tabular-nums;
}

.sample-detail-grid {
  display: grid;
  grid-template-columns: minmax(140px, 0.35fr) minmax(320px, 1fr);
  gap: 14px 24px;
  padding: 16px 48px;
  background: #f7f8fa;
}

.evaluation-reason {
  grid-column: 1 / -1;
}

.evaluation-reason p {
  margin: 0;
  color: #606266;
  font-size: 13px;
  line-height: 1.7;
}

.dimension-score-list {
  display: flex;
  flex-wrap: wrap;
  gap: 8px;
}

.dimension-score-list > span {
  color: #606266;
  font-size: 12px;
}

.dimension-score-list > span:not(:last-child)::after {
  margin-left: 8px;
  color: #c0c4cc;
  content: '/';
}

.dimension-score-list strong {
  margin-left: 4px;
  color: #303133;
  font-variant-numeric: tabular-nums;
}

.load-state {
  min-height: 280px;
  display: flex;
  flex-direction: column;
  align-items: center;
  justify-content: center;
  padding: 32px;
  color: #909399;
  text-align: center;
}

.load-state > i {
  margin-bottom: 12px;
  color: #c0c4cc;
  font-size: 30px;
}

.load-state h2,
.load-state h3 {
  margin: 0;
  color: #303133;
}

.load-state p,
.empty-copy {
  margin: 8px 0 16px;
  color: #909399;
  font-size: 13px;
}

.load-state.is-error > i {
  color: #ef4444;
}

.sample-empty {
  min-height: 210px;
  border: 1px dashed #dcdfe6;
  border-radius: 6px;
  background: #fafafa;
}

@media (max-width: 1100px) {
  .overview-grid {
    grid-template-columns: repeat(2, minmax(0, 1fr));
  }

  .overview-item:nth-child(2) {
    border-right: 0;
  }

  .overview-item:nth-child(-n + 2) {
    border-bottom: 1px solid #e4e7ed;
  }

  .dimension-grid {
    grid-template-columns: repeat(2, minmax(0, 1fr));
  }

  .dimension-item:nth-child(2) {
    border-right: 0;
  }

  .dimension-item:nth-child(-n + 2) {
    border-bottom: 1px solid #ebeef5;
  }

  .sample-toolbar {
    align-items: stretch;
    flex-direction: column;
  }

  .sample-filters {
    flex: 1;
    justify-content: flex-start;
  }

  .sample-detail-grid {
    padding: 16px 24px;
  }
}

@media (max-width: 720px) {
  .task-meta {
    flex-direction: column;
    gap: 5px;
  }

  .overview-grid {
    grid-template-columns: repeat(2, minmax(0, 1fr));
  }

  .overview-item {
    padding: 14px;
  }

  .review-copy,
  .dimension-summary,
  .sample-section {
    padding: 16px;
  }

  .dimension-grid,
  .improvement-list,
  .sample-detail-grid {
    grid-template-columns: minmax(0, 1fr);
  }

  .dimension-item,
  .dimension-item:nth-child(2) {
    border-right: 0;
    border-bottom: 1px solid #ebeef5;
  }

  .dimension-item:last-child {
    border-bottom: 0;
  }

  .sample-filters {
    flex-direction: column;
    align-items: stretch;
  }

  .sample-filters :deep(.el-input),
  .sample-filters :deep(.el-select) {
    width: 100%;
  }

  .evaluation-reason {
    grid-column: auto;
  }
}
</style>