refactor: 评估系统重构 - 前端 API 及评估组件适配

- API 层方法名和路径统一为 dataset/run 语义
- RAGEvaluationTab 重写为全屏面板、支持列宽拖拽
- EvaluationBenchmarks 重写为数据集管理视图
- 生成模态框增加并发数配置；上传模态框适配新接口

web/src/apis/knowledge_api.js

@@ -423,141 +423,64 @@ export const typeApi = {
   }
 }
 
-
 // =============================================================================
 // === RAG评估分组 ===
 // =============================================================================
 
 export const evaluationApi = {
-  /**
-   * 上传评估基准文件
-   * @param {string} dbId - 知识库ID
-   * @param {File} file - JSONL文件
-   * @param {Object} metadata - 基准元数据
-   * @returns {Promise} - 上传结果
-   */
-  uploadBenchmark: async (dbId, file, metadata = {}) => {
+  uploadDataset: async (dbId, file, metadata = {}) => {
     const formData = new FormData()
     formData.append('file', file)
     formData.append('name', metadata.name || '')
     formData.append('description', metadata.description || '')
 
-    return apiAdminPost(`/api/evaluation/databases/${dbId}/benchmarks/upload`, formData)
+    return apiAdminPost(`/api/evaluation/databases/${dbId}/datasets/upload`, formData)
   },
 
-  /**
-   * 获取评估基准列表
-   * @param {string} dbId - 知识库ID
-   * @returns {Promise} - 基准列表
-   */
-  getBenchmarks: async (dbId) => {
-    return apiAdminGet(`/api/evaluation/databases/${dbId}/benchmarks`)
+  listDatasets: async (dbId) => {
+    return apiAdminGet(`/api/evaluation/databases/${dbId}/datasets`)
   },
 
-  /**
-   * 获取评估基准详情
-   * @param {string} benchmarkId - 基准ID
-   * @returns {Promise} - 基准详情
-   */
-  getBenchmark: async (benchmarkId) => {
-    return apiAdminGet(`/api/evaluation/benchmarks/${benchmarkId}`)
-  },
-  /**
-   * 获取评估基准详情（带db_id）
-   * @param {string} dbId - 知识库ID
-   * @param {string} benchmarkId - 基准ID
-   */
-  getBenchmarkByDb: async (dbId, benchmarkId, page = 1, pageSize = 50) => {
+  getDataset: async (dbId, datasetId, page = 1, pageSize = 50) => {
     const params = new URLSearchParams({
       page: page.toString(),
       page_size: pageSize.toString()
     })
-    return apiAdminGet(`/api/evaluation/databases/${dbId}/benchmarks/${benchmarkId}?${params}`)
+    return apiAdminGet(`/api/evaluation/databases/${dbId}/datasets/${datasetId}?${params}`)
   },
 
-  /**
-   * 删除评估基准
-   * @param {string} benchmarkId - 基准ID
-   * @returns {Promise} - 删除结果
-   */
-  deleteBenchmark: async (benchmarkId) => {
-    return apiAdminDelete(`/api/evaluation/benchmarks/${benchmarkId}`)
+  deleteDataset: async (datasetId) => {
+    return apiAdminDelete(`/api/evaluation/datasets/${datasetId}`)
   },
 
-  /**
-   * 下载评估基准
-   * @param {string} benchmarkId - 基准ID
-   * @returns {Promise} - Response对象
-   */
-  downloadBenchmark: async (benchmarkId) => {
-    return apiAdminGet(`/api/evaluation/benchmarks/${benchmarkId}/download`, {}, 'blob')
+  downloadDataset: async (datasetId) => {
+    return apiAdminGet(`/api/evaluation/datasets/${datasetId}/download`, {}, 'blob')
   },
 
-  /**
-   * 自动生成评估基准
-   * @param {string} dbId - 知识库ID
-   * @param {Object} params - 生成参数
-   * @param {number} params.count - 生成问题数量
-   * @param {string} params.llm_model_spec - LLM模型配置
-   * @returns {Promise} - 生成结果
-   */
-  generateBenchmark: async (dbId, params) => {
-    return apiAdminPost(`/api/evaluation/databases/${dbId}/benchmarks/generate`, params)
+  generateDataset: async (dbId, params) => {
+    return apiAdminPost(`/api/evaluation/databases/${dbId}/datasets/generate`, params)
   },
 
-  /**
-   * 运行RAG评估
-   * @param {string} dbId - 知识库ID
-   * @param {Object} params - 评估参数
-   * @param {string} params.benchmark_id - 基准ID
-   * @param {Object} params.retrieval_config - 检索配置
-   * @returns {Promise} - 评估任务ID
-   */
   runEvaluation: async (dbId, params) => {
-    return apiAdminPost(`/api/evaluation/databases/${dbId}/run`, params)
+    return apiAdminPost(`/api/evaluation/databases/${dbId}/runs`, params)
   },
 
-  /**
-   * 获取评估结果
-   * @param {string} taskId - 任务ID
-   * @returns {Promise} - 评估结果
-   */
-  getEvaluationResults: async (taskId) => {
-    // 已废弃：请改用 getEvaluationResultsByDb
-    return apiAdminGet(`/api/evaluation/${taskId}/results`)
+  listRuns: async (dbId) => {
+    return apiAdminGet(`/api/evaluation/databases/${dbId}/runs`)
   },
 
-  /**
-   * 删除评估结果
-   * @param {string} taskId - 任务ID
-   * @returns {Promise} - 删除结果
-   */
-  deleteEvaluationResult: async (taskId) => {
-    // 已废弃：请改用 deleteEvaluationResultByDb
-    return apiAdminDelete(`/api/evaluation/${taskId}`)
-  },
-
-  // 新接口：带 db_id 的评估结果查询与删除
-  getEvaluationResultsByDb: async (dbId, taskId, params = {}) => {
+  getRunResults: async (dbId, runId, params = {}) => {
     const queryParams = new URLSearchParams()
 
     if (params.page) queryParams.append('page', params.page)
     if (params.pageSize) queryParams.append('page_size', params.pageSize)
     if (params.errorOnly !== undefined) queryParams.append('error_only', params.errorOnly)
 
-    const url = `/api/evaluation/databases/${dbId}/results/${taskId}${queryParams.toString() ? '?' + queryParams.toString() : ''}`
+    const url = `/api/evaluation/databases/${dbId}/runs/${runId}${queryParams.toString() ? '?' + queryParams.toString() : ''}`
     return apiAdminGet(url)
   },
-  deleteEvaluationResultByDb: async (dbId, taskId) => {
-    return apiAdminDelete(`/api/evaluation/databases/${dbId}/results/${taskId}`)
-  },
 
-  /**
-   * 获取知识库的评估历史记录
-   * @param {string} dbId - 知识库ID
-   * @returns {Promise} - 评估历史列表
-   */
-  getEvaluationHistory: async (dbId) => {
-    return apiAdminGet(`/api/evaluation/databases/${dbId}/history`)
+  deleteRun: async (dbId, runId) => {
+    return apiAdminDelete(`/api/evaluation/databases/${dbId}/runs/${runId}`)
   }
 }

web/src/components/modals/BenchmarkGenerateModal.vue

@@ -4,6 +4,10 @@
     title="自动生成评估基准"
     width="600px"
     :confirmLoading="generating"
+    :ok-button-props="{ disabled: generating }"
+    :cancel-button-props="{ disabled: generating }"
+    :mask-closable="!generating"
+    :closable="!generating"
     @ok="handleGenerate"
     @cancel="handleCancel"
   >
@@ -54,6 +58,23 @@
               />
             </a-form-item>
           </a-col>
+          <a-col :span="12">
+            <a-form-item
+              label="构建并发数"
+              name="concurrency_count"
+              :labelCol="{ span: 24 }"
+              :wrapperCol="{ span: 24 }"
+              extra="同时生成评估题目的 worker 数，过高可能触发模型服务限流"
+            >
+              <a-input-number
+                v-model:value="formState.concurrency_count"
+                :min="1"
+                :max="20"
+                style="width: 100%"
+                placeholder="默认 10"
+              />
+            </a-form-item>
+          </a-col>
         </a-row>
       </a-form-item>
 
@@ -113,6 +134,7 @@ const formState = reactive({
   description: '',
   count: 10,
   neighbors_count: 1,
+  concurrency_count: 10,
   llm_model_spec: configStore.config?.default_model || ''
 })
 
@@ -122,7 +144,8 @@ const rules = {
     { required: true, message: '请输入基准名称', trigger: 'blur' },
     { min: 2, max: 100, message: '基准名称长度应在2-100个字符之间', trigger: 'blur' }
   ],
-  count: [{ required: true, message: '请输入生成问题数量', trigger: 'blur' }]
+  count: [{ required: true, message: '请输入生成问题数量', trigger: 'blur' }],
+  concurrency_count: [{ required: true, message: '请输入构建并发数', trigger: 'blur' }]
 }
 
 // 双向绑定visible
@@ -149,6 +172,8 @@ const extraText = computed(() =>
 
 // 生成基准
 const handleGenerate = async () => {
+  if (generating.value) return
+
   try {
     // 表单验证
     await formRef.value.validate()
@@ -160,28 +185,31 @@ const handleGenerate = async () => {
       description: formState.description,
       count: formState.count,
       neighbors_count: formState.neighbors_count,
+      concurrency_count: formState.concurrency_count,
       llm_model_spec: formState.llm_model_spec
     }
 
-    const response = await evaluationApi.generateBenchmark(props.databaseId, params)
+    const response = await evaluationApi.generateDataset(props.databaseId, params)
 
     if (response.message === 'success') {
-      message.success('生成任务已提交，请稍后查看结果')
-      handleCancel()
+      message.success('生成任务已提交')
+      visible.value = false
+      resetForm()
       emit('success')
     } else {
+      generating.value = false
       message.error(response.message || '生成失败')
     }
   } catch (error) {
     console.error('生成失败:', error)
-    message.error('生成失败')
-  } finally {
     generating.value = false
+    message.error('生成失败')
   }
 }
 
 // 取消操作
 const handleCancel = () => {
+  if (generating.value) return
   visible.value = false
   resetForm()
 }
@@ -194,6 +222,7 @@ const resetForm = () => {
     description: '',
     count: 10,
     neighbors_count: 1,
+    concurrency_count: 10,
     llm_model_spec: configStore.config?.default_model || ''
   })
   generating.value = false
@@ -206,7 +235,7 @@ const handleSelectLLMModel = (modelSpec) => {
 
 // 监听visible变化
 watch(visible, (val) => {
-  if (val) {
+  if (val && !generating.value) {
     resetForm()
   }
 })

web/src/components/modals/BenchmarkUploadModal.vue

@@ -172,7 +172,7 @@ const handleUpload = async () => {
 
     uploading.value = true
 
-    const response = await evaluationApi.uploadBenchmark(props.databaseId, formState.file, {
+    const response = await evaluationApi.uploadDataset(props.databaseId, formState.file, {
       name: formState.name,
       description: formState.description
     })