feat: add embedding and rerank support

frontend/src/mocks/__tests__/embedding-metadata.spec.ts

@@ -0,0 +1,44 @@
+import { describe, expect, it } from 'vitest'
+
+import { MOCK_API_FORMATS, MOCK_GLOBAL_MODELS } from '../data'
+
+describe('embedding mock metadata', () => {
+  it('exposes embedding model metadata to frontend code without chat treatment', () => {
+    const model = MOCK_GLOBAL_MODELS.find(item => item.name === 'text-embedding-3-small')
+
+    expect(model).toMatchObject({
+      supported_capabilities: ['embedding'],
+      supports_embedding: true,
+      config: {
+        streaming: false,
+        embedding: true,
+        model_type: 'embedding',
+        api_formats: ['openai:embedding'],
+      },
+    })
+  })
+
+  it('includes all embedding API formats as distinct catalog formats', () => {
+    const embeddingFormats = MOCK_API_FORMATS.formats
+      .filter(format => format.value.endsWith(':embedding'))
+      .map(format => [format.value, format.label])
+
+    expect(embeddingFormats).toEqual([
+      ['openai:embedding', 'OpenAI Embedding'],
+      ['gemini:embedding', 'Gemini Embedding'],
+      ['jina:embedding', 'Jina Embedding'],
+      ['doubao:embedding', 'Doubao Embedding'],
+    ])
+  })
+
+  it('includes rerank API formats as distinct catalog formats', () => {
+    const rerankFormats = MOCK_API_FORMATS.formats
+      .filter(format => format.value.endsWith(':rerank'))
+      .map(format => [format.value, format.label])
+
+    expect(rerankFormats).toEqual([
+      ['openai:rerank', 'OpenAI Rerank'],
+      ['jina:rerank', 'Jina Rerank'],
+    ])
+  })
+})

frontend/src/mocks/data.ts

@@ -546,20 +546,21 @@ export const MOCK_PROVIDERS: ProviderWithEndpointsSummary[] = [
     billing_type: 'pay_as_you_go',
     monthly_used_usd: 5.29,
     is_active: true,
-    total_endpoints: 4,
-    active_endpoints: 4,
+    total_endpoints: 5,
+    active_endpoints: 5,
     total_keys: 11,
     active_keys: 11,
-    total_models: 8,
-    active_models: 8,
+    total_models: 9,
+    active_models: 9,
     avg_health_score: 0.863,
     unhealthy_endpoints: 1,
-    api_formats: ['claude:messages', 'gemini:generate_content', 'openai:chat', 'openai:responses'],
+    api_formats: ['claude:messages', 'gemini:generate_content', 'openai:chat', 'openai:responses', 'openai:embedding'],
     endpoint_health_details: [
       { api_format: 'claude:messages', health_score: 1.0, is_active: true, active_keys: 2 },
       { api_format: 'gemini:generate_content', health_score: 1.0, is_active: true, active_keys: 2 },
       { api_format: 'openai:chat', health_score: 0.85, is_active: true, active_keys: 2 },
-      { api_format: 'openai:responses', health_score: 1.0, is_active: true, active_keys: 1 }
+      { api_format: 'openai:responses', health_score: 1.0, is_active: true, active_keys: 1 },
+      { api_format: 'openai:embedding', health_score: 0.98, is_active: true, active_keys: 1 }
     ],
     created_at: '2024-12-07T22:56:09.712806+08:00',
     updated_at: new Date().toISOString()
@@ -805,6 +806,46 @@ export const MOCK_GLOBAL_MODELS: GlobalModelResponse[] = [
     },
     provider_count: 2,
     created_at: '2024-01-01T00:00:00Z'
+  },
+  {
+    id: 'gm-010',
+    name: 'text-embedding-3-small',
+    display_name: 'text-embedding-3-small',
+    is_active: true,
+    default_tiered_pricing: {
+      tiers: [{ up_to: null, input_price_per_1m: 0.02, output_price_per_1m: 0 }]
+    },
+    supported_capabilities: ['embedding'],
+    supports_embedding: true,
+    config: {
+      streaming: false,
+      embedding: true,
+      model_type: 'embedding',
+      api_formats: ['openai:embedding'],
+      dimensions: 1536,
+      description: 'OpenAI 文本向量嵌入模型'
+    },
+    provider_count: 1,
+    created_at: '2024-01-01T00:00:00Z'
+  },
+  {
+    id: 'gm-rerank-001',
+    name: 'bge-reranker-base',
+    display_name: 'bge-reranker-base',
+    is_active: true,
+    default_tiered_pricing: {
+      tiers: [{ up_to: null, input_price_per_1m: 0.05, output_price_per_1m: 0 }]
+    },
+    supported_capabilities: ['rerank'],
+    config: {
+      streaming: false,
+      rerank: true,
+      model_type: 'rerank',
+      api_formats: ['openai:rerank'],
+      description: '文本重排序模型'
+    },
+    provider_count: 1,
+    created_at: '2024-01-01T00:00:00Z'
   }
 ]
 
@@ -878,9 +919,15 @@ export const MOCK_API_FORMATS = {
     { value: 'openai:chat', label: 'OpenAI Chat', default_path: '/v1/chat/completions', aliases: [] },
     { value: 'openai:responses', label: 'OpenAI Responses', default_path: '/v1/responses', aliases: [] },
     { value: 'openai:responses:compact', label: 'OpenAI Responses Compact', default_path: '/v1/responses/compact', aliases: [] },
+    { value: 'openai:embedding', label: 'OpenAI Embedding', default_path: '/v1/embeddings', aliases: [] },
+    { value: 'openai:rerank', label: 'OpenAI Rerank', default_path: '/v1/rerank', aliases: [] },
     { value: 'openai:image', label: 'OpenAI Image', default_path: '/v1/images/generations', aliases: [] },
     { value: 'openai:video', label: 'OpenAI Video', default_path: '/v1/videos', aliases: [] },
     { value: 'gemini:generate_content', label: 'Gemini Generate Content', default_path: '/v1beta/models/{model}:{action}', aliases: [] },
-    { value: 'gemini:video', label: 'Gemini Video', default_path: '/v1beta/models/{model}:predictLongRunning', aliases: [] }
+    { value: 'gemini:embedding', label: 'Gemini Embedding', default_path: '/v1beta/models/{model}:embedContent', aliases: [] },
+    { value: 'gemini:video', label: 'Gemini Video', default_path: '/v1beta/models/{model}:predictLongRunning', aliases: [] },
+    { value: 'jina:embedding', label: 'Jina Embedding', default_path: '/v1/embeddings', aliases: [] },
+    { value: 'jina:rerank', label: 'Jina Rerank', default_path: '/v1/rerank', aliases: [] },
+    { value: 'doubao:embedding', label: 'Doubao Embedding', default_path: '/embeddings/multimodal', aliases: [] }
   ]
 }

frontend/src/mocks/handler.ts

@@ -228,6 +228,19 @@ const MOCK_ENDPOINT_STATUS = {
       last_event_at: new Date().toISOString(),
       // 94.0% 成功率：successRate=0.940, failRate=0.043, skipRate=0.017
       events: generateHealthEvents(100, 0.940, 0.043, 0.017, 800, 600)
+    },
+    {
+      api_format: 'openai:embedding',
+      api_path: '/v1/embeddings',
+      total_attempts: 620,
+      success_count: 612,
+      failed_count: 6,
+      skipped_count: 2,
+      success_rate: 0.987,
+      provider_count: 1,
+      key_count: 1,
+      last_event_at: new Date().toISOString(),
+      events: generateHealthEvents(40, 0.987, 0.01, 0.003, 320, 140)
     }
   ]
 }
@@ -447,6 +460,12 @@ function getMockEndpointExtras(apiFormat: string) {
     extras.config = { upstream_stream_policy: 'force_stream' }
   } else if (normalizedFormat === 'openai:responses') {
     extras.config = { upstream_stream_policy: 'force_non_stream' }
+  } else if (normalizedFormat === 'openai:embedding') {
+    extras.custom_path = '/v1/embeddings'
+    extras.config = { route_kind: 'embedding' }
+  } else if (normalizedFormat === 'openai:rerank' || normalizedFormat === 'jina:rerank') {
+    extras.custom_path = '/v1/rerank'
+    extras.config = { route_kind: 'rerank' }
   } else if (normalizedFormat === 'gemini:generate_content') {
     extras.custom_path = '/v1beta/models/gemini-3-pro-preview:generateContent'
     extras.body_rules = [
@@ -720,6 +739,23 @@ const mockHandlers: Record<string, (config: AxiosRequestConfig) => Promise<Axios
     })))
   },
 
+  'GET /api/users/me/available-models': async () => {
+    await delay()
+    const models = MOCK_GLOBAL_MODELS.filter(model => model.is_active).map(model => ({
+      id: model.id,
+      name: model.name,
+      display_name: model.display_name,
+      is_active: model.is_active,
+      default_price_per_request: model.default_price_per_request ?? null,
+      default_tiered_pricing: model.default_tiered_pricing,
+      supported_capabilities: model.supported_capabilities ?? null,
+      supports_embedding: model.supports_embedding ?? null,
+      config: model.config ?? null,
+      usage_count: model.usage_count ?? 0,
+    }))
+    return createMockResponse({ models, total: models.length })
+  },
+
   'GET /api/users/me/preferences': async () => {
     await delay()
     return createMockResponse(getCurrentProfile().preferences || { theme: 'auto', language: 'zh-CN' })
@@ -1146,6 +1182,7 @@ const mockHandlers: Record<string, (config: AxiosRequestConfig) => Promise<Axios
         default_tiered_pricing: m.default_tiered_pricing,
         default_price_per_request: m.default_price_per_request,
         supported_capabilities: m.supported_capabilities,
+        supports_embedding: m.supports_embedding,
         config: m.config
       })),
       total: MOCK_GLOBAL_MODELS.length
@@ -1419,6 +1456,8 @@ function generateMockModelsForProvider(providerId: string) {
   const hasClaude = provider.api_formats.some(f => f.includes('claude'))
   const hasOpenAI = provider.api_formats.some(f => f.includes('openai'))
   const hasGemini = provider.api_formats.some(f => f.includes('gemini'))
+  const hasEmbedding = provider.api_formats.some(f => f.endsWith(':embedding'))
+  const hasRerank = provider.api_formats.some(f => f.endsWith(':rerank'))
 
   const models: Record<string, unknown>[] = []
   const now = new Date().toISOString()
@@ -1503,6 +1542,66 @@ function generateMockModelsForProvider(providerId: string) {
       }
     )
   }
+  if (hasEmbedding) {
+    models.push({
+      id: `pm-${providerId}-embedding-1`,
+      provider_id: providerId,
+      global_model_id: 'gm-010',
+      provider_model_name: 'text-embedding-3-small',
+      global_model_name: 'text-embedding-3-small',
+      global_model_display_name: 'text-embedding-3-small',
+      effective_input_price: 0.02,
+      effective_output_price: 0,
+      supports_embedding: true,
+      effective_supports_embedding: true,
+      supports_streaming: false,
+      effective_supports_streaming: false,
+      config: {
+        embedding: true,
+        model_type: 'embedding',
+        api_formats: ['openai:embedding'],
+      },
+      effective_config: {
+        embedding: true,
+        model_type: 'embedding',
+        api_formats: ['openai:embedding'],
+        streaming: false,
+      },
+      is_active: true,
+      is_available: true,
+      created_at: provider.created_at,
+      updated_at: now
+    })
+  }
+  if (hasRerank) {
+    models.push({
+      id: `pm-${providerId}-rerank-1`,
+      provider_id: providerId,
+      global_model_id: 'gm-rerank-001',
+      provider_model_name: 'bge-reranker-base',
+      global_model_name: 'bge-reranker-base',
+      global_model_display_name: 'bge-reranker-base',
+      effective_input_price: 0.05,
+      effective_output_price: 0,
+      supports_streaming: false,
+      effective_supports_streaming: false,
+      config: {
+        rerank: true,
+        model_type: 'rerank',
+        api_formats: ['openai:rerank'],
+      },
+      effective_config: {
+        rerank: true,
+        model_type: 'rerank',
+        api_formats: ['openai:rerank'],
+        streaming: false,
+      },
+      is_active: true,
+      is_available: true,
+      created_at: provider.created_at,
+      updated_at: now
+    })
+  }
   if (hasGemini) {
     models.push(
       {