feat(stage-ui): add nvidia nim provider (#1196)

* feat: add nvidia nim provider * fix: restore onboarding no-models help text * fix: limit nvidia provider to desktop runtime * [autofix.ci] apply automated fixes * fix: update validators check list Co-authored-by: gemini-code-assist[bot] <176961590+gemini-code-assist[bot]@users.noreply.github.com> --------- Co-authored-by: autofix-ci[bot] <114827586+autofix-ci[bot]@users.noreply.github.com> Co-authored-by: Garfield Lee <Garfield550@users.noreply.github.com> Co-authored-by: gemini-code-assist[bot] <176961590+gemini-code-assist[bot]@users.noreply.github.com>
2026-03-09 20:09:36 +08:00
parent a16606f2de
commit 861beacad2
16 changed files with 225 additions and 65 deletions
@@ -743,6 +743,9 @@ pages:
      openai-compatible:
        description: OpenAI Compatible
        title: OpenAI Compatible
+      nvidia:
+        description: build.nvidia.com
+        title: NVIDIA NIM
      n1n:
        description: n1n.ai - High-performance AI API provider.
        title: n1n
@@ -711,6 +711,9 @@ pages:
      openai-compatible:
        description: OpenAI Compatible
        title: OpenAI Compatible
+      nvidia:
+        description: build.nvidia.com
+        title: NVIDIA NIM
      n1n:
        description: n1n.ai - High-performance AI API provider.
        title: n1n
@@ -711,6 +711,9 @@ pages:
      openai-compatible:
        description: Compatible avec OpenAI
        title: Compatible avec OpenAI
+      nvidia:
+        description: build.nvidia.com
+        title: NVIDIA NIM
      n1n:
        description: n1n.ai - Fournisseur d'API d'IA haute performance.
        title: n1n
@@ -711,6 +711,9 @@ pages:
      openai-compatible:
        description: OpenAI互換
        title: OpenAI互換
+      nvidia:
+        description: build.nvidia.com
+        title: NVIDIA NIM
      n1n:
        description: n1n.ai - 高性能なAI APIプロバイダー。
        title: n1n
@@ -711,6 +711,9 @@ pages:
      openai-compatible:
        description: OpenAI 호환
        title: OpenAI 호환
+      nvidia:
+        description: build.nvidia.com
+        title: NVIDIA NIM
      n1n:
        description: n1n.ai - 고성능 AI API 제공자.
        title: n1n
@@ -711,6 +711,9 @@ pages:
      openai-compatible:
        description: Совместимый с OpenAI
        title: OpenAI Compatible
+      nvidia:
+        description: build.nvidia.com
+        title: NVIDIA NIM
      n1n:
        description: n1n.ai - High-performance AI API provider.
        title: n1n
@@ -711,6 +711,9 @@ pages:
      openai-compatible:
        description: Tương thích OpenAI
        title: OpenAI Compatible
+      nvidia:
+        description: build.nvidia.com
+        title: NVIDIA NIM
      n1n:
        description: n1n.ai - High-performance AI API provider.
        title: n1n
@@ -711,6 +711,9 @@ pages:
      openai-compatible:
        description: OpenAI 兼容 API
        title: OpenAI 兼容 API
+      nvidia:
+        description: build.nvidia.com
+        title: NVIDIA NIM
      n1n:
        description: n1n.ai - 高性能的 AI API 提供商
        title: n1n
@@ -711,6 +711,9 @@ pages:
      openai-compatible:
        description: OpenAI 相容
        title: OpenAI Compatible
+      nvidia:
+        description: build.nvidia.com
+        title: NVIDIA NIM
      n1n:
        description: n1n.ai - High-performance AI API provider.
        title: n1n
@@ -30,7 +30,7 @@ const {

 // Popular providers for first-time setup
 const popularProviders = computed(() => {
-  const popular = ['openai', 'anthropic', 'google-generative-ai', 'groq', 'openrouter-ai', 'ollama', 'deepseek', 'player2', 'openai-compatible']
+  const popular = ['openai', 'anthropic', 'google-generative-ai', 'groq', 'nvidia', 'openrouter-ai', 'ollama', 'deepseek', 'player2', 'openai-compatible']
  return allChatProvidersMetadata.value
    .filter(provider => popular.includes(provider.id))
    .sort((a, b) => popular.indexOf(a.id) - popular.indexOf(b.id))
@@ -35,9 +35,22 @@ const {
    </div>

    <!-- Using the new RadioCardManySelect component -->
-    <div flex-1>
+    <div flex flex-1 flex-col gap-4>
+      <Alert
+        v-if="providerModels.length === 0 && !isLoadingActiveProviderModels"
+        type="error"
+      >
+        <template #title>
+          {{ t('settings.dialogs.onboarding.no-models') }}
+        </template>
+        <template #content>
+          <div class="whitespace-pre-wrap break-all">
+            {{ t('settings.dialogs.onboarding.no-models-help') }}
+          </div>
+        </template>
+      </Alert>
+
      <RadioCardManySelect
-        v-if="providerModels.length > 0"
        v-model="activeModel"
        v-model:search-query="modelSearchQuery"
        :items="providerModels.toSorted((a, b) => a.id === activeModel ? -1 : b.id === activeModel ? 1 : 0)"
@@ -52,17 +65,6 @@ const {
        :collapse-button-text="t('settings.pages.modules.consciousness.sections.section.provider-model-selection.collapse')"
        list-class="max-h-[calc(100dvh-17rem)] sm:max-h-120 overflow-y-auto"
      />
-
-      <Alert v-else type="error">
-        <template #title>
-          {{ t('settings.dialogs.onboarding.no-models') }}
-        </template>
-        <template #content>
-          <div class="whitespace-pre-wrap break-all">
-            {{ t('settings.dialogs.onboarding.no-models-help') }}
-          </div>
-        </template>
-      </Alert>
    </div>

    <!-- Action Buttons -->
@@ -146,6 +146,7 @@ function getApiKeyPlaceholder(providerId: string): string {
    'modelscope': 'ms-...',
    'fireworks-ai': 'fw-...',
    'featherless-ai': 'fw-...',
+    'nvidia': 'nvapi-...',
    'novita-ai': 'nvt-...',
  }

@@ -4,6 +4,7 @@ import './lm-studio'
 import './openai-compatible'
 import './n1n'
 import './openrouter-ai'
+import './nvidia'
 import './groq'
 import './anthropic'
 import './google-generative-ai'
@@ -0,0 +1,54 @@
+import { isStageTamagotchi } from '@proj-airi/stage-shared'
+import { createOpenAI } from '@xsai-ext/providers/create'
+import { z } from 'zod'
+
+import { createOpenAICompatibleValidators } from '../../validators/openai-compatible'
+import { defineProvider } from '../registry'
+
+const nvidiaConfigSchema = z.object({
+  apiKey: z
+    .string('API Key'),
+  baseUrl: z
+    .string('Base URL')
+    .optional()
+    .default('https://integrate.api.nvidia.com/v1/'),
+})
+
+type NvidiaConfig = z.input<typeof nvidiaConfigSchema>
+
+export const providerNvidia = defineProvider<NvidiaConfig>({
+  id: 'nvidia',
+  name: 'NVIDIA NIM',
+  nameLocalize: ({ t }) => t('settings.pages.providers.provider.nvidia.title'),
+  description: 'build.nvidia.com',
+  descriptionLocalize: ({ t }) => t('settings.pages.providers.provider.nvidia.description'),
+  tasks: ['chat'],
+  icon: 'i-simple-icons:nvidia',
+  isAvailableBy: isStageTamagotchi,
+
+  createProviderConfig: ({ t }) => nvidiaConfigSchema.extend({
+    apiKey: nvidiaConfigSchema.shape.apiKey.meta({
+      labelLocalized: t('settings.pages.providers.catalog.edit.config.common.fields.field.api-key.label'),
+      descriptionLocalized: t('settings.pages.providers.catalog.edit.config.common.fields.field.api-key.description'),
+      placeholderLocalized: t('settings.pages.providers.catalog.edit.config.common.fields.field.api-key.placeholder'),
+      type: 'password',
+    }),
+    baseUrl: nvidiaConfigSchema.shape.baseUrl.meta({
+      labelLocalized: t('settings.pages.providers.catalog.edit.config.common.fields.field.base-url.label'),
+      descriptionLocalized: t('settings.pages.providers.catalog.edit.config.common.fields.field.base-url.description'),
+      placeholderLocalized: t('settings.pages.providers.catalog.edit.config.common.fields.field.base-url.placeholder'),
+    }),
+  }),
+  createProvider(config) {
+    return createOpenAI(config.apiKey, config.baseUrl)
+  },
+
+  validationRequiredWhen(config) {
+    return !!config.apiKey?.trim()
+  },
+  validators: {
+    ...createOpenAICompatibleValidators({
+      checks: ['connectivity', 'model_list'],
+    }),
+  },
+})
@@ -0,0 +1,77 @@
+import { beforeEach, describe, expect, it, vi } from 'vitest'
+
+import { createOpenAICompatibleValidators } from './openai-compatible'
+
+const {
+  generateTextMock,
+  listModelsMock,
+} = vi.hoisted(() => ({
+  generateTextMock: vi.fn(),
+  listModelsMock: vi.fn(),
+}))
+
+vi.mock('@xsai/generate-text', () => ({
+  generateText: generateTextMock,
+}))
+
+vi.mock('@xsai/model', () => ({
+  listModels: listModelsMock,
+}))
+
+function getProviderValidators(options?: Parameters<typeof createOpenAICompatibleValidators>[0]) {
+  const validators = createOpenAICompatibleValidators(options)
+
+  return (validators?.validateProvider || []).map(create => create({
+    t: (input: string) => input,
+  } as any))
+}
+
+describe('createOpenAICompatibleValidators', () => {
+  const config = {
+    apiKey: 'test-key',
+    baseUrl: 'https://example.com/v1/',
+  }
+  const provider = {
+    model: () => ({
+      apiKey: config.apiKey,
+      baseURL: config.baseUrl,
+    }),
+  }
+
+  beforeEach(() => {
+    vi.clearAllMocks()
+  })
+
+  it('does not probe chat completions with a synthetic fallback model', async () => {
+    listModelsMock.mockResolvedValue([])
+
+    const [connectivityValidator, chatValidator] = getProviderValidators({
+      checks: ['connectivity', 'chat_completions'],
+    })
+
+    const connectivityResult = await connectivityValidator.validator(config, provider as any, undefined as any, undefined as any)
+    const chatResult = await chatValidator.validator(config, provider as any, undefined as any, undefined as any)
+
+    expect(connectivityResult.valid).toBe(false)
+    expect(connectivityResult.reason).toContain('No model available for validation.')
+    expect(chatResult.valid).toBe(false)
+    expect(chatResult.reason).toContain('No model available for validation.')
+    expect(generateTextMock).not.toHaveBeenCalled()
+  })
+
+  it('allows providers to skip chat probing when they do not expose model listing', async () => {
+    listModelsMock.mockResolvedValue([])
+
+    const [connectivityValidator, chatValidator] = getProviderValidators({
+      checks: ['connectivity', 'chat_completions'],
+      allowValidationWithoutModel: true,
+    })
+
+    const connectivityResult = await connectivityValidator.validator(config, provider as any, undefined as any, undefined as any)
+    const chatResult = await chatValidator.validator(config, provider as any, undefined as any, undefined as any)
+
+    expect(connectivityResult.valid).toBe(true)
+    expect(chatResult.valid).toBe(true)
+    expect(generateTextMock).not.toHaveBeenCalled()
+  })
+})
@@ -15,6 +15,7 @@ type OpenAICompatibleValidationCheck = 'connectivity' | 'model_list' | 'chat_com
 interface OpenAICompatibleValidationOptions<TConfig extends { apiKey?: string, baseUrl?: string }> {
  checks?: OpenAICompatibleValidationCheck[]
  additionalHeaders?: Record<string, string>
+  allowValidationWithoutModel?: boolean
  schedule?: {
    mode: 'once' | 'interval'
    intervalMs?: number
@@ -88,16 +89,14 @@ async function pickValidationModel<TConfig extends { apiKey?: string | null, bas
  config: TConfig,
  provider: ProviderInstance,
  providerExtra: ProviderExtraMethods<TConfig> | undefined,
-): Promise<string> {
-  const fallback = 'test'
-
+): Promise<string | null> {
  try {
    const models = await resolveModels(config, provider, providerExtra)
    const modelId = extractModelId(models.find(model => !shouldSkipModelId(extractModelId(model))))
-    return modelId || fallback
+    return modelId || null
  }
  catch {
-    return fallback
+    return null
  }
 }

@@ -106,6 +105,7 @@ export function createOpenAICompatibleValidators<TConfig extends { apiKey?: stri
 ): ProviderDefinition<TConfig>['validators'] {
  const checks = options?.checks ?? ['connectivity', 'model_list', 'chat_completions']
  const additionalHeaders = options?.additionalHeaders
+  const missingValidationModelReason = 'No model available for validation. Configure a model manually and try again.'

  interface ChatCheckResult {
    connectivityOk: boolean
@@ -114,6 +114,48 @@ export function createOpenAICompatibleValidators<TConfig extends { apiKey?: stri
    error?: unknown
  }

+  async function runChatCheck(
+    config: TConfig,
+    provider: ProviderInstance,
+    providerExtra: ProviderExtraMethods<TConfig> | undefined,
+  ): Promise<ChatCheckResult> {
+    const model = await pickValidationModel(config, provider, providerExtra)
+
+    if (!model) {
+      if (options?.allowValidationWithoutModel) {
+        return { connectivityOk: true, chatOk: true }
+      }
+
+      return {
+        connectivityOk: false,
+        chatOk: false,
+        errorMessage: missingValidationModelReason,
+      }
+    }
+
+    try {
+      await generateText({
+        apiKey: config.apiKey,
+        baseURL: config.baseUrl!,
+        headers: additionalHeaders,
+        model,
+        messages: message.messages(message.user('ping')),
+        max_tokens: 1,
+      })
+
+      return { connectivityOk: true, chatOk: true }
+    }
+    catch (e) {
+      if (isNetworkError(e)) {
+        return { connectivityOk: false, chatOk: false, error: e, errorMessage: errorMessageFrom(e) }
+      }
+
+      const status = extractStatusCode(e)
+      const chatOk = status === 400 || Boolean(status && status >= 200 && status < 300)
+      return { connectivityOk: true, chatOk, errorMessage: errorMessageFrom(e) }
+    }
+  }
+
  const chatCheckCacheKey = 'openai-compatible:chat-check'
  const chatCheckMutexKey = 'openai-compatible:chat-check:mutex'
  const getChatCheckResult = async (
@@ -128,28 +170,7 @@ export function createOpenAICompatibleValidators<TConfig extends { apiKey?: stri
      return existing

    if (!cache) {
-      const model = await pickValidationModel(config, provider, providerExtra)
-      try {
-        await generateText({
-          apiKey: config.apiKey,
-          baseURL: config.baseUrl!,
-          headers: additionalHeaders,
-          model,
-          messages: message.messages(message.user('ping')),
-          max_tokens: 1,
-        })
-
-        return { connectivityOk: true, chatOk: true }
-      }
-      catch (e) {
-        if (isNetworkError(e)) {
-          return { connectivityOk: false, chatOk: false, error: e, errorMessage: errorMessageFrom(e) }
-        }
-
-        const status = extractStatusCode(e)
-        const chatOk = status === 400 || Boolean(status && status >= 200 && status < 300)
-        return { connectivityOk: true, chatOk, errorMessage: errorMessageFrom(e) }
-      }
+      return runChatCheck(config, provider, providerExtra)
    }

    let mutex = cache.get(chatCheckMutexKey) as Mutex | undefined
@@ -165,30 +186,7 @@ export function createOpenAICompatibleValidators<TConfig extends { apiKey?: stri
      if (cached)
        return cached

-      const sharedCheck = (async () => {
-        const model = await pickValidationModel(config, provider, providerExtra)
-        try {
-          await generateText({
-            apiKey: config.apiKey,
-            baseURL: config.baseUrl!,
-            headers: additionalHeaders,
-            model,
-            messages: message.messages(message.user('ping')),
-            max_tokens: 1,
-          })
-
-          return { connectivityOk: true, chatOk: true }
-        }
-        catch (e) {
-          if (isNetworkError(e)) {
-            return { connectivityOk: false, chatOk: false, error: e, errorMessage: errorMessageFrom(e) }
-          }
-
-          const status = extractStatusCode(e)
-          const chatOk = status === 400 || Boolean(status && status >= 200 && status < 300)
-          return { connectivityOk: true, chatOk, errorMessage: errorMessageFrom(e) }
-        }
-      })()
+      const sharedCheck = runChatCheck(config, provider, providerExtra)

      cache.set(chatCheckCacheKey, sharedCheck)