✨ feat: add Nvidia NIM provider support

Dockerfile

@@ -191,6 +191,8 @@ ENV \
     MOONSHOT_API_KEY="" MOONSHOT_MODEL_LIST="" MOONSHOT_PROXY_URL="" \
     # Novita
     NOVITA_API_KEY="" NOVITA_MODEL_LIST="" \
+    # Nvidia NIM
+    NVIDIA_API_KEY="" NVIDIA_MODEL_LIST="" NVIDIA_PROXY_URL="" \
     # Ollama
     ENABLED_OLLAMA="" OLLAMA_MODEL_LIST="" OLLAMA_PROXY_URL="" \
     # OpenAI

Dockerfile.database

@@ -228,6 +228,8 @@ ENV \
     MOONSHOT_API_KEY="" MOONSHOT_MODEL_LIST="" MOONSHOT_PROXY_URL="" \
     # Novita
     NOVITA_API_KEY="" NOVITA_MODEL_LIST="" \
+    # Nvidia NIM
+    NVIDIA_API_KEY="" NVIDIA_MODEL_LIST="" NVIDIA_PROXY_URL="" \
     # Ollama
     ENABLED_OLLAMA="" OLLAMA_MODEL_LIST="" OLLAMA_PROXY_URL="" \
     # OpenAI

src/app/[variants]/(main)/settings/llm/ProviderList/providers.tsx

@@ -17,6 +17,7 @@ import {
   MistralProviderCard,
   MoonshotProviderCard,
   NovitaProviderCard,
+  NvidiaProviderCard,
   OpenRouterProviderCard,
   PerplexityProviderCard,
   QwenProviderCard,
@@ -68,6 +69,7 @@ export const useProviderList = (): ProviderItem[] => {
       TogetherAIProviderCard,
       FireworksAIProviderCard,
       GroqProviderCard,
+      NvidiaProviderCard,
       PerplexityProviderCard,
       MistralProviderCard,
       Ai21ProviderCard,

src/config/aiModels/index.ts

@@ -23,6 +23,7 @@ import { default as minimax } from './minimax';
 import { default as mistral } from './mistral';
 import { default as moonshot } from './moonshot';
 import { default as novita } from './novita';
+import { default as nvidia } from './nvidia';
 import { default as ollama } from './ollama';
 import { default as openai } from './openai';
 import { default as openrouter } from './openrouter';
@@ -84,6 +85,7 @@ export const LOBE_DEFAULT_MODEL_LIST = buildDefaultModelList({
   mistral,
   moonshot,
   novita,
+  nvidia,
   ollama,
   openai,
   openrouter,
@@ -126,6 +128,7 @@ export { default as minimax } from './minimax';
 export { default as mistral } from './mistral';
 export { default as moonshot } from './moonshot';
 export { default as novita } from './novita';
+export { default as nvidia } from './nvidia';
 export { default as ollama } from './ollama';
 export { default as openai } from './openai';
 export { default as openrouter } from './openrouter';

src/config/aiModels/nvidia.ts

@@ -0,0 +1,155 @@
+import { AIChatModelCard } from '@/types/aiModel';
+
+const nvidiaChatModels: AIChatModelCard[] = [
+  {
+    abilities: {
+      functionCall: true,
+    },
+    contextWindowTokens: 128_000,
+    description: '先进的 LLM，擅长推理、数学、常识和函数调用。',
+    displayName: 'Llama 3.3 70B Instruct',
+    enabled: true,
+    id: 'meta/llama-3.3-70b-instruct',
+    type: 'chat'
+  },
+  {
+    contextWindowTokens: 128_000,
+    description: '先进的最尖端小型语言模型，具备语言理解、卓越的推理能力和文本生成能力。',
+    displayName: 'Llama 3.2 1B Instruct',
+    id: 'meta/llama-3.2-1b-instruct',
+    type: 'chat'
+  },
+  {
+    contextWindowTokens: 128_000,
+    description: '先进的最尖端小型语言模型，具备语言理解、卓越的推理能力和文本生成能力。',
+    displayName: 'Llama 3.2 3B Instruct',
+    id: 'meta/llama-3.2-3b-instruct',
+    type: 'chat'
+  },
+  {
+    abilities: {
+      vision: true,
+    },
+    contextWindowTokens: 128_000,
+    description: '尖端的视觉-语言模型，擅长从图像中进行高质量推理。',
+    displayName: 'Llama 3.2 11B Vision Instruct',
+    enabled: true,
+    id: 'meta/llama-3.2-11b-vision-instruct',
+    type: 'chat'
+  },
+  {
+    abilities: {
+      vision: true,
+    },
+    contextWindowTokens: 128_000,
+    description: '尖端的视觉-语言模型，擅长从图像中进行高质量推理。',
+    displayName: 'Llama 3.2 90B Vision Instruct',
+    id: 'meta/llama-3.2-90b-vision-instruct',
+    type: 'chat'
+  },
+  {
+    abilities: {
+      functionCall: true,
+    },
+    contextWindowTokens: 128_000,
+    description: '先进的最尖端模型，具备语言理解、卓越的推理能力和文本生成能力。',
+    displayName: 'Llama 3.1 8B Instruct',
+    id: 'meta/llama-3.1-8b-instruct',
+    type: 'chat'
+  },
+  {
+    abilities: {
+      functionCall: true,
+    },
+    contextWindowTokens: 128_000,
+    description: '赋能复杂对话，具备卓越的上下文理解、推理能力和文本生成能力。',
+    displayName: 'Llama 3.1 70B Instruct',
+    id: 'meta/llama-3.1-70b-instruct',
+    type: 'chat'
+  },
+  {
+    abilities: {
+      functionCall: true,
+    },
+    contextWindowTokens: 128_000,
+    description: '高级 LLM，支持合成数据生成、知识蒸馏和推理，适用于聊天机器人、编程和特定领域任务。',
+    displayName: 'Llama 3.1 405B Instruct',
+    id: 'meta/llama-3.1-405b-instruct',
+    type: 'chat'
+  },
+  {
+    contextWindowTokens: 32_768,
+    description: '独特的语言模型，提供无与伦比的准确性和效率表现。',
+    displayName: 'Llama 3.1 Nemotron 51B Instruct',
+    id: 'nvidia/llama-3.1-nemotron-51b-instruct',
+    type: 'chat'
+  },
+  {
+    contextWindowTokens: 32_768,
+    description: 'Llama-3.1-Nemotron-70B-Instruct 是 NVIDIA 定制的大型语言模型，旨在提高 LLM 生成的响应的帮助性。',
+    displayName: 'Llama 3.1 Nemotron 70B Instruct',
+    id: 'nvidia/llama-3.1-nemotron-70b-instruct',
+    type: 'chat'
+  },
+  {
+    contextWindowTokens: 8192,
+    description: '面向边缘应用的高级小型语言生成 AI 模型。',
+    displayName: 'Gemma 2 2B Instruct',
+    id: 'google/gemma-2-2b-it',
+    type: 'chat'
+  },
+  {
+    contextWindowTokens: 8192,
+    description: '尖端文本生成模型，擅长文本理解、转换和代码生成。',
+    displayName: 'Gemma 2 9B Instruct',
+    id: 'google/gemma-2-9b-it',
+    type: 'chat'
+  },
+  {
+    contextWindowTokens: 8192,
+    description: '尖端文本生成模型，擅长文本理解、转换和代码生成。',
+    displayName: 'Gemma 2 27B Instruct',
+    id: 'google/gemma-2-27b-it',
+    type: 'chat'
+  },
+  {
+    abilities: {
+      reasoning: true,
+    },
+    contextWindowTokens: 128_000,
+    description: '最先进的高效 LLM，擅长推理、数学和编程。',
+    displayName: 'DeepSeek R1',
+    enabled: true,
+    id: 'deepseek-ai/deepseek-r1',
+    type: 'chat'
+  },
+  {
+    abilities: {
+      functionCall: true,
+    },
+    contextWindowTokens: 32_768,
+    description: '面向中文和英文的 LLM，针对语言、编程、数学、推理等领域。',
+    displayName: 'Qwen2.5 7B Instruct',
+    enabled: true,
+    id: 'qwen/qwen2.5-7b-instruct',
+    type: 'chat'
+  },
+  {
+    contextWindowTokens: 32_768,
+    description: '强大的中型代码模型，支持 32K 上下文长度，擅长多语言编程。',
+    displayName: 'Qwen2.5 Coder 7B Instruct',
+    id: 'qwen/qwen2.5-coder-7b-instruct',
+    type: 'chat'
+  },
+  {
+    contextWindowTokens: 32_768,
+    description: '高级 LLM，支持代码生成、推理和修复，涵盖主流编程语言。',
+    displayName: 'Qwen2.5 Coder 32B Instruct',
+    id: 'qwen/qwen2.5-coder-32b-instruct',
+    type: 'chat'
+  },
+]
+
+export const allModels = [...nvidiaChatModels];
+
+export default allModels;

src/config/llm.ts

@@ -77,6 +77,9 @@ export const getLLMConfig = () => {
       ENABLED_NOVITA: z.boolean(),
       NOVITA_API_KEY: z.string().optional(),
 
+      ENABLED_NVIDIA: z.boolean(),
+      NVIDIA_API_KEY: z.string().optional(),
+
       ENABLED_BAICHUAN: z.boolean(),
       BAICHUAN_API_KEY: z.string().optional(),
 
@@ -202,6 +205,9 @@ export const getLLMConfig = () => {
       ENABLED_NOVITA: !!process.env.NOVITA_API_KEY,
       NOVITA_API_KEY: process.env.NOVITA_API_KEY,
 
+      ENABLED_NVIDIA: !!process.env.NVIDIA_API_KEY,
+      NVIDIA_API_KEY: process.env.NVIDIA_API_KEY,
+
       ENABLED_BAICHUAN: !!process.env.BAICHUAN_API_KEY,
       BAICHUAN_API_KEY: process.env.BAICHUAN_API_KEY,
 

src/config/modelProviders/index.ts

@@ -23,6 +23,7 @@ import MinimaxProvider from './minimax';
 import MistralProvider from './mistral';
 import MoonshotProvider from './moonshot';
 import NovitaProvider from './novita';
+import NvidiaProvider from './nvidia';
 import OllamaProvider from './ollama';
 import OpenAIProvider from './openai';
 import OpenRouterProvider from './openrouter';
@@ -67,6 +68,7 @@ export const LOBE_DEFAULT_MODEL_LIST: ChatModelCard[] = [
   ZeroOneProvider.chatModels,
   StepfunProvider.chatModels,
   NovitaProvider.chatModels,
+  NvidiaProvider.chatModels,
   BaichuanProvider.chatModels,
   TaichuProvider.chatModels,
   CloudflareProvider.chatModels,
@@ -96,6 +98,7 @@ export const DEFAULT_MODEL_PROVIDER_LIST = [
   CloudflareProvider,
   GithubProvider,
   NovitaProvider,
+  NvidiaProvider,
   TogetherAIProvider,
   FireworksAIProvider,
   GroqProvider,
@@ -158,6 +161,7 @@ export { default as MinimaxProviderCard } from './minimax';
 export { default as MistralProviderCard } from './mistral';
 export { default as MoonshotProviderCard } from './moonshot';
 export { default as NovitaProviderCard } from './novita';
+export { default as NvidiaProviderCard } from './nvidia';
 export { default as OllamaProviderCard } from './ollama';
 export { default as OpenAIProviderCard } from './openai';
 export { default as OpenRouterProviderCard } from './openrouter';

src/config/modelProviders/nvidia.ts

@@ -0,0 +1,21 @@
+import { ModelProviderCard } from '@/types/llm';
+
+const Nvidia: ModelProviderCard = {
+  chatModels: [],
+  checkModel: 'meta/llama-3.2-1b-instruct',
+  description: 'NVIDIA NIM™ 提供容器，可用于自托管 GPU 加速推理微服务，支持在云端、数据中心、RTX™ AI 个人电脑和工作站上部署预训练和自定义 AI 模型。',
+  id: 'nvidia',
+  modelList: { showModelFetcher: true },
+  modelsUrl: 'https://build.nvidia.com/models',
+  name: 'Nvidia',
+  settings: {
+    proxyUrl: {
+      placeholder: 'https://integrate.api.nvidia.com/v1',
+    },
+    sdkType: 'openai',
+    showModelFetcher: true,
+  },
+  url: 'https://build.nvidia.com',
+};
+
+export default Nvidia; 

src/libs/agent-runtime/AgentRuntime.ts

@@ -26,6 +26,7 @@ import { LobeMinimaxAI } from './minimax';
 import { LobeMistralAI } from './mistral';
 import { LobeMoonshotAI } from './moonshot';
 import { LobeNovitaAI } from './novita';
+import { LobeNvidiaAI } from './nvidia';
 import { LobeOllamaAI } from './ollama';
 import { LobeOpenAI } from './openai';
 import { LobeOpenRouterAI } from './openrouter';
@@ -157,6 +158,7 @@ class AgentRuntime {
       mistral: Partial<ClientOptions>;
       moonshot: Partial<ClientOptions>;
       novita: Partial<ClientOptions>;
+      nvidia: Partial<ClientOptions>;
       ollama: Partial<ClientOptions>;
       openai: Partial<ClientOptions>;
       openrouter: Partial<ClientOptions>;
@@ -300,6 +302,11 @@ class AgentRuntime {
         break;
       }
 
+      case ModelProvider.Nvidia: {
+        runtimeModel = new LobeNvidiaAI(params.nvidia);
+        break;
+      }
+
       case ModelProvider.Baichuan: {
         runtimeModel = new LobeBaichuanAI(params.baichuan ?? {});
         break;

src/libs/agent-runtime/nvidia/index.ts

@@ -0,0 +1,44 @@
+import { ModelProvider } from '../types';
+import { LobeOpenAICompatibleFactory } from '../utils/openaiCompatibleFactory';
+
+import type { ChatModelCard } from '@/types/llm';
+
+export interface NvidiaModelCard {
+  id: string;
+}
+
+export const LobeNvidiaAI = LobeOpenAICompatibleFactory({
+  baseURL: 'https://integrate.api.nvidia.com/v1',
+  debug: {
+    chatCompletion: () => process.env.DEBUG_NVIDIA_CHAT_COMPLETION === '1',
+  },
+  models: async ({ client }) => {
+    const { LOBE_DEFAULT_MODEL_LIST } = await import('@/config/aiModels');
+
+    const modelsPage = await client.models.list() as any;
+    const modelList: NvidiaModelCard[] = modelsPage.data;
+
+    return modelList
+      .map((model) => {
+        const knownModel = LOBE_DEFAULT_MODEL_LIST.find((m) => model.id.toLowerCase() === m.id.toLowerCase());
+
+        return {
+          contextWindowTokens: knownModel?.contextWindowTokens ?? undefined,
+          displayName: knownModel?.displayName ?? undefined,
+          enabled: knownModel?.enabled || false,
+          functionCall:
+            knownModel?.abilities?.functionCall
+            || false,
+          id: model.id,
+          reasoning:
+            knownModel?.abilities?.reasoning
+            || false,
+          vision:
+            knownModel?.abilities?.vision
+            || false,
+        };
+      })
+      .filter(Boolean) as ChatModelCard[];
+  },
+  provider: ModelProvider.Nvidia,
+});

src/libs/agent-runtime/types/type.ts

@@ -45,6 +45,7 @@ export enum ModelProvider {
   Mistral = 'mistral',
   Moonshot = 'moonshot',
   Novita = 'novita',
+  Nvidia = 'nvidia',
   Ollama = 'ollama',
   OpenAI = 'openai',
   OpenRouter = 'openrouter',

src/types/user/settings/keyVaults.ts

@@ -50,6 +50,7 @@ export interface UserKeyVaults {
   mistral?: OpenAICompatibleKeyVault;
   moonshot?: OpenAICompatibleKeyVault;
   novita?: OpenAICompatibleKeyVault;
+  nvidia?: OpenAICompatibleKeyVault;
   ollama?: OpenAICompatibleKeyVault;
   openai?: OpenAICompatibleKeyVault;
   openrouter?: OpenAICompatibleKeyVault;