✨ feat: Add Fireworks AI Model Provider (lobehub#3392)

* ✨ feat: Add Fireworks AI Model Provider * 💄 style: add FireworksAI self-models * 🐛 fix: support function call * Revert "🐛 fix: support function call" (lobehub#48) This reverts commit c8422f4. * 💄 style: update model info * 🐛 fix: disable stream if contains tools * 👷 build: add ENV for Fireworks AI provider * 👷 build: update ENV * ✨ feat: support proxy url * 🐛 fix: fix typo * Update providers.tsx * ♻️ refactor: cleanup * 🔨 chore: remove proxyUrl for FireworksAI * 🐛 fix: tools calling issue in Fireworks AI * 🔨 chore: remove work around method for tool calling * 🔨 chore: rollback changes * 🐛 fix: fix CI error * 💄 style: update model list * 🔨 chore: fix rebase conflicts
bentwnghk · Sep 10, 2024 · fa0d84d · fa0d84d
1 parent fc85c20
commit fa0d84d
Show file tree

Hide file tree

Showing 16 changed files with 468 additions and 4 deletions.
diff --git a/Dockerfile b/Dockerfile
@@ -117,6 +117,8 @@ ENV \
     BAICHUAN_API_KEY="" \
     # DeepSeek
     DEEPSEEK_API_KEY="" \
+    # Fireworks AI
+    FIREWORKSAI_API_KEY="" FIREWORKSAI_MODEL_LIST="" \
     # Google
     GOOGLE_API_KEY="" GOOGLE_PROXY_URL="" \
     # Groq

diff --git a/Dockerfile.database b/Dockerfile.database
@@ -149,6 +149,8 @@ ENV \
     BAICHUAN_API_KEY="" \
     # DeepSeek
     DEEPSEEK_API_KEY="" \
+    # Fireworks AI
+    FIREWORKSAI_API_KEY="" FIREWORKSAI_MODEL_LIST="" \
     # Google
     GOOGLE_API_KEY="" GOOGLE_PROXY_URL="" \
     # Groq

diff --git a/src/app/(main)/settings/llm/ProviderList/providers.tsx b/src/app/(main)/settings/llm/ProviderList/providers.tsx
@@ -5,6 +5,7 @@ import {
   AnthropicProviderCard,
   BaichuanProviderCard,
   DeepSeekProviderCard,
+  FireworksAIProviderCard,
   GoogleProviderCard,
   GroqProviderCard,
   MinimaxProviderCard,
@@ -48,6 +49,7 @@ export const useProviderList = (): ProviderItem[] => {
       OpenRouterProviderCard,
       NovitaProviderCard,
       TogetherAIProviderCard,
+      FireworksAIProviderCard,
       QwenProviderCard,
       DeepSeekProviderCard,
       MinimaxProviderCard,

diff --git a/src/app/api/chat/agentRuntime.ts b/src/app/api/chat/agentRuntime.ts
@@ -151,6 +151,13 @@ const getLlmOptionsFromPayload = (provider: string, payload: JWTPayload) => {
 
       return { apiKey };
     }
+    case ModelProvider.FireworksAI: {
+      const { FIREWORKSAI_API_KEY } = getLLMConfig();
+
+      const apiKey = apiKeyManager.pick(payload?.apiKey || FIREWORKSAI_API_KEY);
+
+      return { apiKey };
+    }
     case ModelProvider.ZeroOne: {
       const { ZEROONE_API_KEY } = getLLMConfig();
 

diff --git a/src/config/llm.ts b/src/config/llm.ts
@@ -64,6 +64,10 @@ export const getLLMConfig = () => {
       TOGETHERAI_API_KEY: z.string().optional(),
       TOGETHERAI_MODEL_LIST: z.string().optional(),
 
+      ENABLED_FIREWORKSAI: z.boolean(),
+      FIREWORKSAI_API_KEY: z.string().optional(),
+      FIREWORKSAI_MODEL_LIST: z.string().optional(),
+
       ENABLED_AWS_BEDROCK: z.boolean(),
       AWS_BEDROCK_MODEL_LIST: z.string().optional(),
       AWS_REGION: z.string().optional(),
@@ -152,6 +156,10 @@ export const getLLMConfig = () => {
       TOGETHERAI_API_KEY: process.env.TOGETHERAI_API_KEY,
       TOGETHERAI_MODEL_LIST: process.env.TOGETHERAI_MODEL_LIST,
 
+      ENABLED_FIREWORKSAI: !!process.env.FIREWORKSAI_API_KEY,
+      FIREWORKSAI_API_KEY: process.env.FIREWORKSAI_API_KEY,
+      FIREWORKSAI_MODEL_LIST: process.env.FIREWORKSAI_MODEL_LIST,
+
       ENABLED_MOONSHOT: !!process.env.MOONSHOT_API_KEY,
       MOONSHOT_API_KEY: process.env.MOONSHOT_API_KEY,
       MOONSHOT_PROXY_URL: process.env.MOONSHOT_PROXY_URL,

diff --git a/src/config/modelProviders/fireworksai.ts b/src/config/modelProviders/fireworksai.ts
@@ -0,0 +1,143 @@
+import { ModelProviderCard } from '@/types/llm';
+
+// ref https://fireworks.ai/models?show=Serverless
+// ref https://fireworks.ai/pricing
+const FireworksAI: ModelProviderCard = {
+  chatModels: [
+    {
+      description: 'Fireworks latest and most performant function-calling model. Firefunction-v2 is based on Llama-3 and trained to excel at function-calling as well as chat and instruction-following. See blog post for more details https://fireworks.ai/blog/firefunction-v2-launch-post',
+      displayName: 'Firefunction V2',
+      enabled: true,
+      functionCall: true,
+      id: 'accounts/fireworks/models/firefunction-v2',
+      tokens: 8192,
+    },
+    {
+      description: 'Fireworks open-source function calling model.',
+      displayName: 'Firefunction V1',
+      functionCall: true,
+      id: 'accounts/fireworks/models/firefunction-v1',
+      tokens: 32_768,
+    },
+    {
+      description: 'Vision-language model allowing both image and text as inputs (single image is recommended), trained on OSS model generated training data and open sourced on huggingface at fireworks-ai/FireLLaVA-13b',
+      displayName: 'FireLLaVA-13B',
+      enabled: true,
+      functionCall: false,
+      id: 'accounts/fireworks/models/firellava-13b',
+      tokens: 4096,
+      vision: true,
+    },
+    {
+      displayName: 'Llama 3.1 8B Instruct',
+      enabled: true,
+      functionCall: false,
+      id: 'accounts/fireworks/models/llama-v3p1-8b-instruct',
+      tokens: 131_072,
+    },
+    {
+      displayName: 'Llama 3.1 70B Instruct',
+      enabled: true,
+      functionCall: false,
+      id: 'accounts/fireworks/models/llama-v3p1-70b-instruct',
+      tokens: 131_072,
+    },
+    {
+      displayName: 'Llama 3.1 405B Instruct',
+      enabled: true,
+      functionCall: false,
+      id: 'accounts/fireworks/models/llama-v3p1-405b-instruct',
+      tokens: 131_072,
+    },
+    {
+      displayName: 'Llama 3 8B Instruct',
+      functionCall: false,
+      id: 'accounts/fireworks/models/llama-v3-8b-instruct',
+      tokens: 8192,
+    },
+    {
+      displayName: 'Llama 3 70B Instruct',
+      functionCall: false,
+      id: 'accounts/fireworks/models/llama-v3-70b-instruct',
+      tokens: 8192,
+    },
+    {
+      displayName: 'Llama 3 8B Instruct (HF version)',
+      functionCall: false,
+      id: 'accounts/fireworks/models/llama-v3-8b-instruct-hf',
+      tokens: 8192,
+    },
+    {
+      displayName: 'Llama 3 70B Instruct (HF version)',
+      functionCall: false,
+      id: 'accounts/fireworks/models/llama-v3-70b-instruct-hf',
+      tokens: 8192,
+    },
+    {
+      displayName: 'Gemma 2 9B Instruct',
+      enabled: true,
+      functionCall: false,
+      id: 'accounts/fireworks/models/gemma2-9b-it',
+      tokens: 8192,
+    },
+    {
+      displayName: 'Mixtral MoE 8x7B Instruct',
+      enabled: true,
+      functionCall: false,
+      id: 'accounts/fireworks/models/mixtral-8x7b-instruct',
+      tokens: 32_768,
+    },
+    {
+      displayName: 'Mixtral MoE 8x22B Instruct',
+      enabled: true,
+      functionCall: false,
+      id: 'accounts/fireworks/models/mixtral-8x22b-instruct',
+      tokens: 65_536,
+    },
+    {
+      displayName: 'Mixtral MoE 8x7B Instruct (HF version)',
+      functionCall: false,
+      id: 'accounts/fireworks/models/mixtral-8x7b-instruct-hf',
+      tokens: 32_768,
+    },
+    {
+      displayName: 'Phi 3 Vision Instruct',
+      enabled: true,
+      functionCall: false,
+      id: 'accounts/fireworks/models/phi-3-vision-128k-instruct',
+      tokens: 8192,
+      vision: true,
+    },
+    {
+      displayName: 'Yi-Large',
+      enabled: true,
+      functionCall: false,
+      id: 'accounts/yi-01-ai/models/yi-large',
+      tokens: 32_768,
+    },
+    {
+      displayName: 'StarCoder 7B',
+      functionCall: false,
+      id: 'accounts/fireworks/models/starcoder-7b',
+      tokens: 8192,
+    },
+    {
+      displayName: 'StarCoder 15.5B',
+      functionCall: false,
+      id: 'accounts/fireworks/models/starcoder-16b',
+      tokens: 8192,
+    },
+    {
+      displayName: 'MythoMax L2 13b',
+      functionCall: false,
+      id: 'accounts/fireworks/models/mythomax-l2-13b',
+      tokens: 4096,
+    },
+  ],
+  checkModel: 'accounts/fireworks/models/firefunction-v2',
+  id: 'fireworksai',
+  modelList: { showModelFetcher: true },
+  name: 'Fireworks AI',
+};
+
+export default FireworksAI;
diff --git a/src/config/modelProviders/index.ts b/src/config/modelProviders/index.ts
@@ -6,6 +6,7 @@ import AzureProvider from './azure';
 import BaichuanProvider from './baichuan';
 import BedrockProvider from './bedrock';
 import DeepSeekProvider from './deepseek';
+import FireworksAIProvider from './fireworksai';
 import GoogleProvider from './google';
 import GroqProvider from './groq';
 import MinimaxProvider from './minimax';
@@ -40,6 +41,7 @@ export const LOBE_DEFAULT_MODEL_LIST: ChatModelCard[] = [
   OllamaProvider.chatModels,
   OpenRouterProvider.chatModels,
   TogetherAIProvider.chatModels,
+  FireworksAIProvider.chatModels,
   PerplexityProvider.chatModels,
   AnthropicProvider.chatModels,
   ZeroOneProvider.chatModels,
@@ -63,6 +65,7 @@ export const DEFAULT_MODEL_PROVIDER_LIST = [
   GoogleProvider,
   OpenRouterProvider,
   TogetherAIProvider,
+  FireworksAIProvider,
   BedrockProvider,
   PerplexityProvider,
   MinimaxProvider,
@@ -96,6 +99,7 @@ export { default as AzureProviderCard } from './azure';
 export { default as BaichuanProviderCard } from './baichuan';
 export { default as BedrockProviderCard } from './bedrock';
 export { default as DeepSeekProviderCard } from './deepseek';
+export { default as FireworksAIProviderCard } from './fireworksai';
 export { default as GoogleProviderCard } from './google';
 export { default as GroqProviderCard } from './groq';
 export { default as MinimaxProviderCard } from './minimax';

diff --git a/src/const/settings/llm.ts b/src/const/settings/llm.ts
@@ -4,6 +4,7 @@ import {
   BaichuanProviderCard,
   BedrockProviderCard,
   DeepSeekProviderCard,
+  FireworksAIProviderCard,
   GoogleProviderCard,
   GroqProviderCard,
   MinimaxProviderCard,
@@ -52,6 +53,10 @@ export const DEFAULT_LLM_CONFIG: UserModelProviderConfig = {
     enabled: false,
     enabledModels: filterEnabledModels(DeepSeekProviderCard),
   },
+  fireworksai: {
+    enabled: false,
+    enabledModels: filterEnabledModels(FireworksAIProviderCard),
+  },
   google: {
     enabled: false,
     enabledModels: filterEnabledModels(GoogleProviderCard),

diff --git a/src/libs/agent-runtime/AgentRuntime.ts b/src/libs/agent-runtime/AgentRuntime.ts
@@ -9,6 +9,7 @@ import { LobeAzureOpenAI } from './azureOpenai';
 import { LobeBaichuanAI } from './baichuan';
 import { LobeBedrockAI, LobeBedrockAIParams } from './bedrock';
 import { LobeDeepSeekAI } from './deepseek';
+import { LobeFireworksAI } from './fireworksai';
 import { LobeGoogleAI } from './google';
 import { LobeGroq } from './groq';
 import { LobeMinimaxAI } from './minimax';
@@ -121,6 +122,7 @@ class AgentRuntime {
       baichuan: Partial<ClientOptions>;
       bedrock: Partial<LobeBedrockAIParams>;
       deepseek: Partial<ClientOptions>;
+      fireworksai: Partial<ClientOptions>;
       google: { apiKey?: string; baseURL?: string };
       groq: Partial<ClientOptions>;
       minimax: Partial<ClientOptions>;
@@ -226,6 +228,11 @@ class AgentRuntime {
         break;
       }
 
+      case ModelProvider.FireworksAI: {
+        runtimeModel = new LobeFireworksAI(params.fireworksai);
+        break
+      }
+
       case ModelProvider.ZeroOne: {
         runtimeModel = new LobeZeroOneAI(params.zeroone);
         break;