From 0c062124b6679bfd071e337f48f35cfa9ef02b80 Mon Sep 17 00:00:00 2001 From: Zhijie He Date: Tue, 4 Feb 2025 20:45:24 +0800 Subject: [PATCH 1/4] =?UTF-8?q?=E2=99=BB=EF=B8=8F=20refactor:=20refactor?= =?UTF-8?q?=20Wenxin=20with=20LobeOpenAICompatibleFactory?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit --- Dockerfile | 2 +- Dockerfile.database | 2 +- package.json | 1 - .../webapi/chat/wenxin/route.test.ts | 27 ---- src/app/(backend)/webapi/chat/wenxin/route.ts | 30 ---- .../llm/ProviderList/Wenxin/index.tsx | 44 ----- .../settings/llm/ProviderList/providers.tsx | 6 +- .../provider/(detail)/wenxin/page.tsx | 61 ------- src/config/aiModels/wenxin.ts | 24 +-- src/config/llm.ts | 8 +- src/config/modelProviders/wenxin.ts | 30 ++-- src/const/auth.ts | 3 - .../Conversation/Error/APIKeyForm/Wenxin.tsx | 49 ------ .../Conversation/Error/APIKeyForm/index.tsx | 3 - src/libs/agent-runtime/AgentRuntime.test.ts | 1 - src/libs/agent-runtime/AgentRuntime.ts | 7 + .../utils/streams/wenxin.test.ts | 153 ------------------ .../agent-runtime/utils/streams/wenxin.ts | 38 ----- src/libs/agent-runtime/wenxin/index.ts | 117 ++------------ src/libs/agent-runtime/wenxin/type.ts | 84 ---------- src/locales/default/modelProvider.ts | 20 --- src/server/modules/AgentRuntime/index.test.ts | 21 --- src/services/_auth.ts | 14 -- .../slices/modelList/selectors/keyVaults.ts | 2 - src/types/aiProvider.ts | 1 - src/types/user/settings/keyVaults.ts | 7 +- 26 files changed, 51 insertions(+), 704 deletions(-) delete mode 100644 src/app/(backend)/webapi/chat/wenxin/route.test.ts delete mode 100644 src/app/(backend)/webapi/chat/wenxin/route.ts delete mode 100644 src/app/(main)/settings/llm/ProviderList/Wenxin/index.tsx delete mode 100644 src/app/(main)/settings/provider/(detail)/wenxin/page.tsx delete mode 100644 src/features/Conversation/Error/APIKeyForm/Wenxin.tsx delete mode 100644 src/libs/agent-runtime/utils/streams/wenxin.test.ts delete mode 100644 src/libs/agent-runtime/utils/streams/wenxin.ts delete mode 100644 src/libs/agent-runtime/wenxin/type.ts diff --git a/Dockerfile b/Dockerfile index 56f30488d8c37..f4137cdc9b688 100644 --- a/Dockerfile +++ b/Dockerfile @@ -212,7 +212,7 @@ ENV \ # Upstage UPSTAGE_API_KEY="" UPSTAGE_MODEL_LIST="" \ # Wenxin - WENXIN_ACCESS_KEY="" WENXIN_SECRET_KEY="" WENXIN_MODEL_LIST="" \ + WENXIN_API_KEY="" WENXIN_MODEL_LIST="" \ # xAI XAI_API_KEY="" XAI_MODEL_LIST="" XAI_PROXY_URL="" \ # 01.AI diff --git a/Dockerfile.database b/Dockerfile.database index f3fd4b4aed99c..73acc79e61819 100644 --- a/Dockerfile.database +++ b/Dockerfile.database @@ -247,7 +247,7 @@ ENV \ # Upstage UPSTAGE_API_KEY="" UPSTAGE_MODEL_LIST="" \ # Wenxin - WENXIN_ACCESS_KEY="" WENXIN_SECRET_KEY="" WENXIN_MODEL_LIST="" \ + WENXIN_API_KEY="" WENXIN_MODEL_LIST="" \ # xAI XAI_API_KEY="" XAI_MODEL_LIST="" XAI_PROXY_URL="" \ # 01.AI diff --git a/package.json b/package.json index d5d2c60784481..3a1da04bffe37 100644 --- a/package.json +++ b/package.json @@ -111,7 +111,6 @@ "@aws-sdk/s3-request-presigner": "^3.723.0", "@azure/core-rest-pipeline": "1.16.0", "@azure/openai": "1.0.0-beta.12", - "@baiducloud/qianfan": "^0.1.9", "@cfworker/json-schema": "^4.1.0", "@clerk/localizations": "^3.9.6", "@clerk/nextjs": "^6.10.6", diff --git a/src/app/(backend)/webapi/chat/wenxin/route.test.ts b/src/app/(backend)/webapi/chat/wenxin/route.test.ts deleted file mode 100644 index cb25976b70fd4..0000000000000 --- a/src/app/(backend)/webapi/chat/wenxin/route.test.ts +++ /dev/null @@ -1,27 +0,0 @@ -// @vitest-environment edge-runtime -import { describe, expect, it, vi } from 'vitest'; - -import { POST as UniverseRoute } from '../[provider]/route'; -import { POST, runtime } from './route'; - -// 模拟 '../[provider]/route' -vi.mock('../[provider]/route', () => ({ - POST: vi.fn().mockResolvedValue('mocked response'), -})); - -describe('Configuration tests', () => { - it('should have runtime set to "edge"', () => { - expect(runtime).toBe('nodejs'); - }); -}); - -describe('Wenxin POST function tests', () => { - it('should call UniverseRoute with correct parameters', async () => { - const mockRequest = new Request('https://example.com', { method: 'POST' }); - await POST(mockRequest); - expect(UniverseRoute).toHaveBeenCalledWith(mockRequest, { - createRuntime: expect.anything(), - params: Promise.resolve({ provider: 'wenxin' }), - }); - }); -}); diff --git a/src/app/(backend)/webapi/chat/wenxin/route.ts b/src/app/(backend)/webapi/chat/wenxin/route.ts deleted file mode 100644 index b048f62e5ec4b..0000000000000 --- a/src/app/(backend)/webapi/chat/wenxin/route.ts +++ /dev/null @@ -1,30 +0,0 @@ -import { getLLMConfig } from '@/config/llm'; -import { AgentRuntime, ModelProvider } from '@/libs/agent-runtime'; -import LobeWenxinAI from '@/libs/agent-runtime/wenxin'; - -import { POST as UniverseRoute } from '../[provider]/route'; - -export const runtime = 'nodejs'; - -export const maxDuration = 30; - -export const POST = async (req: Request) => - UniverseRoute(req, { - createRuntime: (payload) => { - const { WENXIN_ACCESS_KEY, WENXIN_SECRET_KEY } = getLLMConfig(); - let accessKey: string | undefined = WENXIN_ACCESS_KEY; - let secretKey: string | undefined = WENXIN_SECRET_KEY; - - // if the payload has the api key, use user - if (payload.apiKey) { - accessKey = payload?.wenxinAccessKey; - secretKey = payload?.wenxinSecretKey; - } - - const params = { accessKey, secretKey }; - const instance = new LobeWenxinAI(params); - - return new AgentRuntime(instance); - }, - params: Promise.resolve({ provider: ModelProvider.Wenxin }), - }); diff --git a/src/app/(main)/settings/llm/ProviderList/Wenxin/index.tsx b/src/app/(main)/settings/llm/ProviderList/Wenxin/index.tsx deleted file mode 100644 index ffde4872d07cd..0000000000000 --- a/src/app/(main)/settings/llm/ProviderList/Wenxin/index.tsx +++ /dev/null @@ -1,44 +0,0 @@ -'use client'; - -import { Input } from 'antd'; -import { useTranslation } from 'react-i18next'; - -import { WenxinProviderCard } from '@/config/modelProviders'; -import { GlobalLLMProviderKey } from '@/types/user/settings'; - -import { KeyVaultsConfigKey } from '../../const'; -import { ProviderItem } from '../../type'; - -const providerKey: GlobalLLMProviderKey = 'wenxin'; - -export const useWenxinProvider = (): ProviderItem => { - const { t } = useTranslation('modelProvider'); - - return { - ...WenxinProviderCard, - apiKeyItems: [ - { - children: ( - - ), - desc: t(`${providerKey}.accessKey.desc`), - label: t(`${providerKey}.accessKey.title`), - name: [KeyVaultsConfigKey, providerKey, 'accessKey'], - }, - { - children: ( - - ), - desc: t(`${providerKey}.secretKey.desc`), - label: t(`${providerKey}.secretKey.title`), - name: [KeyVaultsConfigKey, providerKey, 'secretKey'], - }, - ], - }; -}; diff --git a/src/app/(main)/settings/llm/ProviderList/providers.tsx b/src/app/(main)/settings/llm/ProviderList/providers.tsx index f108683bfa07f..333864ec7fc37 100644 --- a/src/app/(main)/settings/llm/ProviderList/providers.tsx +++ b/src/app/(main)/settings/llm/ProviderList/providers.tsx @@ -27,6 +27,7 @@ import { TaichuProviderCard, TogetherAIProviderCard, UpstageProviderCard, + WenxinProviderCard, XAIProviderCard, ZeroOneProviderCard, ZhiPuProviderCard, @@ -40,7 +41,6 @@ import { useGithubProvider } from './Github'; import { useHuggingFaceProvider } from './HuggingFace'; import { useOllamaProvider } from './Ollama'; import { useOpenAIProvider } from './OpenAI'; -import { useWenxinProvider } from './Wenxin'; export const useProviderList = (): ProviderItem[] => { const AzureProvider = useAzureProvider(); @@ -50,7 +50,6 @@ export const useProviderList = (): ProviderItem[] => { const CloudflareProvider = useCloudflareProvider(); const GithubProvider = useGithubProvider(); const HuggingFaceProvider = useHuggingFaceProvider(); - const WenxinProvider = useWenxinProvider(); return useMemo( () => [ @@ -75,7 +74,7 @@ export const useProviderList = (): ProviderItem[] => { UpstageProviderCard, XAIProviderCard, QwenProviderCard, - WenxinProvider, + WenxinProviderCard, HunyuanProviderCard, SparkProviderCard, ZhiPuProviderCard, @@ -99,7 +98,6 @@ export const useProviderList = (): ProviderItem[] => { BedrockProvider, CloudflareProvider, GithubProvider, - WenxinProvider, HuggingFaceProvider, ], ); diff --git a/src/app/(main)/settings/provider/(detail)/wenxin/page.tsx b/src/app/(main)/settings/provider/(detail)/wenxin/page.tsx deleted file mode 100644 index 2d820e1a7df39..0000000000000 --- a/src/app/(main)/settings/provider/(detail)/wenxin/page.tsx +++ /dev/null @@ -1,61 +0,0 @@ -'use client'; - -import { useTranslation } from 'react-i18next'; - -import { FormPassword } from '@/components/FormInput'; -import { WenxinProviderCard } from '@/config/modelProviders'; -import { aiProviderSelectors, useAiInfraStore } from '@/store/aiInfra'; -import { GlobalLLMProviderKey } from '@/types/user/settings'; - -import { KeyVaultsConfigKey } from '../../const'; -import { SkeletonInput } from '../../features/ProviderConfig'; -import { ProviderItem } from '../../type'; -import ProviderDetail from '../[id]'; - -const providerKey: GlobalLLMProviderKey = 'wenxin'; - -const useProviderCard = (): ProviderItem => { - const { t } = useTranslation('modelProvider'); - - const isLoading = useAiInfraStore(aiProviderSelectors.isAiProviderConfigLoading(providerKey)); - - return { - ...WenxinProviderCard, - apiKeyItems: [ - { - children: isLoading ? ( - - ) : ( - - ), - desc: t(`${providerKey}.accessKey.desc`), - label: t(`${providerKey}.accessKey.title`), - name: [KeyVaultsConfigKey, 'accessKey'], - }, - { - children: isLoading ? ( - - ) : ( - - ), - desc: t(`${providerKey}.secretKey.desc`), - label: t(`${providerKey}.secretKey.title`), - name: [KeyVaultsConfigKey, 'secretKey'], - }, - ], - }; -}; - -const Page = () => { - const card = useProviderCard(); - - return ; -}; - -export default Page; diff --git a/src/config/aiModels/wenxin.ts b/src/config/aiModels/wenxin.ts index f773c6b6f37a8..47ba9d7a689ee 100644 --- a/src/config/aiModels/wenxin.ts +++ b/src/config/aiModels/wenxin.ts @@ -7,7 +7,7 @@ const wenxinChatModels: AIChatModelCard[] = [ '百度自研的旗舰级大规模⼤语⾔模型，覆盖海量中英文语料，具有强大的通用能力，可满足绝大部分对话问答、创作生成、插件应用场景要求；支持自动对接百度搜索插件，保障问答信息时效。', displayName: 'ERNIE 3.5 8K', enabled: true, - id: 'ERNIE-3.5-8K', + id: 'ernie-3.5-8k', pricing: { currency: 'CNY', input: 0.8, @@ -20,7 +20,7 @@ const wenxinChatModels: AIChatModelCard[] = [ description: '百度自研的旗舰级大规模⼤语⾔模型，覆盖海量中英文语料，具有强大的通用能力，可满足绝大部分对话问答、创作生成、插件应用场景要求；支持自动对接百度搜索插件，保障问答信息时效。', displayName: 'ERNIE 3.5 8K Preview', - id: 'ERNIE-3.5-8K-Preview', + id: 'ernie-3.5-8k-preview', pricing: { currency: 'CNY', input: 0.8, @@ -34,7 +34,7 @@ const wenxinChatModels: AIChatModelCard[] = [ '百度自研的旗舰级大规模⼤语⾔模型，覆盖海量中英文语料，具有强大的通用能力，可满足绝大部分对话问答、创作生成、插件应用场景要求；支持自动对接百度搜索插件，保障问答信息时效。', displayName: 'ERNIE 3.5 128K', enabled: true, - id: 'ERNIE-3.5-128K', + id: 'ernie-3.5-128k', pricing: { currency: 'CNY', input: 0.8, @@ -48,7 +48,7 @@ const wenxinChatModels: AIChatModelCard[] = [ '百度自研的旗舰级超大规模⼤语⾔模型，相较ERNIE 3.5实现了模型能力全面升级，广泛适用于各领域复杂任务场景；支持自动对接百度搜索插件，保障问答信息时效。', displayName: 'ERNIE 4.0 8K', enabled: true, - id: 'ERNIE-4.0-8K-Latest', + id: 'ernie-4.0-8k-latest', pricing: { currency: 'CNY', input: 30, @@ -61,7 +61,7 @@ const wenxinChatModels: AIChatModelCard[] = [ description: '百度自研的旗舰级超大规模⼤语⾔模型，相较ERNIE 3.5实现了模型能力全面升级，广泛适用于各领域复杂任务场景；支持自动对接百度搜索插件，保障问答信息时效。', displayName: 'ERNIE 4.0 8K Preview', - id: 'ERNIE-4.0-8K-Preview', + id: 'ernie-4.0-8k-preview', pricing: { currency: 'CNY', input: 30, @@ -75,7 +75,7 @@ const wenxinChatModels: AIChatModelCard[] = [ '百度自研的旗舰级超大规模⼤语⾔模型，综合效果表现出色，广泛适用于各领域复杂任务场景；支持自动对接百度搜索插件，保障问答信息时效。相较于ERNIE 4.0在性能表现上更优秀', displayName: 'ERNIE 4.0 Turbo 8K', enabled: true, - id: 'ERNIE-4.0-Turbo-8K-Latest', + id: 'ernie-4.0-turbo-8k-latest', pricing: { currency: 'CNY', input: 20, @@ -89,7 +89,7 @@ const wenxinChatModels: AIChatModelCard[] = [ '百度自研的旗舰级超大规模⼤语⾔模型，综合效果表现出色，广泛适用于各领域复杂任务场景；支持自动对接百度搜索插件，保障问答信息时效。相较于ERNIE 4.0在性能表现上更优秀', displayName: 'ERNIE 4.0 Turbo 128K', enabled: true, - id: 'ERNIE-4.0-Turbo-128K', + id: 'ernie-4.0-turbo-128k', pricing: { currency: 'CNY', input: 20, @@ -102,7 +102,7 @@ const wenxinChatModels: AIChatModelCard[] = [ description: '百度自研的旗舰级超大规模⼤语⾔模型，综合效果表现出色，广泛适用于各领域复杂任务场景；支持自动对接百度搜索插件，保障问答信息时效。相较于ERNIE 4.0在性能表现上更优秀', displayName: 'ERNIE 4.0 Turbo 8K Preview', - id: 'ERNIE-4.0-Turbo-8K-Preview', + id: 'ernie-4.0-turbo-8k-preview', pricing: { currency: 'CNY', input: 20, @@ -116,7 +116,7 @@ const wenxinChatModels: AIChatModelCard[] = [ '百度自研的轻量级大语言模型，兼顾优异的模型效果与推理性能，效果比ERNIE Lite更优，适合低算力AI加速卡推理使用。', displayName: 'ERNIE Lite Pro 128K', enabled: true, - id: 'ERNIE-Lite-Pro-128K', + id: 'ernie-lite-pro-128k', pricing: { currency: 'CNY', input: 0.2, @@ -130,7 +130,7 @@ const wenxinChatModels: AIChatModelCard[] = [ '百度2024年最新发布的自研高性能大语言模型，通用能力优异，效果比ERNIE Speed更优，适合作为基座模型进行精调，更好地处理特定场景问题，同时具备极佳的推理性能。', displayName: 'ERNIE Speed Pro 128K', enabled: true, - id: 'ERNIE-Speed-Pro-128K', + id: 'ernie-speed-pro-128k', pricing: { currency: 'CNY', input: 0.3, @@ -143,7 +143,7 @@ const wenxinChatModels: AIChatModelCard[] = [ description: '百度2024年最新发布的自研高性能大语言模型，通用能力优异，适合作为基座模型进行精调，更好地处理特定场景问题，同时具备极佳的推理性能。', displayName: 'ERNIE Speed 128K', - id: 'ERNIE-Speed-128K', + id: 'ernie-speed-128k', pricing: { currency: 'CNY', input: 0, @@ -156,7 +156,7 @@ const wenxinChatModels: AIChatModelCard[] = [ description: '百度自研的垂直场景大语言模型，适合游戏NPC、客服对话、对话角色扮演等应用场景，人设风格更为鲜明、一致，指令遵循能力更强，推理性能更优。', displayName: 'ERNIE Character 8K', - id: 'ERNIE-Character-8K', + id: 'ernie-character-8k', pricing: { currency: 'CNY', input: 4, diff --git a/src/config/llm.ts b/src/config/llm.ts index c5f15c3f0acf6..f6908756c7fc6 100644 --- a/src/config/llm.ts +++ b/src/config/llm.ts @@ -64,8 +64,7 @@ export const getLLMConfig = () => { AWS_SESSION_TOKEN: z.string().optional(), ENABLED_WENXIN: z.boolean(), - WENXIN_ACCESS_KEY: z.string().optional(), - WENXIN_SECRET_KEY: z.string().optional(), + WENXIN_API_KEY: z.string().optional(), ENABLED_OLLAMA: z.boolean(), @@ -186,9 +185,8 @@ export const getLLMConfig = () => { AWS_SECRET_ACCESS_KEY: process.env.AWS_SECRET_ACCESS_KEY, AWS_SESSION_TOKEN: process.env.AWS_SESSION_TOKEN, - ENABLED_WENXIN: !!process.env.WENXIN_ACCESS_KEY && !!process.env.WENXIN_SECRET_KEY, - WENXIN_ACCESS_KEY: process.env.WENXIN_ACCESS_KEY, - WENXIN_SECRET_KEY: process.env.WENXIN_SECRET_KEY, + ENABLED_WENXIN: !!process.env.WENXIN_API_KEY, + WENXIN_API_KEY: process.env.WENXIN_API_KEY, ENABLED_OLLAMA: process.env.ENABLED_OLLAMA !== '0', diff --git a/src/config/modelProviders/wenxin.ts b/src/config/modelProviders/wenxin.ts index 25a71e60b0497..13124a990dd3e 100644 --- a/src/config/modelProviders/wenxin.ts +++ b/src/config/modelProviders/wenxin.ts @@ -9,7 +9,7 @@ const BaiduWenxin: ModelProviderCard = { '百度自研的旗舰级大规模⼤语⾔模型，覆盖海量中英文语料，具有强大的通用能力，可满足绝大部分对话问答、创作生成、插件应用场景要求；支持自动对接百度搜索插件，保障问答信息时效。', displayName: 'ERNIE 3.5 8K', enabled: true, - id: 'ERNIE-3.5-8K', + id: 'ernie-3.5-8k', pricing: { currency: 'CNY', input: 0.8, @@ -21,7 +21,7 @@ const BaiduWenxin: ModelProviderCard = { description: '百度自研的旗舰级大规模⼤语⾔模型，覆盖海量中英文语料，具有强大的通用能力，可满足绝大部分对话问答、创作生成、插件应用场景要求；支持自动对接百度搜索插件，保障问答信息时效。', displayName: 'ERNIE 3.5 8K Preview', - id: 'ERNIE-3.5-8K-Preview', + id: 'ernie-3.5-8k-preview', pricing: { currency: 'CNY', input: 0.8, @@ -34,7 +34,7 @@ const BaiduWenxin: ModelProviderCard = { '百度自研的旗舰级大规模⼤语⾔模型，覆盖海量中英文语料，具有强大的通用能力，可满足绝大部分对话问答、创作生成、插件应用场景要求；支持自动对接百度搜索插件，保障问答信息时效。', displayName: 'ERNIE 3.5 128K', enabled: true, - id: 'ERNIE-3.5-128K', + id: 'ernie-3.5-128k', pricing: { currency: 'CNY', input: 0.8, @@ -47,7 +47,7 @@ const BaiduWenxin: ModelProviderCard = { '百度自研的旗舰级超大规模⼤语⾔模型，相较ERNIE 3.5实现了模型能力全面升级，广泛适用于各领域复杂任务场景；支持自动对接百度搜索插件，保障问答信息时效。', displayName: 'ERNIE 4.0 8K', enabled: true, - id: 'ERNIE-4.0-8K-Latest', + id: 'ernie-4.0-8k-latest', pricing: { currency: 'CNY', input: 30, @@ -59,7 +59,7 @@ const BaiduWenxin: ModelProviderCard = { description: '百度自研的旗舰级超大规模⼤语⾔模型，相较ERNIE 3.5实现了模型能力全面升级，广泛适用于各领域复杂任务场景；支持自动对接百度搜索插件，保障问答信息时效。', displayName: 'ERNIE 4.0 8K Preview', - id: 'ERNIE-4.0-8K-Preview', + id: 'ernie-4.0-8k-preview', pricing: { currency: 'CNY', input: 30, @@ -72,7 +72,7 @@ const BaiduWenxin: ModelProviderCard = { '百度自研的旗舰级超大规模⼤语⾔模型，综合效果表现出色，广泛适用于各领域复杂任务场景；支持自动对接百度搜索插件，保障问答信息时效。相较于ERNIE 4.0在性能表现上更优秀', displayName: 'ERNIE 4.0 Turbo 8K', enabled: true, - id: 'ERNIE-4.0-Turbo-8K-Latest', + id: 'ernie-4.0-turbo-8k-latest', pricing: { currency: 'CNY', input: 20, @@ -85,7 +85,7 @@ const BaiduWenxin: ModelProviderCard = { '百度自研的旗舰级超大规模⼤语⾔模型，综合效果表现出色，广泛适用于各领域复杂任务场景；支持自动对接百度搜索插件，保障问答信息时效。相较于ERNIE 4.0在性能表现上更优秀', displayName: 'ERNIE 4.0 Turbo 128K', enabled: true, - id: 'ERNIE-4.0-Turbo-128K', + id: 'ernie-4.0-turbo-128k', pricing: { currency: 'CNY', input: 20, @@ -97,7 +97,7 @@ const BaiduWenxin: ModelProviderCard = { description: '百度自研的旗舰级超大规模⼤语⾔模型，综合效果表现出色，广泛适用于各领域复杂任务场景；支持自动对接百度搜索插件，保障问答信息时效。相较于ERNIE 4.0在性能表现上更优秀', displayName: 'ERNIE 4.0 Turbo 8K Preview', - id: 'ERNIE-4.0-Turbo-8K-Preview', + id: 'ernie-4.0-turbo-8k-preview', pricing: { currency: 'CNY', input: 20, @@ -110,7 +110,7 @@ const BaiduWenxin: ModelProviderCard = { '百度自研的轻量级大语言模型，兼顾优异的模型效果与推理性能，效果比ERNIE Lite更优，适合低算力AI加速卡推理使用。', displayName: 'ERNIE Lite Pro 128K', enabled: true, - id: 'ERNIE-Lite-Pro-128K', + id: 'ernie-lite-pro-128k', pricing: { currency: 'CNY', input: 0.2, @@ -123,7 +123,7 @@ const BaiduWenxin: ModelProviderCard = { '百度2024年最新发布的自研高性能大语言模型，通用能力优异，效果比ERNIE Speed更优，适合作为基座模型进行精调，更好地处理特定场景问题，同时具备极佳的推理性能。', displayName: 'ERNIE Speed Pro 128K', enabled: true, - id: 'ERNIE-Speed-Pro-128K', + id: 'ernie-speed-pro-128k', pricing: { currency: 'CNY', input: 0.3, @@ -135,7 +135,7 @@ const BaiduWenxin: ModelProviderCard = { description: '百度2024年最新发布的自研高性能大语言模型，通用能力优异，适合作为基座模型进行精调，更好地处理特定场景问题，同时具备极佳的推理性能。', displayName: 'ERNIE Speed 128K', - id: 'ERNIE-Speed-128K', + id: 'ernie-speed-128k', pricing: { currency: 'CNY', input: 0, @@ -147,7 +147,7 @@ const BaiduWenxin: ModelProviderCard = { description: '百度自研的垂直场景大语言模型，适合游戏NPC、客服对话、对话角色扮演等应用场景，人设风格更为鲜明、一致，指令遵循能力更强，推理性能更优。', displayName: 'ERNIE Character 8K', - id: 'ERNIE-Character-8K', + id: 'ernie-character-8k', pricing: { currency: 'CNY', input: 4, @@ -155,16 +155,14 @@ const BaiduWenxin: ModelProviderCard = { }, }, ], - checkModel: 'ERNIE-Speed-128K', + checkModel: 'ernie-speed-128k', description: '企业级一站式大模型与AI原生应用开发及服务平台，提供最全面易用的生成式人工智能模型开发、应用开发全流程工具链', - disableBrowserRequest: true, id: 'wenxin', modelsUrl: 'https://cloud.baidu.com/doc/WENXINWORKSHOP/s/Nlks5zkzu#%E5%AF%B9%E8%AF%9Dchat', name: 'Wenxin', settings: { - disableBrowserRequest: true, - sdkType: 'wenxin', + sdkType: 'openai', smoothing: { speed: 2, text: true, diff --git a/src/const/auth.ts b/src/const/auth.ts index 5bec0201d5a2f..9a633da5459ec 100644 --- a/src/const/auth.ts +++ b/src/const/auth.ts @@ -39,9 +39,6 @@ export interface JWTPayload { cloudflareBaseURLOrAccountID?: string; - wenxinAccessKey?: string; - wenxinSecretKey?: string; - /** * user id * in client db mode it's a uuid diff --git a/src/features/Conversation/Error/APIKeyForm/Wenxin.tsx b/src/features/Conversation/Error/APIKeyForm/Wenxin.tsx deleted file mode 100644 index c109cf6afbd27..0000000000000 --- a/src/features/Conversation/Error/APIKeyForm/Wenxin.tsx +++ /dev/null @@ -1,49 +0,0 @@ -import { Wenxin } from '@lobehub/icons'; -import { Input } from 'antd'; -import { memo } from 'react'; -import { useTranslation } from 'react-i18next'; - -import { ModelProvider } from '@/libs/agent-runtime'; -import { useUserStore } from '@/store/user'; -import { keyVaultsConfigSelectors } from '@/store/user/selectors'; - -import { FormAction } from '../style'; - -const WenxinForm = memo(() => { - const { t } = useTranslation('modelProvider'); - - const [accessKey, secretKey, setConfig] = useUserStore((s) => [ - keyVaultsConfigSelectors.wenxinConfig(s).accessKey, - keyVaultsConfigSelectors.wenxinConfig(s).secretKey, - s.updateKeyVaultConfig, - ]); - - return ( - } - description={t('wenxin.unlock.description')} - title={t('wenxin.unlock.title')} - > - { - setConfig(ModelProvider.Wenxin, { accessKey: e.target.value }); - }} - placeholder={'Access Key'} - type={'block'} - value={accessKey} - /> - { - setConfig(ModelProvider.Wenxin, { secretKey: e.target.value }); - }} - placeholder={'Secret Key'} - type={'block'} - value={secretKey} - /> - - ); -}); - -export default WenxinForm; diff --git a/src/features/Conversation/Error/APIKeyForm/index.tsx b/src/features/Conversation/Error/APIKeyForm/index.tsx index 5ba78f4f0ba34..703b079948f35 100644 --- a/src/features/Conversation/Error/APIKeyForm/index.tsx +++ b/src/features/Conversation/Error/APIKeyForm/index.tsx @@ -10,7 +10,6 @@ import { GlobalLLMProviderKey } from '@/types/user/settings'; import BedrockForm from './Bedrock'; import ProviderApiKeyForm from './ProviderApiKeyForm'; -import WenxinForm from './Wenxin'; interface APIKeyFormProps { id: string; @@ -66,8 +65,6 @@ const APIKeyForm = memo(({ id, provider }) => {
{provider === ModelProvider.Bedrock ? ( - ) : provider === ModelProvider.Wenxin ? ( - ) : ( ; togetherai: Partial; upstage: Partial; + wenxin: Partial; xai: Partial; zeroone: Partial; zhipu: Partial; @@ -370,6 +372,11 @@ class AgentRuntime { runtimeModel = new LobeDoubaoAI(params.doubao); break; } + + case ModelProvider.Wenxin: { + runtimeModel = new LobeWenxinAI(params.wenxin); + break; + } } return new AgentRuntime(runtimeModel); } diff --git a/src/libs/agent-runtime/utils/streams/wenxin.test.ts b/src/libs/agent-runtime/utils/streams/wenxin.test.ts deleted file mode 100644 index da0072db84419..0000000000000 --- a/src/libs/agent-runtime/utils/streams/wenxin.test.ts +++ /dev/null @@ -1,153 +0,0 @@ -import { describe, expect, it, vi } from 'vitest'; - -import * as uuidModule from '@/utils/uuid'; - -import { convertIterableToStream } from '../../utils/streams/protocol'; -import { ChatResp } from '../../wenxin/type'; -import { WenxinStream } from './wenxin'; - -const dataStream = [ - { - id: 'as-vb0m37ti8y', - object: 'chat.completion', - created: 1709089502, - sentence_id: 0, - is_end: false, - is_truncated: false, - result: '当然可以，', - need_clear_history: false, - finish_reason: 'normal', - usage: { prompt_tokens: 5, completion_tokens: 2, total_tokens: 7 }, - }, - { - id: 'as-vb0m37ti8y', - object: 'chat.completion', - created: 1709089504, - sentence_id: 1, - is_end: false, - is_truncated: false, - result: - '以下是一些建议的自驾游路线，它们涵盖了各种不同的风景和文化体验：\n\n1. **西安-敦煌历史文化之旅**：\n\n\n\t* 路线：西安', - need_clear_history: false, - finish_reason: 'normal', - usage: { prompt_tokens: 5, completion_tokens: 2, total_tokens: 7 }, - }, - { - id: 'as-vb0m37ti8y', - object: 'chat.completion', - created: 1709089506, - sentence_id: 2, - is_end: false, - is_truncated: false, - result: ' - 天水 - 兰州 - 嘉峪关 - 敦煌\n\t* 特点：此路线让您领略到中国西北的丰富历史文化。', - need_clear_history: false, - finish_reason: 'normal', - usage: { prompt_tokens: 5, completion_tokens: 2, total_tokens: 7 }, - }, - { - id: 'as-vb0m37ti8y', - object: 'chat.completion', - created: 1709089508, - sentence_id: 3, - is_end: false, - is_truncated: false, - result: '您可以参观西安的兵马俑、大雁塔，体验兰州的黄河风情，以及在敦煌欣赏壮丽的莫高窟。', - need_clear_history: false, - finish_reason: 'normal', - usage: { prompt_tokens: 5, completion_tokens: 2, total_tokens: 7 }, - }, - { - id: 'as-vb0m37ti8y', - object: 'chat.completion', - created: 1709089511, - sentence_id: 4, - is_end: false, - is_truncated: false, - result: '\n2. **海南环岛热带风情游**：\n\n\n\t* 路线：海口 - 三亚 - 陵水 - 万宁 - 文昌 - 海', - need_clear_history: false, - finish_reason: 'normal', - usage: { prompt_tokens: 5, completion_tokens: 2, total_tokens: 7 }, - }, - { - id: 'as-vb0m37ti8y', - object: 'chat.completion', - created: 1709089512, - sentence_id: 5, - is_end: false, - is_truncated: false, - result: - '口\n\t* 特点：海南岛是中国唯一的黎族聚居区，这里有独特的热带风情、美丽的海滩和丰富的水果。', - need_clear_history: false, - finish_reason: 'normal', - usage: { prompt_tokens: 5, completion_tokens: 153, total_tokens: 158 }, - }, -]; - -describe('WenxinStream', () => { - it('should transform Wenxin stream to protocol stream', async () => { - vi.spyOn(uuidModule, 'nanoid').mockReturnValueOnce('1'); - - const mockWenxinStream: AsyncIterable = { - // @ts-ignore - async *[Symbol.asyncIterator]() { - for (const item of dataStream) { - yield item; - } - }, - }; - - const stream = convertIterableToStream(mockWenxinStream); - - const onStartMock = vi.fn(); - const onTextMock = vi.fn(); - const onTokenMock = vi.fn(); - const onCompletionMock = vi.fn(); - - const protocolStream = WenxinStream(stream, { - onStart: onStartMock, - onText: onTextMock, - onToken: onTokenMock, - onCompletion: onCompletionMock, - }); - - const decoder = new TextDecoder(); - const chunks = []; - - // @ts-ignore - for await (const chunk of protocolStream) { - chunks.push(decoder.decode(chunk, { stream: true })); - } - - expect(chunks).toEqual( - [ - 'id: as-vb0m37ti8y', - 'event: text', - `data: "当然可以，"\n`, - 'id: as-vb0m37ti8y', - 'event: text', - `data: "以下是一些建议的自驾游路线，它们涵盖了各种不同的风景和文化体验：\\n\\n1. **西安-敦煌历史文化之旅**：\\n\\n\\n\\t* 路线：西安"\n`, - 'id: as-vb0m37ti8y', - 'event: text', - `data: " - 天水 - 兰州 - 嘉峪关 - 敦煌\\n\\t* 特点：此路线让您领略到中国西北的丰富历史文化。"\n`, - 'id: as-vb0m37ti8y', - 'event: text', - `data: "您可以参观西安的兵马俑、大雁塔，体验兰州的黄河风情，以及在敦煌欣赏壮丽的莫高窟。"\n`, - 'id: as-vb0m37ti8y', - 'event: text', - `data: "\\n2. **海南环岛热带风情游**：\\n\\n\\n\\t* 路线：海口 - 三亚 - 陵水 - 万宁 - 文昌 - 海"\n`, - 'id: as-vb0m37ti8y', - 'event: text', - `data: "口\\n\\t* 特点：海南岛是中国唯一的黎族聚居区，这里有独特的热带风情、美丽的海滩和丰富的水果。"\n`, - ].map((item) => `${item}\n`), - ); - - expect(onStartMock).toHaveBeenCalledTimes(1); - expect(onTextMock).toHaveBeenNthCalledWith(1, '"当然可以，"'); - expect(onTextMock).toHaveBeenNthCalledWith( - 2, - '"以下是一些建议的自驾游路线，它们涵盖了各种不同的风景和文化体验：\\n\\n1. **西安-敦煌历史文化之旅**：\\n\\n\\n\\t* 路线：西安"', - ); - expect(onTokenMock).toHaveBeenCalledTimes(6); - expect(onCompletionMock).toHaveBeenCalledTimes(1); - }); -}); diff --git a/src/libs/agent-runtime/utils/streams/wenxin.ts b/src/libs/agent-runtime/utils/streams/wenxin.ts deleted file mode 100644 index 418ede9e7f265..0000000000000 --- a/src/libs/agent-runtime/utils/streams/wenxin.ts +++ /dev/null @@ -1,38 +0,0 @@ -import { ChatStreamCallbacks } from '@/libs/agent-runtime'; -import { nanoid } from '@/utils/uuid'; - -import { ChatResp } from '../../wenxin/type'; -import { - StreamProtocolChunk, - StreamStack, - createCallbacksTransformer, - createSSEProtocolTransformer, -} from './protocol'; - -const transformERNIEBotStream = (chunk: ChatResp): StreamProtocolChunk => { - const finished = chunk.is_end; - if (finished) { - return { data: chunk.finish_reason || 'stop', id: chunk.id, type: 'stop' }; - } - - if (chunk.result) { - return { data: chunk.result, id: chunk.id, type: 'text' }; - } - - return { - data: chunk, - id: chunk.id, - type: 'data', - }; -}; - -export const WenxinStream = ( - rawStream: ReadableStream, - callbacks?: ChatStreamCallbacks, -) => { - const streamStack: StreamStack = { id: 'chat_' + nanoid() }; - - return rawStream - .pipeThrough(createSSEProtocolTransformer(transformERNIEBotStream, streamStack)) - .pipeThrough(createCallbacksTransformer(callbacks)); -}; diff --git a/src/libs/agent-runtime/wenxin/index.ts b/src/libs/agent-runtime/wenxin/index.ts index f597fcf92d923..c324e9b9bc695 100644 --- a/src/libs/agent-runtime/wenxin/index.ts +++ b/src/libs/agent-runtime/wenxin/index.ts @@ -1,107 +1,10 @@ -import { ChatCompletion } from '@baiducloud/qianfan'; - -// 如果引入了这个类型，那么在跑 type-check 的 tsc 检查中就会抛错，大无语 -// import type QianFanClient from '@baiducloud/qianfan/src/ChatCompletion/index'; -import { safeParseJSON } from '@/utils/safeParseJSON'; - -import { LobeRuntimeAI } from '../BaseAI'; -import { AgentRuntimeErrorType } from '../error'; -import { ChatCompetitionOptions, ChatStreamPayload } from '../types'; -import { AgentRuntimeError } from '../utils/createError'; -import { debugStream } from '../utils/debugStream'; -import { StreamingResponse } from '../utils/response'; -import { convertIterableToStream } from '../utils/streams'; -import { WenxinStream } from '../utils/streams/wenxin'; -import { ChatResp } from './type'; - -interface ChatErrorCode { - error_code: number; - error_msg: string; -} - -export interface LobeWenxinAIParams { - accessKey?: string; - baseURL?: string; - secretKey?: string; -} - -export class LobeWenxinAI implements LobeRuntimeAI { - private client: any; - baseURL?: string; - - constructor({ accessKey, baseURL, secretKey }: LobeWenxinAIParams = {}) { - if (!accessKey || !secretKey) - throw AgentRuntimeError.createError(AgentRuntimeErrorType.InvalidProviderAPIKey); - - this.client = new ChatCompletion({ - QIANFAN_ACCESS_KEY: accessKey, - QIANFAN_SECRET_KEY: secretKey, - }); - this.baseURL = baseURL; - } - - async chat(payload: ChatStreamPayload, options?: ChatCompetitionOptions) { - try { - const result = await this.client.chat( - { messages: payload.messages as any, stream: true, user_id: options?.user }, - payload.model, - ); - - const wenxinStream = convertIterableToStream(result as AsyncIterable); - - const [prod, useForDebug] = wenxinStream.tee(); - - if (process.env.DEBUG_WENXIN_CHAT_COMPLETION === '1') { - debugStream(useForDebug).catch(); - } - - const stream = WenxinStream(prod, options?.callback); - - // Respond with the stream - return StreamingResponse(stream, { headers: options?.headers }); - } catch (e) { - const err = e as Error; - - const error: ChatErrorCode | undefined = safeParseJSON(err.message); - - if (!error) { - throw AgentRuntimeError.createError(AgentRuntimeErrorType.AgentRuntimeError, { - message: err.message, - name: err.name, - }); - } - - // 文心一言错误码 - // https://cloud.baidu.com/doc/WENXINWORKSHOP/s/tlmyncueh - switch (error.error_code) { - // Invalid API key or access key - case 100: - case 13: - case 14: { - throw AgentRuntimeError.createError(AgentRuntimeErrorType.InvalidProviderAPIKey, error); - } - - // quota limit - case 4: - case 17: - case 18: - case 19: - case 336_501: - case 336_502: - case 336_503: - case 336_504: - case 336_505: - case 336_507: { - throw AgentRuntimeError.createError(AgentRuntimeErrorType.QuotaLimitReached, { - errorCode: error.error_code, - message: `${error.error_msg} | you can visit https://cloud.baidu.com/doc/WENXINWORKSHOP/s/tlmyncueh for more information about the error code`, - }); - } - } - - throw AgentRuntimeError.createError(AgentRuntimeErrorType.ProviderBizError, error); - } - } -} - -export default LobeWenxinAI; +import { ModelProvider } from '../types'; +import { LobeOpenAICompatibleFactory } from '../utils/openaiCompatibleFactory'; + +export const LobeWenxinAI = LobeOpenAICompatibleFactory({ + baseURL: 'https://qianfan.baidubce.com/v2', + debug: { + chatCompletion: () => process.env.DEBUG_WENXIN_CHAT_COMPLETION === '1', + }, + provider: ModelProvider.Wenxin, +}); diff --git a/src/libs/agent-runtime/wenxin/type.ts b/src/libs/agent-runtime/wenxin/type.ts deleted file mode 100644 index 63bafbba7e33e..0000000000000 --- a/src/libs/agent-runtime/wenxin/type.ts +++ /dev/null @@ -1,84 +0,0 @@ -/** - * token 用量基类 - */ -export interface TokenUsage { - /** - * 回答tokens数 - */ - completion_tokens?: number; - /** - * 问题tokens数 - */ - prompt_tokens: number; - /** - * tokens总数 - */ - total_tokens: number; -} - -/** - * 响应基类 - */ -export interface RespBase { - /** - * 时间戳 - */ - created: number; - /** - * 本轮对话的id - */ - id: string; - /** - * 表示当前子句是否是最后一句。只有在流式接口模式下会返回该字段 - */ - is_end?: boolean; - /** - * 1：表示输入内容无安全风险 - * 0：表示输入内容有安全风险 - */ - is_safe?: number; - /** - * 回包类型。 - * - * chat.completion：多轮对话返回 - */ - object: string; - /** - * 对话返回结果 - */ - result: string; - /** - * 表示当前子句的序号。只有在流式接口模式下会返回该字段 - */ - sentence_id?: number; - /** - * token统计信息，token数 = 汉字数+单词数*1.3 （仅为估算逻辑） - */ - usage: TokenUsage; -} - -export interface ChatResp extends RespBase { - /** - * 当 need_clear_history 为 true 时，此字段会告知第几轮对话有敏感信息，如果是当前问题，ban_round=-1 - */ - ban_round: number; - /** - * 输出内容标识，说明： - * · normal：输出内容完全由大模型生成，未触发截断、替换 - * · stop：输出结果命中入参stop中指定的字段后被截断 - * · length：达到了最大的token数，根据EB返回结果is_truncated来截断 - * · content_filter：输出内容被截断、兜底、替换为**等 - */ - finish_reason: string; - /** - * 当前生成的结果是否被截断 - */ - is_truncated?: boolean; - /** - * 表示用户输入是否存在安全，是否关闭当前会话，清理历史会话信息 - * - * true：是，表示用户输入存在安全风险，建议关闭当前会话，清理历史会话信息 - * false：否，表示用户输入无安全风险 - */ - need_clear_history: boolean; -} diff --git a/src/locales/default/modelProvider.ts b/src/locales/default/modelProvider.ts index f86f94035e9b6..b554a822ce9ac 100644 --- a/src/locales/default/modelProvider.ts +++ b/src/locales/default/modelProvider.ts @@ -291,26 +291,6 @@ export default { tooltip: '更新服务商基础配置', updateSuccess: '更新成功', }, - wenxin: { - accessKey: { - desc: '填入百度千帆平台的 Access Key', - placeholder: 'Qianfan Access Key', - title: 'Access Key', - }, - checker: { - desc: '测试 AccessKey / SecretAccess 是否填写正确', - }, - secretKey: { - desc: '填入百度千帆平台 Secret Key', - placeholder: 'Qianfan Secret Key', - title: 'Secret Key', - }, - unlock: { - customRegion: '自定义服务区域', - description: '输入你的 AccessKey / SecretKey 即可开始会话。应用不会记录你的鉴权配置', - title: '使用自定义文心一言鉴权信息', - }, - }, zeroone: { title: '01.AI 零一万物', }, diff --git a/src/server/modules/AgentRuntime/index.test.ts b/src/server/modules/AgentRuntime/index.test.ts index dc4601c59a798..1d853315d74da 100644 --- a/src/server/modules/AgentRuntime/index.test.ts +++ b/src/server/modules/AgentRuntime/index.test.ts @@ -25,7 +25,6 @@ import { } from '@/libs/agent-runtime'; import { AgentRuntime } from '@/libs/agent-runtime'; import { LobeStepfunAI } from '@/libs/agent-runtime/stepfun'; -import LobeWenxinAI from '@/libs/agent-runtime/wenxin'; import { initAgentRuntimeWithUserPayload } from './index'; @@ -55,9 +54,6 @@ vi.mock('@/config/llm', () => ({ TOGETHERAI_API_KEY: 'test-togetherai-key', QWEN_API_KEY: 'test-qwen-key', STEPFUN_API_KEY: 'test-stepfun-key', - - WENXIN_ACCESS_KEY: 'test-wenxin-access-key', - WENXIN_SECRET_KEY: 'test-wenxin-secret-key', })), })); @@ -207,16 +203,6 @@ describe('initAgentRuntimeWithUserPayload method', () => { expect(runtime['_runtime']).toBeInstanceOf(LobeStepfunAI); }); - it.skip('Wenxin AI provider: with apikey', async () => { - const jwtPayload: JWTPayload = { - wenxinAccessKey: 'user-wenxin-accessKey', - wenxinSecretKey: 'wenxin-secret-key', - }; - const runtime = await initAgentRuntimeWithUserPayload(ModelProvider.Wenxin, jwtPayload); - expect(runtime).toBeInstanceOf(AgentRuntime); - expect(runtime['_runtime']).toBeInstanceOf(LobeWenxinAI); - }); - it('Unknown Provider: with apikey and endpoint, should initialize to OpenAi', async () => { const jwtPayload: JWTPayload = { apiKey: 'user-unknown-key', @@ -364,13 +350,6 @@ describe('initAgentRuntimeWithUserPayload method', () => { expect(runtime['_runtime']).toBeInstanceOf(LobeTogetherAI); }); - it.skip('Wenxin AI provider: without apikey', async () => { - const jwtPayload = {}; - const runtime = await initAgentRuntimeWithUserPayload(ModelProvider.Wenxin, jwtPayload); - expect(runtime).toBeInstanceOf(AgentRuntime); - expect(runtime['_runtime']).toBeInstanceOf(LobeWenxinAI); - }); - it('OpenAI provider: without apikey with OPENAI_PROXY_URL', async () => { process.env.OPENAI_PROXY_URL = 'https://proxy.example.com/v1'; diff --git a/src/services/_auth.ts b/src/services/_auth.ts index 098f42959e66b..7a3fc2fa49c5e 100644 --- a/src/services/_auth.ts +++ b/src/services/_auth.ts @@ -9,7 +9,6 @@ import { AzureOpenAIKeyVault, CloudflareKeyVault, OpenAICompatibleKeyVault, - WenxinKeyVault, } from '@/types/user/settings'; import { createJWT } from '@/utils/jwt'; @@ -18,7 +17,6 @@ export const getProviderAuthPayload = ( keyVaults: OpenAICompatibleKeyVault & AzureOpenAIKeyVault & AWSBedrockKeyVault & - WenxinKeyVault & CloudflareKeyVault, ) => { switch (provider) { @@ -47,18 +45,6 @@ export const getProviderAuthPayload = ( }; } - case ModelProvider.Wenxin: { - const { secretKey, accessKey } = keyVaults; - - const apiKey = (accessKey || '') + (secretKey || ''); - - return { - apiKey, - wenxinAccessKey: accessKey, - wenxinSecretKey: secretKey, - }; - } - case ModelProvider.Azure: { return { apiKey: keyVaults.apiKey, diff --git a/src/store/user/slices/modelList/selectors/keyVaults.ts b/src/store/user/slices/modelList/selectors/keyVaults.ts index 684c95baa28b1..a564cfde4117f 100644 --- a/src/store/user/slices/modelList/selectors/keyVaults.ts +++ b/src/store/user/slices/modelList/selectors/keyVaults.ts @@ -14,7 +14,6 @@ export const keyVaultsSettings = (s: UserStore): UserKeyVaults => const openAIConfig = (s: UserStore) => keyVaultsSettings(s).openai || {}; const bedrockConfig = (s: UserStore) => keyVaultsSettings(s).bedrock || {}; -const wenxinConfig = (s: UserStore) => keyVaultsSettings(s).wenxin || {}; const ollamaConfig = (s: UserStore) => keyVaultsSettings(s).ollama || {}; const azureConfig = (s: UserStore) => keyVaultsSettings(s).azure || {}; const cloudflareConfig = (s: UserStore) => keyVaultsSettings(s).cloudflare || {}; @@ -45,5 +44,4 @@ export const keyVaultsConfigSelectors = { ollamaConfig, openAIConfig, password, - wenxinConfig, }; diff --git a/src/types/aiProvider.ts b/src/types/aiProvider.ts index a5d2ae9cd0bad..403a711c63d6a 100644 --- a/src/types/aiProvider.ts +++ b/src/types/aiProvider.ts @@ -23,7 +23,6 @@ export const AiProviderSDKEnum = { Huggingface: 'huggingface', Ollama: 'ollama', Openai: 'openai', - Wenxin: 'wenxin', } as const; export type AiProviderSDKType = (typeof AiProviderSDKEnum)[keyof typeof AiProviderSDKEnum]; diff --git a/src/types/user/settings/keyVaults.ts b/src/types/user/settings/keyVaults.ts index 97c910c37bdd3..ec4ffc23e1e31 100644 --- a/src/types/user/settings/keyVaults.ts +++ b/src/types/user/settings/keyVaults.ts @@ -25,11 +25,6 @@ export interface CloudflareKeyVault { baseURLOrAccountID?: string; } -export interface WenxinKeyVault { - accessKey?: string; - secretKey?: string; -} - export interface UserKeyVaults { ai21?: OpenAICompatibleKeyVault; ai360?: OpenAICompatibleKeyVault; @@ -68,7 +63,7 @@ export interface UserKeyVaults { taichu?: OpenAICompatibleKeyVault; togetherai?: OpenAICompatibleKeyVault; upstage?: OpenAICompatibleKeyVault; - wenxin?: WenxinKeyVault; + wenxin?: OpenAICompatibleKeyVault; xai?: OpenAICompatibleKeyVault; zeroone?: OpenAICompatibleKeyVault; zhipu?: OpenAICompatibleKeyVault; From 9c28a583d9e552bb69809e6aa437a88994bf1a11 Mon Sep 17 00:00:00 2001 From: Zhijie He Date: Tue, 4 Feb 2025 21:15:32 +0800 Subject: [PATCH 2/4] =?UTF-8?q?=F0=9F=92=84=20style:=20update=20model=20li?= =?UTF-8?q?st?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit --- src/config/aiModels/wenxin.ts | 97 +++++++++++++++++++++++++++++ src/config/modelProviders/wenxin.ts | 71 +++++++++++++++++++++ 2 files changed, 168 insertions(+) diff --git a/src/config/aiModels/wenxin.ts b/src/config/aiModels/wenxin.ts index 47ba9d7a689ee..94cfeccab313f 100644 --- a/src/config/aiModels/wenxin.ts +++ b/src/config/aiModels/wenxin.ts @@ -2,6 +2,9 @@ import { AIChatModelCard } from '@/types/aiModel'; const wenxinChatModels: AIChatModelCard[] = [ { + abilities: { + functionCall: true, + }, contextWindowTokens: 8192, description: '百度自研的旗舰级大规模⼤语⾔模型，覆盖海量中英文语料，具有强大的通用能力，可满足绝大部分对话问答、创作生成、插件应用场景要求；支持自动对接百度搜索插件，保障问答信息时效。', @@ -16,6 +19,9 @@ const wenxinChatModels: AIChatModelCard[] = [ type: 'chat', }, { + abilities: { + functionCall: true, + }, contextWindowTokens: 8192, description: '百度自研的旗舰级大规模⼤语⾔模型，覆盖海量中英文语料，具有强大的通用能力，可满足绝大部分对话问答、创作生成、插件应用场景要求；支持自动对接百度搜索插件，保障问答信息时效。', @@ -29,6 +35,9 @@ const wenxinChatModels: AIChatModelCard[] = [ type: 'chat', }, { + abilities: { + functionCall: true, + }, contextWindowTokens: 128_000, description: '百度自研的旗舰级大规模⼤语⾔模型，覆盖海量中英文语料，具有强大的通用能力，可满足绝大部分对话问答、创作生成、插件应用场景要求；支持自动对接百度搜索插件，保障问答信息时效。', @@ -43,6 +52,9 @@ const wenxinChatModels: AIChatModelCard[] = [ type: 'chat', }, { + abilities: { + functionCall: true, + }, contextWindowTokens: 8192, description: '百度自研的旗舰级超大规模⼤语⾔模型，相较ERNIE 3.5实现了模型能力全面升级，广泛适用于各领域复杂任务场景；支持自动对接百度搜索插件，保障问答信息时效。', @@ -57,6 +69,9 @@ const wenxinChatModels: AIChatModelCard[] = [ type: 'chat', }, { + abilities: { + functionCall: true, + }, contextWindowTokens: 8192, description: '百度自研的旗舰级超大规模⼤语⾔模型，相较ERNIE 3.5实现了模型能力全面升级，广泛适用于各领域复杂任务场景；支持自动对接百度搜索插件，保障问答信息时效。', @@ -70,6 +85,9 @@ const wenxinChatModels: AIChatModelCard[] = [ type: 'chat', }, { + abilities: { + functionCall: true, + }, contextWindowTokens: 8192, description: '百度自研的旗舰级超大规模⼤语⾔模型，综合效果表现出色，广泛适用于各领域复杂任务场景；支持自动对接百度搜索插件，保障问答信息时效。相较于ERNIE 4.0在性能表现上更优秀', @@ -84,6 +102,9 @@ const wenxinChatModels: AIChatModelCard[] = [ type: 'chat', }, { + abilities: { + functionCall: true, + }, contextWindowTokens: 128_000, description: '百度自研的旗舰级超大规模⼤语⾔模型，综合效果表现出色，广泛适用于各领域复杂任务场景；支持自动对接百度搜索插件，保障问答信息时效。相较于ERNIE 4.0在性能表现上更优秀', @@ -98,6 +119,9 @@ const wenxinChatModels: AIChatModelCard[] = [ type: 'chat', }, { + abilities: { + functionCall: true, + }, contextWindowTokens: 8192, description: '百度自研的旗舰级超大规模⼤语⾔模型，综合效果表现出色，广泛适用于各领域复杂任务场景；支持自动对接百度搜索插件，保障问答信息时效。相较于ERNIE 4.0在性能表现上更优秀', @@ -111,6 +135,23 @@ const wenxinChatModels: AIChatModelCard[] = [ type: 'chat', }, { + contextWindowTokens: 8192, + description: + 'ERNIE Lite是百度自研的轻量级大语言模型，兼顾优异的模型效果与推理性能，适合低算力AI加速卡推理使用。', + displayName: 'ERNIE Lite 8K', + enabled: true, + id: 'ernie-lite-8k', + pricing: { + currency: 'CNY', + input: 0, + output: 0, + }, + type: 'chat', + }, + { + abilities: { + functionCall: true, + }, contextWindowTokens: 128_000, description: '百度自研的轻量级大语言模型，兼顾优异的模型效果与推理性能，效果比ERNIE Lite更优，适合低算力AI加速卡推理使用。', @@ -124,6 +165,20 @@ const wenxinChatModels: AIChatModelCard[] = [ }, type: 'chat', }, + { + contextWindowTokens: 8192, + description: + 'ERNIE Tiny是百度自研的超高性能大语言模型，部署与精调成本在文心系列模型中最低。', + displayName: 'ERNIE Tiny 8K', + enabled: true, + id: 'ernie-tiny-8k', + pricing: { + currency: 'CNY', + input: 0, + output: 0, + }, + type: 'chat', + }, { contextWindowTokens: 128_000, description: @@ -164,6 +219,48 @@ const wenxinChatModels: AIChatModelCard[] = [ }, type: 'chat', }, + { + contextWindowTokens: 8192, + description: + '2024年5月23日发布的版本，支持8K上下文长度，在情节演绎和规则化文本等场景下指令遵循能力更强', + displayName: 'ERNIE Character Fiction 8K', + id: 'ernie-char-fiction-8k', + pricing: { + currency: 'CNY', + input: 4, + output: 8, + }, + type: 'chat', + }, + { + contextWindowTokens: 65_536, + description: + 'DeepSeek-V3 为杭州深度求索人工智能基础技术研究有限公司自研的 MoE 模型，其多项评测成绩突出，在主流榜单中位列开源模型榜首。V3 相比 V2.5 模型生成速度实现 3 倍提升，为用户带来更加迅速流畅的使用体验。', + displayName: 'DeepSeek V3', + id: 'deepseek-v3', + pricing: { + currency: 'CNY', + input: 0.8, + output: 1.6, + }, + type: 'chat', + }, + { + abilities: { + reasoning: true, + }, + contextWindowTokens: 65_536, + description: + 'DeepSeek-R1 在后训练阶段大规模使用了强化学习技术，在仅有极少标注数据的情况下，极大提升了模型推理能力。在数学、代码、自然语言推理等任务上，性能比肩 OpenAI o1 正式版。', + displayName: 'DeepSeek R1', + id: 'deepseek-r1', + pricing: { + currency: 'CNY', + input: 2, + output: 8, + }, + type: 'chat', + }, ]; export const allModels = [...wenxinChatModels]; diff --git a/src/config/modelProviders/wenxin.ts b/src/config/modelProviders/wenxin.ts index 13124a990dd3e..5b1e0317124ac 100644 --- a/src/config/modelProviders/wenxin.ts +++ b/src/config/modelProviders/wenxin.ts @@ -9,6 +9,7 @@ const BaiduWenxin: ModelProviderCard = { '百度自研的旗舰级大规模⼤语⾔模型，覆盖海量中英文语料，具有强大的通用能力，可满足绝大部分对话问答、创作生成、插件应用场景要求；支持自动对接百度搜索插件，保障问答信息时效。', displayName: 'ERNIE 3.5 8K', enabled: true, + functionCall: true, id: 'ernie-3.5-8k', pricing: { currency: 'CNY', @@ -21,6 +22,7 @@ const BaiduWenxin: ModelProviderCard = { description: '百度自研的旗舰级大规模⼤语⾔模型，覆盖海量中英文语料，具有强大的通用能力，可满足绝大部分对话问答、创作生成、插件应用场景要求；支持自动对接百度搜索插件，保障问答信息时效。', displayName: 'ERNIE 3.5 8K Preview', + functionCall: true, id: 'ernie-3.5-8k-preview', pricing: { currency: 'CNY', @@ -34,6 +36,7 @@ const BaiduWenxin: ModelProviderCard = { '百度自研的旗舰级大规模⼤语⾔模型，覆盖海量中英文语料，具有强大的通用能力，可满足绝大部分对话问答、创作生成、插件应用场景要求；支持自动对接百度搜索插件，保障问答信息时效。', displayName: 'ERNIE 3.5 128K', enabled: true, + functionCall: true, id: 'ernie-3.5-128k', pricing: { currency: 'CNY', @@ -47,6 +50,7 @@ const BaiduWenxin: ModelProviderCard = { '百度自研的旗舰级超大规模⼤语⾔模型，相较ERNIE 3.5实现了模型能力全面升级，广泛适用于各领域复杂任务场景；支持自动对接百度搜索插件，保障问答信息时效。', displayName: 'ERNIE 4.0 8K', enabled: true, + functionCall: true, id: 'ernie-4.0-8k-latest', pricing: { currency: 'CNY', @@ -59,6 +63,7 @@ const BaiduWenxin: ModelProviderCard = { description: '百度自研的旗舰级超大规模⼤语⾔模型，相较ERNIE 3.5实现了模型能力全面升级，广泛适用于各领域复杂任务场景；支持自动对接百度搜索插件，保障问答信息时效。', displayName: 'ERNIE 4.0 8K Preview', + functionCall: true, id: 'ernie-4.0-8k-preview', pricing: { currency: 'CNY', @@ -72,6 +77,7 @@ const BaiduWenxin: ModelProviderCard = { '百度自研的旗舰级超大规模⼤语⾔模型，综合效果表现出色，广泛适用于各领域复杂任务场景；支持自动对接百度搜索插件，保障问答信息时效。相较于ERNIE 4.0在性能表现上更优秀', displayName: 'ERNIE 4.0 Turbo 8K', enabled: true, + functionCall: true, id: 'ernie-4.0-turbo-8k-latest', pricing: { currency: 'CNY', @@ -85,6 +91,7 @@ const BaiduWenxin: ModelProviderCard = { '百度自研的旗舰级超大规模⼤语⾔模型，综合效果表现出色，广泛适用于各领域复杂任务场景；支持自动对接百度搜索插件，保障问答信息时效。相较于ERNIE 4.0在性能表现上更优秀', displayName: 'ERNIE 4.0 Turbo 128K', enabled: true, + functionCall: true, id: 'ernie-4.0-turbo-128k', pricing: { currency: 'CNY', @@ -97,6 +104,7 @@ const BaiduWenxin: ModelProviderCard = { description: '百度自研的旗舰级超大规模⼤语⾔模型，综合效果表现出色，广泛适用于各领域复杂任务场景；支持自动对接百度搜索插件，保障问答信息时效。相较于ERNIE 4.0在性能表现上更优秀', displayName: 'ERNIE 4.0 Turbo 8K Preview', + functionCall: true, id: 'ernie-4.0-turbo-8k-preview', pricing: { currency: 'CNY', @@ -104,12 +112,26 @@ const BaiduWenxin: ModelProviderCard = { output: 60, }, }, + { + contextWindowTokens: 8192, + description: + 'ERNIE Lite是百度自研的轻量级大语言模型，兼顾优异的模型效果与推理性能，适合低算力AI加速卡推理使用。', + displayName: 'ERNIE Lite 8K', + enabled: true, + id: 'ernie-lite-8k', + pricing: { + currency: 'CNY', + input: 0, + output: 0, + }, + }, { contextWindowTokens: 128_000, description: '百度自研的轻量级大语言模型，兼顾优异的模型效果与推理性能，效果比ERNIE Lite更优，适合低算力AI加速卡推理使用。', displayName: 'ERNIE Lite Pro 128K', enabled: true, + functionCall: true, id: 'ernie-lite-pro-128k', pricing: { currency: 'CNY', @@ -117,6 +139,19 @@ const BaiduWenxin: ModelProviderCard = { output: 0.4, }, }, + { + contextWindowTokens: 8192, + description: + 'ERNIE Tiny是百度自研的超高性能大语言模型，部署与精调成本在文心系列模型中最低。', + displayName: 'ERNIE Tiny 8K', + enabled: true, + id: 'ernie-tiny-8k', + pricing: { + currency: 'CNY', + input: 0, + output: 0, + }, + }, { contextWindowTokens: 128_000, description: @@ -154,6 +189,42 @@ const BaiduWenxin: ModelProviderCard = { output: 8, }, }, + { + contextWindowTokens: 8192, + description: + '2024年5月23日发布的版本，支持8K上下文长度，在情节演绎和规则化文本等场景下指令遵循能力更强', + displayName: 'ERNIE Character Fiction 8K', + id: 'ernie-char-fiction-8k', + pricing: { + currency: 'CNY', + input: 4, + output: 8, + }, + }, + { + contextWindowTokens: 65_536, + description: + 'DeepSeek-V3 为杭州深度求索人工智能基础技术研究有限公司自研的 MoE 模型，其多项评测成绩突出，在主流榜单中位列开源模型榜首。V3 相比 V2.5 模型生成速度实现 3 倍提升，为用户带来更加迅速流畅的使用体验。', + displayName: 'DeepSeek V3', + id: 'deepseek-v3', + pricing: { + currency: 'CNY', + input: 0.8, + output: 1.6, + }, + }, + { + contextWindowTokens: 65_536, + description: + 'DeepSeek-R1 在后训练阶段大规模使用了强化学习技术，在仅有极少标注数据的情况下，极大提升了模型推理能力。在数学、代码、自然语言推理等任务上，性能比肩 OpenAI o1 正式版。', + displayName: 'DeepSeek R1', + id: 'deepseek-r1', + pricing: { + currency: 'CNY', + input: 2, + output: 8, + }, + }, ], checkModel: 'ernie-speed-128k', description: From 15a64cfaa4cfc95cefdbe2cc294b8c1bf17083e6 Mon Sep 17 00:00:00 2001 From: Zhijie He Date: Tue, 4 Feb 2025 21:51:23 +0800 Subject: [PATCH 3/4] =?UTF-8?q?=F0=9F=93=9D=20docs:=20update=20docs?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit --- docs/usage/providers/wenxin.mdx | 29 +++++++++++++++------------ docs/usage/providers/wenxin.zh-CN.mdx | 19 ++++++++++-------- 2 files changed, 27 insertions(+), 21 deletions(-) diff --git a/docs/usage/providers/wenxin.mdx b/docs/usage/providers/wenxin.mdx index c20b35dcf4541..9e78d32c98d5c 100644 --- a/docs/usage/providers/wenxin.mdx +++ b/docs/usage/providers/wenxin.mdx @@ -22,27 +22,30 @@ This article will guide you on how to use Wenxin Qianfan in LobeChat. ### Step 1: Obtain the Wenxin Qianfan API Key - - Register and log in to the [Baidu Intelligent Cloud Console](https://console.bce.baidu.com/) - - Navigate to `Baidu Intelligent Cloud Qianfan ModelBuilder` - - Choose `Application Access` from the left-side menu - - Create an application + - Register and log in to the [Baidu AI Cloud Console](https://console.bce.baidu.com/) + - Navigate to `Baidu AI Cloud Qianfan ModelBuilder` + - Select `API Key` from the left menu - {'Create + {'API - - Enter the `Security Authentication` -> `Access Key` management page from the user account menu - - Copy the `Access Key` and `Secret Key`, and store them safely + - Click `Create API Key` + - In `Service`, select `Qianfan ModelBuilder` + - In `Resource`, choose `All Resources` + - Click the `Confirm` button + - Copy the `API Key` and keep it safe - {'Save + {'Create + {'Copy ### Step 2: Configure Wenxin Qianfan in LobeChat - - Go to the `Settings` interface in LobeChat - - Locate the settings for `Wenxin Qianfan` under `Language Model` + - Go to the `Settings` page of LobeChat + - Under `Language Models`, find the `Wenxin Qianfan` settings - {'Enter + {'Enter - - Enter the obtained `Access Key` and `Secret Key` - - Select a Wenxin Qianfan model for your AI assistant to start interacting + - Enter the obtained `API Key` + - Select a Wenxin Qianfan model for your AI assistant, and you're ready to start chatting! {'Select diff --git a/docs/usage/providers/wenxin.zh-CN.mdx b/docs/usage/providers/wenxin.zh-CN.mdx index a2b44bd047650..184f574d75e53 100644 --- a/docs/usage/providers/wenxin.zh-CN.mdx +++ b/docs/usage/providers/wenxin.zh-CN.mdx @@ -22,24 +22,27 @@ tags: - 注册并登录 [百度智能云控制台](https://console.bce.baidu.com/) - 进入 `百度智能云千帆 ModelBuilder` - - 在左侧菜单中选择`应用接入` - - 创建一个应用 + - 在左侧菜单中选择 `API Key` - {'创建应用'} + {'API - - 在用户账号菜单进入 `安全认证` -> `Access Key` 管理页面 - - 复制 `Access Key` 和 `Secret Key`，并妥善保存 + - 点击创建 API Key + - 在 `服务` 中选择 `千帆ModelBuilder` + - 在 `资源` 中选择 `所有资源` + - 点击 `确定` 按钮 + - 复制 `API Key` 并妥善保存 - {'保存密钥'} + {'创建密钥'} + {'复制密钥'} ### 步骤二：在 LobeChat 中配置文心千帆 - 访问 LobeChat 的`设置`界面 - 在`语言模型`下找到 `文心千帆` 的设置项 - {'填入 + {'填入 - - 填入获得的 `Access Key` 和 `Secret Key` + - 填入获得的 `API Key` - 为你的 AI 助手选择一个文心千帆的模型即可开始对话 {'选择文心千帆模型并开始对话'} From 17d89ffc7ef3880157b1d4226a743ca3c7ed590c Mon Sep 17 00:00:00 2001 From: Zhijie He Date: Wed, 5 Feb 2025 06:33:18 +0800 Subject: [PATCH 4/4] =?UTF-8?q?=F0=9F=92=84=20style:=20update=20model=20li?= =?UTF-8?q?st?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit --- src/config/aiModels/wenxin.ts | 41 ++++++++++++++++++----------- src/config/modelProviders/wenxin.ts | 40 +++++++++++++++++----------- 2 files changed, 49 insertions(+), 32 deletions(-) diff --git a/src/config/aiModels/wenxin.ts b/src/config/aiModels/wenxin.ts index 94cfeccab313f..cd193788f9e53 100644 --- a/src/config/aiModels/wenxin.ts +++ b/src/config/aiModels/wenxin.ts @@ -139,7 +139,6 @@ const wenxinChatModels: AIChatModelCard[] = [ description: 'ERNIE Lite是百度自研的轻量级大语言模型，兼顾优异的模型效果与推理性能，适合低算力AI加速卡推理使用。', displayName: 'ERNIE Lite 8K', - enabled: true, id: 'ernie-lite-8k', pricing: { currency: 'CNY', @@ -156,7 +155,6 @@ const wenxinChatModels: AIChatModelCard[] = [ description: '百度自研的轻量级大语言模型，兼顾优异的模型效果与推理性能，效果比ERNIE Lite更优，适合低算力AI加速卡推理使用。', displayName: 'ERNIE Lite Pro 128K', - enabled: true, id: 'ernie-lite-pro-128k', pricing: { currency: 'CNY', @@ -170,7 +168,6 @@ const wenxinChatModels: AIChatModelCard[] = [ description: 'ERNIE Tiny是百度自研的超高性能大语言模型，部署与精调成本在文心系列模型中最低。', displayName: 'ERNIE Tiny 8K', - enabled: true, id: 'ernie-tiny-8k', pricing: { currency: 'CNY', @@ -182,27 +179,26 @@ const wenxinChatModels: AIChatModelCard[] = [ { contextWindowTokens: 128_000, description: - '百度2024年最新发布的自研高性能大语言模型，通用能力优异，效果比ERNIE Speed更优，适合作为基座模型进行精调，更好地处理特定场景问题，同时具备极佳的推理性能。', - displayName: 'ERNIE Speed Pro 128K', - enabled: true, - id: 'ernie-speed-pro-128k', + '百度2024年最新发布的自研高性能大语言模型，通用能力优异，适合作为基座模型进行精调，更好地处理特定场景问题，同时具备极佳的推理性能。', + displayName: 'ERNIE Speed 128K', + id: 'ernie-speed-128k', pricing: { currency: 'CNY', - input: 0.3, - output: 0.6, + input: 0, + output: 0, }, type: 'chat', }, { contextWindowTokens: 128_000, description: - '百度2024年最新发布的自研高性能大语言模型，通用能力优异，适合作为基座模型进行精调，更好地处理特定场景问题，同时具备极佳的推理性能。', - displayName: 'ERNIE Speed 128K', - id: 'ernie-speed-128k', + '百度2024年最新发布的自研高性能大语言模型，通用能力优异，效果比ERNIE Speed更优，适合作为基座模型进行精调，更好地处理特定场景问题，同时具备极佳的推理性能。', + displayName: 'ERNIE Speed Pro 128K', + id: 'ernie-speed-pro-128k', pricing: { currency: 'CNY', - input: 0, - output: 0, + input: 0.3, + output: 0.6, }, type: 'chat', }, @@ -211,7 +207,7 @@ const wenxinChatModels: AIChatModelCard[] = [ description: '百度自研的垂直场景大语言模型，适合游戏NPC、客服对话、对话角色扮演等应用场景，人设风格更为鲜明、一致，指令遵循能力更强，推理性能更优。', displayName: 'ERNIE Character 8K', - id: 'ernie-character-8k', + id: 'ernie-char-8k', pricing: { currency: 'CNY', input: 4, @@ -222,7 +218,7 @@ const wenxinChatModels: AIChatModelCard[] = [ { contextWindowTokens: 8192, description: - '2024年5月23日发布的版本，支持8K上下文长度，在情节演绎和规则化文本等场景下指令遵循能力更强', + '百度自研的垂直场景大语言模型，适合游戏NPC、客服对话、对话角色扮演等应用场景，人设风格更为鲜明、一致，指令遵循能力更强，推理性能更优。', displayName: 'ERNIE Character Fiction 8K', id: 'ernie-char-fiction-8k', pricing: { @@ -232,6 +228,19 @@ const wenxinChatModels: AIChatModelCard[] = [ }, type: 'chat', }, + { + contextWindowTokens: 8192, + description: + '百度自研通用大语言模型，在小说续写能力上有明显优势，也可用在短剧、电影等场景。', + displayName: 'ERNIE Novel 8K', + id: 'ernie-novel-8k', + pricing: { + currency: 'CNY', + input: 40, + output: 120, + }, + type: 'chat', + }, { contextWindowTokens: 65_536, description: diff --git a/src/config/modelProviders/wenxin.ts b/src/config/modelProviders/wenxin.ts index 5b1e0317124ac..02305b0479e8b 100644 --- a/src/config/modelProviders/wenxin.ts +++ b/src/config/modelProviders/wenxin.ts @@ -117,7 +117,6 @@ const BaiduWenxin: ModelProviderCard = { description: 'ERNIE Lite是百度自研的轻量级大语言模型，兼顾优异的模型效果与推理性能，适合低算力AI加速卡推理使用。', displayName: 'ERNIE Lite 8K', - enabled: true, id: 'ernie-lite-8k', pricing: { currency: 'CNY', @@ -130,7 +129,6 @@ const BaiduWenxin: ModelProviderCard = { description: '百度自研的轻量级大语言模型，兼顾优异的模型效果与推理性能，效果比ERNIE Lite更优，适合低算力AI加速卡推理使用。', displayName: 'ERNIE Lite Pro 128K', - enabled: true, functionCall: true, id: 'ernie-lite-pro-128k', pricing: { @@ -144,7 +142,6 @@ const BaiduWenxin: ModelProviderCard = { description: 'ERNIE Tiny是百度自研的超高性能大语言模型，部署与精调成本在文心系列模型中最低。', displayName: 'ERNIE Tiny 8K', - enabled: true, id: 'ernie-tiny-8k', pricing: { currency: 'CNY', @@ -155,26 +152,25 @@ const BaiduWenxin: ModelProviderCard = { { contextWindowTokens: 128_000, description: - '百度2024年最新发布的自研高性能大语言模型，通用能力优异，效果比ERNIE Speed更优，适合作为基座模型进行精调，更好地处理特定场景问题，同时具备极佳的推理性能。', - displayName: 'ERNIE Speed Pro 128K', - enabled: true, - id: 'ernie-speed-pro-128k', + '百度2024年最新发布的自研高性能大语言模型，通用能力优异，适合作为基座模型进行精调，更好地处理特定场景问题，同时具备极佳的推理性能。', + displayName: 'ERNIE Speed 128K', + id: 'ernie-speed-128k', pricing: { currency: 'CNY', - input: 0.3, - output: 0.6, + input: 0, + output: 0, }, }, { contextWindowTokens: 128_000, description: - '百度2024年最新发布的自研高性能大语言模型，通用能力优异，适合作为基座模型进行精调，更好地处理特定场景问题，同时具备极佳的推理性能。', - displayName: 'ERNIE Speed 128K', - id: 'ernie-speed-128k', + '百度2024年最新发布的自研高性能大语言模型，通用能力优异，效果比ERNIE Speed更优，适合作为基座模型进行精调，更好地处理特定场景问题，同时具备极佳的推理性能。', + displayName: 'ERNIE Speed Pro 128K', + id: 'ernie-speed-pro-128k', pricing: { currency: 'CNY', - input: 0, - output: 0, + input: 0.3, + output: 0.6, }, }, { @@ -182,7 +178,7 @@ const BaiduWenxin: ModelProviderCard = { description: '百度自研的垂直场景大语言模型，适合游戏NPC、客服对话、对话角色扮演等应用场景，人设风格更为鲜明、一致，指令遵循能力更强，推理性能更优。', displayName: 'ERNIE Character 8K', - id: 'ernie-character-8k', + id: 'ernie-char-8k', pricing: { currency: 'CNY', input: 4, @@ -192,7 +188,7 @@ const BaiduWenxin: ModelProviderCard = { { contextWindowTokens: 8192, description: - '2024年5月23日发布的版本，支持8K上下文长度，在情节演绎和规则化文本等场景下指令遵循能力更强', + '百度自研的垂直场景大语言模型，适合游戏NPC、客服对话、对话角色扮演等应用场景，人设风格更为鲜明、一致，指令遵循能力更强，推理性能更优。', displayName: 'ERNIE Character Fiction 8K', id: 'ernie-char-fiction-8k', pricing: { @@ -201,6 +197,18 @@ const BaiduWenxin: ModelProviderCard = { output: 8, }, }, + { + contextWindowTokens: 8192, + description: + '百度自研通用大语言模型，在小说续写能力上有明显优势，也可用在短剧、电影等场景。', + displayName: 'ERNIE Novel 8K', + id: 'ernie-novel-8k', + pricing: { + currency: 'CNY', + input: 40, + output: 120, + }, + }, { contextWindowTokens: 65_536, description: