@lobehub/chat 1.20.7 → 1.21.0

package/src/config/modelProviders/wenxin.ts

@@ -0,0 +1,159 @@
+import { ModelProviderCard } from '@/types/llm';
+
+// ref https://cloud.baidu.com/doc/WENXINWORKSHOP/s/Nlks5zkzu
+const BaiduWenxin: ModelProviderCard = {
+  chatModels: [
+    {
+      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',
+      enabled: true,
+      id: 'ERNIE-3.5-8K',
+      pricing: {
+        currency: 'CNY',
+        input: 0.8,
+        output: 2,
+      },
+      tokens: 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。',
+      displayName: 'ERNIE 3.5 8K Preview',
+      id: 'ERNIE-3.5-8K-Preview',
+      pricing: {
+        currency: 'CNY',
+        input: 0.8,
+        output: 2,
+      },
+      tokens: 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。',
+      displayName: 'ERNIE 3.5 128K',
+      enabled: true,
+      id: 'ERNIE-3.5-128K',
+      pricing: {
+        currency: 'CNY',
+        input: 0.8,
+        output: 2,
+      },
+      tokens: 128_000,
+    },
+    {
+      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',
+      enabled: true,
+      id: 'ERNIE-4.0-8K-Latest',
+      pricing: {
+        currency: 'CNY',
+        input: 30,
+        output: 90,
+      },
+      tokens: 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。',
+      displayName: 'ERNIE 4.0 8K Preview',
+      id: 'ERNIE-4.0-8K-Preview',
+      pricing: {
+        currency: 'CNY',
+        input: 30,
+        output: 90,
+      },
+      tokens: 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',
+      displayName: 'ERNIE 4.0 Turbo 8K',
+      enabled: true,
+      id: 'ERNIE-4.0-Turbo-8K',
+      pricing: {
+        currency: 'CNY',
+        input: 20,
+        output: 60,
+      },
+      tokens: 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',
+      displayName: 'ERNIE 4.0 Turbo 8K Preview',
+      id: 'ERNIE-4.0-Turbo-8K-Preview',
+      pricing: {
+        currency: 'CNY',
+        input: 20,
+        output: 60,
+      },
+      tokens: 8192,
+    },
+    {
+      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',
+        input: 0.2,
+        output: 0.4,
+      },
+      tokens: 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',
+      pricing: {
+        currency: 'CNY',
+        input: 0.3,
+        output: 0.6,
+      },
+      tokens: 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',
+      pricing: {
+        currency: 'CNY',
+        input: 0,
+        output: 0,
+      },
+      tokens: 128_000,
+    },
+    {
+      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',
+      pricing: {
+        currency: 'CNY',
+        input: 4,
+        output: 8,
+      },
+      tokens: 8192,
+    },
+  ],
+  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',
+  smoothing: {
+    speed: 2,
+    text: true,
+  },
+  url: 'https://cloud.baidu.com/wenxin.html',
+};
+
+export default BaiduWenxin;

package/src/const/auth.ts

@@ -36,6 +36,10 @@ export interface JWTPayload {
   awsRegion?: string;
   awsSecretAccessKey?: string;
   awsSessionToken?: string;
+
+  wenxinAccessKey?: string;
+  wenxinSecretKey?: string;
+
   /**
    * user id
    * in client db mode it's a uuid

package/src/const/settings/llm.ts

@@ -25,6 +25,7 @@ import {
   TaichuProviderCard,
   TogetherAIProviderCard,
   UpstageProviderCard,
+  WenxinProviderCard,
   ZeroOneProviderCard,
   ZhiPuProviderCard,
   filterEnabledModels,
@@ -141,6 +142,10 @@ export const DEFAULT_LLM_CONFIG: UserModelProviderConfig = {
     enabled: false,
     enabledModels: filterEnabledModels(UpstageProviderCard),
   },
+  wenxin: {
+    enabled: false,
+    enabledModels: filterEnabledModels(WenxinProviderCard),
+  },
   zeroone: {
     enabled: false,
     enabledModels: filterEnabledModels(ZeroOneProviderCard),

package/src/features/Conversation/Error/APIKeyForm/Wenxin.tsx

@@ -0,0 +1,49 @@
+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 (
+    <FormAction
+      avatar={<Wenxin.Color size={56} />}
+      description={t('wenxin.unlock.description')}
+      title={t('wenxin.unlock.title')}
+    >
+      <Input.Password
+        autoComplete={'new-password'}
+        onChange={(e) => {
+          setConfig(ModelProvider.Wenxin, { accessKey: e.target.value });
+        }}
+        placeholder={'Access Key'}
+        type={'block'}
+        value={accessKey}
+      />
+      <Input.Password
+        autoComplete={'new-password'}
+        onChange={(e) => {
+          setConfig(ModelProvider.Wenxin, { secretKey: e.target.value });
+        }}
+        placeholder={'Secret Key'}
+        type={'block'}
+        value={secretKey}
+      />
+    </FormAction>
+  );
+});
+
+export default WenxinForm;

package/src/features/Conversation/Error/APIKeyForm/index.tsx

@@ -10,6 +10,7 @@ import { GlobalLLMProviderKey } from '@/types/user/settings';
 
 import BedrockForm from './Bedrock';
 import ProviderApiKeyForm from './ProviderApiKeyForm';
+import WenxinForm from './Wenxin';
 
 interface APIKeyFormProps {
   id: string;
@@ -65,6 +66,8 @@ const APIKeyForm = memo<APIKeyFormProps>(({ id, provider }) => {
     <Center gap={16} style={{ maxWidth: 300 }}>
       {provider === ModelProvider.Bedrock ? (
         <BedrockForm />
+      ) : provider === ModelProvider.Wenxin ? (
+        <WenxinForm />
       ) : (
         <ProviderApiKeyForm
           apiKeyPlaceholder={apiKeyPlaceholder}

package/src/features/Conversation/Error/index.tsx

@@ -34,6 +34,7 @@ const getErrorAlertConfig = (
     };
 
   switch (errorType) {
+    case AgentRuntimeErrorType.QuotaLimitReached:
     case AgentRuntimeErrorType.LocationNotSupportError: {
       return {
         type: 'warning',

package/src/libs/agent-runtime/AgentRuntime.test.ts

@@ -27,6 +27,7 @@ import {
   ModelProvider,
 } from '@/libs/agent-runtime';
 import { LobeStepfunAI } from '@/libs/agent-runtime/stepfun';
+import LobeWenxinAI from '@/libs/agent-runtime/wenxin';
 
 import { AgentChatOptions } from './AgentRuntime';
 import { LobeBedrockAIParams } from './bedrock';

package/src/libs/agent-runtime/error.ts

@@ -3,6 +3,7 @@
 export const AgentRuntimeErrorType = {
   AgentRuntimeError: 'AgentRuntimeError', // Agent Runtime 模块运行时错误
   LocationNotSupportError: 'LocationNotSupportError',
+  QuotaLimitReached: 'QuotaLimitReached',
 
   InvalidProviderAPIKey: 'InvalidProviderAPIKey',
   ProviderBizError: 'ProviderBizError',

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

@@ -49,6 +49,7 @@ export enum ModelProvider {
   Taichu = 'taichu',
   TogetherAI = 'togetherai',
   Upstage = 'upstage',
+  Wenxin = 'wenxin',
   ZeroOne = 'zeroone',
   ZhiPu = 'zhipu',
 }

package/src/libs/agent-runtime/utils/streams/wenxin.test.ts

@@ -0,0 +1,149 @@
+import { describe, expect, it, vi } from 'vitest';
+
+import * as uuidModule from '@/utils/uuid';
+
+import { ChatResp } from '../../wenxin/type';
+import { WenxinResultToStream, 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<ChatResp> = {
+      // @ts-ignore
+      async *[Symbol.asyncIterator]() {
+        for (const item of dataStream) {
+          yield item;
+        }
+      },
+    };
+
+    const stream = WenxinResultToStream(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);
+  });
+});

package/src/libs/agent-runtime/utils/streams/wenxin.ts

@@ -0,0 +1,46 @@
+import { readableFromAsyncIterable } from 'ai';
+
+import { ChatStreamCallbacks } from '@/libs/agent-runtime';
+import { nanoid } from '@/utils/uuid';
+
+import { ChatResp } from '../../wenxin/type';
+import {
+  StreamProtocolChunk,
+  StreamStack,
+  chatStreamable,
+  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 WenxinResultToStream = (stream: AsyncIterable<ChatResp>) => {
+  // make the response to the streamable format
+  return readableFromAsyncIterable(chatStreamable(stream));
+};
+
+export const WenxinStream = (
+  rawStream: ReadableStream<ChatResp>,
+  callbacks?: ChatStreamCallbacks,
+) => {
+  const streamStack: StreamStack = { id: 'chat_' + nanoid() };
+
+  return rawStream
+    .pipeThrough(createSSEProtocolTransformer(transformERNIEBotStream, streamStack))
+    .pipeThrough(createCallbacksTransformer(callbacks));
+};

package/src/libs/agent-runtime/wenxin/index.ts

@@ -0,0 +1,106 @@
+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 { WenxinResultToStream, 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 = WenxinResultToStream(result as AsyncIterable<ChatResp>);
+
+      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;

package/src/libs/agent-runtime/wenxin/type.ts

@@ -0,0 +1,84 @@
+/**
+ * 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;
+}

package/src/locales/default/error.ts

@@ -79,6 +79,8 @@ export default {
     InvalidClerkUser: '很抱歉，你当前尚未登录，请先登录或注册账号后继续操作',
     LocationNotSupportError:
       '很抱歉，你的所在地区不支持此模型服务，可能是由于区域限制或服务未开通。请确认当前地区是否支持使用此服务，或尝试使用切换到其他地区后重试。',
+    QuotaLimitReached:
+      '很抱歉，当前 Token 用量或请求次数已达该秘钥的配额(quota)上限，请增加该秘钥的配额或稍后再试',
 
     InvalidProviderAPIKey: '{{provider}} API Key 不正确或为空，请检查 {{provider}} API Key 后重试',
     ProviderBizError: '请求 {{provider}} 服务出错，请根据以下信息排查或重试',