@lobehub/chat 1.91.3 → 1.92.1

package/src/config/aiModels/qwen.ts

@@ -59,7 +59,7 @@ const qwenChatModels: AIChatModelCard[] = [
     contextWindowTokens: 131_072,
     description:
       'Qwen3是一款能力大幅提升的新一代通义千问大模型，在推理、通用、Agent和多语言等多个核心能力上均达到业界领先水平，并支持思考模式切换。',
-    displayName: 'Qwen3 30B',
+    displayName: 'Qwen3 30B A3B',
     enabled: true,
     id: 'qwen3-30b-a3b',
     maxOutput: 8192,
@@ -543,7 +543,7 @@ const qwenChatModels: AIChatModelCard[] = [
     config: {
       deploymentName: 'qvq-max-latest',
     },
-    contextWindowTokens: 122_880,
+    contextWindowTokens: 131_072,
     description:
       '通义千问QVQ视觉推理模型，支持视觉输入及思维链输出，在数学、编程、视觉分析、创作以及通用任务上都表现了更强的能力。',
     displayName: 'QVQ Max',
@@ -555,7 +555,30 @@ const qwenChatModels: AIChatModelCard[] = [
       input: 8,
       output: 32,
     },
-    releasedAt: '2025-03-25',
+    releasedAt: '2025-05-15',
+    type: 'chat',
+  },
+  {
+    abilities: {
+      reasoning: true,
+      vision: true,
+    },
+    config: {
+      deploymentName: 'qvq-plus-latest',
+    },
+    contextWindowTokens: 131_072,
+    description:
+      '视觉推理模型。支持视觉输入及思维链输出，继qvq-max模型后推出的plus版本，相较于qvq-max模型，qvq-plus系列模型推理速度更快，效果和成本更均衡。',
+    displayName: 'QVQ Plus',
+    id: 'qvq-plus',
+    maxOutput: 8192,
+    organization: 'Qwen',
+    pricing: {
+      currency: 'CNY',
+      input: 2,
+      output: 5,
+    },
+    releasedAt: '2025-05-15',
     type: 'chat',
   },
   {
@@ -795,9 +818,28 @@ const qwenChatModels: AIChatModelCard[] = [
     abilities: {
       reasoning: true,
     },
-    contextWindowTokens: 65_792,
+    contextWindowTokens: 65_536,
+    description:
+      '685B 满血版模型，2025年5月28日发布。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，性能较高，能力较强。',
+    displayName: 'DeepSeek R1 0528',
+    id: 'deepseek-r1-0528',
+    maxOutput: 8192,
+    organization: 'DeepSeek',
+    pricing: {
+      currency: 'CNY',
+      input: 4,
+      output: 16,
+    },
+    releasedAt: '2025-05-28',
+    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，性能较高，能力较强。',
+      '671B 满血版模型，2025年1月20日发布。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，性能较高，能力较强。',
     displayName: 'DeepSeek R1',
     id: 'deepseek-r1',
     maxOutput: 8192,
@@ -811,7 +853,7 @@ const qwenChatModels: AIChatModelCard[] = [
     type: 'chat',
   },
   {
-    contextWindowTokens: 65_792,
+    contextWindowTokens: 65_536,
     description:
       'DeepSeek-V3 为自研 MoE 模型，671B 参数，激活 37B，在 14.8T token 上进行了预训练，在长文本、代码、数学、百科、中文能力上表现优秀。',
     displayName: 'DeepSeek V3',

package/src/config/aiModels/siliconcloud.ts

@@ -215,7 +215,6 @@ const siliconcloudChatModels: AIChatModelCard[] = [
   },
   {
     abilities: {
-      functionCall: true,
       reasoning: true,
     },
     contextWindowTokens: 131_072,
@@ -381,23 +380,6 @@ const siliconcloudChatModels: AIChatModelCard[] = [
     },
     type: 'chat',
   },
-  {
-    abilities: {
-      functionCall: true,
-      reasoning: true,
-    },
-    contextWindowTokens: 131_072,
-    description:
-      'DeepSeek-R1-Distill-Qwen-1.5B 是基于 Qwen2.5-Math-1.5B 通过知识蒸馏得到的模型。该模型使用 DeepSeek-R1 生成的 80 万个精选样本进行微调，在多个基准测试中展现出不错的性能。作为一个轻量级模型，在 MATH-500 上达到了 83.9% 的准确率，在 AIME 2024 上达到了 28.9% 的通过率，在 CodeForces 上获得了 954 的评分，显示出超出其参数规模的推理能力。',
-    displayName: 'DeepSeek-R1-Distill-Qwen-1.5B (Free)',
-    id: 'deepseek-ai/DeepSeek-R1-Distill-Qwen-1.5B', // 将于 2025-06-05 下线
-    pricing: {
-      currency: 'CNY',
-      input: 0,
-      output: 0,
-    },
-    type: 'chat',
-  },
   {
     abilities: {
       functionCall: true,
@@ -415,23 +397,6 @@ const siliconcloudChatModels: AIChatModelCard[] = [
     },
     type: 'chat',
   },
-  {
-    abilities: {
-      functionCall: true,
-      reasoning: true,
-    },
-    contextWindowTokens: 131_072,
-    description:
-      'DeepSeek-R1-Distill-Qwen-1.5B 是基于 Qwen2.5-Math-1.5B 通过知识蒸馏得到的模型。该模型使用 DeepSeek-R1 生成的 80 万个精选样本进行微调，在多个基准测试中展现出不错的性能。作为一个轻量级模型，在 MATH-500 上达到了 83.9% 的准确率，在 AIME 2024 上达到了 28.9% 的通过率，在 CodeForces 上获得了 954 的评分，显示出超出其参数规模的推理能力。',
-    displayName: 'DeepSeek-R1-Distill-Qwen-1.5B (Pro)',
-    id: 'Pro/deepseek-ai/DeepSeek-R1-Distill-Qwen-1.5B', // 将于 2025-06-05 下线
-    pricing: {
-      currency: 'CNY',
-      input: 0.14,
-      output: 0.14,
-    },
-    type: 'chat',
-  },
   {
     abilities: {
       functionCall: true,
@@ -649,32 +614,6 @@ const siliconcloudChatModels: AIChatModelCard[] = [
     },
     type: 'chat',
   },
-  {
-    contextWindowTokens: 32_768,
-    description:
-      'Qwen2-1.5B-Instruct 是 Qwen2 系列中的指令微调大语言模型，参数规模为 1.5B。该模型基于 Transformer 架构，采用了 SwiGLU 激活函数、注意力 QKV 偏置和组查询注意力等技术。它在语言理解、生成、多语言能力、编码、数学和推理等多个基准测试中表现出色，超越了大多数开源模型。与 Qwen1.5-1.8B-Chat 相比，Qwen2-1.5B-Instruct 在 MMLU、HumanEval、GSM8K、C-Eval 和 IFEval 等测试中均显示出显著的性能提升，尽管参数量略少',
-    displayName: 'Qwen2 1.5B Instruct (Free)',
-    id: 'Qwen/Qwen2-1.5B-Instruct', // 将于 2025-06-05 下线
-    pricing: {
-      currency: 'CNY',
-      input: 0,
-      output: 0,
-    },
-    type: 'chat',
-  },
-  {
-    contextWindowTokens: 32_768,
-    description:
-      'Qwen2-1.5B-Instruct 是 Qwen2 系列中的指令微调大语言模型，参数规模为 1.5B。该模型基于 Transformer 架构，采用了 SwiGLU 激活函数、注意力 QKV 偏置和组查询注意力等技术。它在语言理解、生成、多语言能力、编码、数学和推理等多个基准测试中表现出色，超越了大多数开源模型。与 Qwen1.5-1.8B-Chat 相比，Qwen2-1.5B-Instruct 在 MMLU、HumanEval、GSM8K、C-Eval 和 IFEval 等测试中均显示出显著的性能提升，尽管参数量略少',
-    displayName: 'Qwen2 1.5B Instruct (Pro)',
-    id: 'Pro/Qwen/Qwen2-1.5B-Instruct', // 将于 2025-06-05 下线
-    pricing: {
-      currency: 'CNY',
-      input: 0.14,
-      output: 0.14,
-    },
-    type: 'chat',
-  },
   {
     contextWindowTokens: 32_768,
     description:
@@ -701,22 +640,6 @@ const siliconcloudChatModels: AIChatModelCard[] = [
     },
     type: 'chat',
   },
-  {
-    abilities: {
-      vision: true,
-    },
-    contextWindowTokens: 32_768,
-    description:
-      'Qwen2-VL-7B-Instruct 是 Qwen-VL 模型的最新迭代版本，在视觉理解基准测试中达到了最先进的性能，包括 MathVista、DocVQA、RealWorldQA 和 MTVQA 等。Qwen2-VL 能够用于高质量的基于视频的问答、对话和内容创作，还具备复杂推理和决策能力，可以与移动设备、机器人等集成，基于视觉环境和文本指令进行自动操作。除了英语和中文，Qwen2-VL 现在还支持理解图像中不同语言的文本，包括大多数欧洲语言、日语、韩语、阿拉伯语和越南语等',
-    displayName: 'Qwen2 VL 7B Instruct (Pro)',
-    id: 'Pro/Qwen/Qwen2-VL-7B-Instruct', // 将于 2025-06-05 下线
-    pricing: {
-      currency: 'CNY',
-      input: 0.35,
-      output: 0.35,
-    },
-    type: 'chat',
-  },
   {
     abilities: {
       vision: true,
@@ -797,22 +720,6 @@ const siliconcloudChatModels: AIChatModelCard[] = [
     },
     type: 'chat',
   },
-  {
-    abilities: {
-      functionCall: true,
-    },
-    contextWindowTokens: 32_768,
-    description:
-      'InternLM2.5-20B-Chat 是一个开源的大规模对话模型，基于 InternLM2 架构开发。该模型拥有 200 亿参数，在数学推理方面表现出色，超越了同量级的 Llama3 和 Gemma2-27B 模型。InternLM2.5-20B-Chat 在工具调用能力方面有显著提升，支持从上百个网页收集信息进行分析推理，并具备更强的指令理解、工具选择和结果反思能力。它适用于构建复杂智能体，可进行多轮工具调用以完成复杂任务',
-    displayName: 'InternLM2.5 20B Chat',
-    id: 'internlm/internlm2_5-20b-chat', // 将于 2025-06-05 下线
-    pricing: {
-      currency: 'CNY',
-      input: 1,
-      output: 1,
-    },
-    type: 'chat',
-  },
   {
     abilities: {
       functionCall: true,
@@ -845,19 +752,6 @@ const siliconcloudChatModels: AIChatModelCard[] = [
     },
     type: 'chat',
   },
-  {
-    contextWindowTokens: 32_768,
-    description:
-      'ChatGLM3-6B 是 ChatGLM 系列的开源模型，由智谱 AI 开发。该模型保留了前代模型的优秀特性，如对话流畅和部署门槛低，同时引入了新的特性。它采用了更多样的训练数据、更充分的训练步数和更合理的训练策略，在 10B 以下的预训练模型中表现出色。ChatGLM3-6B 支持多轮对话、工具调用、代码执行和 Agent 任务等复杂场景。除对话模型外，还开源了基础模型 ChatGLM-6B-Base 和长文本对话模型 ChatGLM3-6B-32K。该模型对学术研究完全开放，在登记后也允许免费商业使用',
-    displayName: 'ChatGLM3 6B (Free)',
-    id: 'THUDM/chatglm3-6b', // 将于 2025-06-05 下线
-    pricing: {
-      currency: 'CNY',
-      input: 0,
-      output: 0,
-    },
-    type: 'chat',
-  },
 ];
 
 export const allModels = [...siliconcloudChatModels];

package/src/features/AgentSetting/AgentModal/index.tsx

@@ -1,7 +1,7 @@
 'use client';
 
 import { Form, type FormGroupItemType, SliderWithInput } from '@lobehub/ui';
-import { Switch } from 'antd';
+import { Form as AntdForm, Switch } from 'antd';
 import isEqual from 'fast-deep-equal';
 import { memo } from 'react';
 import { useTranslation } from 'react-i18next';
@@ -15,6 +15,7 @@ import { selectors, useStore } from '../store';
 const AgentModal = memo(() => {
   const { t } = useTranslation('setting');
   const [form] = Form.useForm();
+  const enableMaxTokens = AntdForm.useWatch(['chatConfig', 'enableMaxTokens'], form);
   const config = useStore(selectors.currentAgentConfig, isEqual);
 
   const updateConfig = useStore((s) => s.setAgentConfig);
@@ -69,7 +70,7 @@ const AgentModal = memo(() => {
         children: <SliderWithInput max={32_000} min={0} step={100} unlimitedInput={true} />,
         desc: t('settingModel.maxTokens.desc'),
         divider: false,
-        hidden: !config.chatConfig.enableMaxTokens,
+        hidden: !enableMaxTokens,
         label: t('settingModel.maxTokens.title'),
         name: ['params', 'max_tokens'],
         tag: 'max_tokens',

package/src/features/ChatInput/ActionBar/Search/Controls.tsx

@@ -9,7 +9,7 @@ import { Center, Flexbox } from 'react-layout-kit';
 
 import { useAgentStore } from '@/store/agent';
 import { agentChatConfigSelectors, agentSelectors } from '@/store/agent/slices/chat';
-import { aiModelSelectors, useAiInfraStore } from '@/store/aiInfra';
+import { aiModelSelectors, aiProviderSelectors, useAiInfraStore } from '@/store/aiInfra';
 import { SearchMode } from '@/types/search';
 
 import FCSearchModel from './FCSearchModel';
@@ -99,6 +99,9 @@ const Controls = memo(() => {
   ]);
 
   const supportFC = useAiInfraStore(aiModelSelectors.isModelSupportToolUse(model, provider));
+  const isProviderHasBuiltinSearchConfig = useAiInfraStore(
+    aiProviderSelectors.isProviderHasBuiltinSearchConfig(provider),
+  );
   const isModelHasBuiltinSearchConfig = useAiInfraStore(
     aiModelSelectors.isModelHasBuiltinSearchConfig(model, provider),
   );
@@ -119,6 +122,7 @@ const Controls = memo(() => {
   ];
 
   const showDivider = isModelHasBuiltinSearchConfig || !supportFC;
+  const showModelBuiltinSearch = isModelHasBuiltinSearchConfig || isProviderHasBuiltinSearchConfig;
 
   return (
     <Flexbox gap={4}>
@@ -126,7 +130,7 @@ const Controls = memo(() => {
         <Item {...option} key={option.value} />
       ))}
       {showDivider && <Divider style={{ margin: 0 }} />}
-      {isModelHasBuiltinSearchConfig && <ModelBuiltinSearch />}
+      {showModelBuiltinSearch && <ModelBuiltinSearch />}
       {!supportFC && <FCSearchModel />}
     </Flexbox>
   );

package/src/libs/model-runtime/utils/streams/vertex-ai.ts

@@ -48,6 +48,18 @@ const transformVertexAIStream = (
     );
   }
 
+  if (
+    candidate && // 首先检查是否为 reasoning 内容 (thought: true)
+    Array.isArray(candidate.content.parts) &&
+    candidate.content.parts.length > 0
+  ) {
+    for (const part of candidate.content.parts) {
+      if (part && part.text && (part as any).thought === true) {
+        return { data: part.text, id: context.id, type: 'reasoning' };
+      }
+    }
+  }
+
   const candidates = chunk.candidates;
   if (!candidates)
     return {

package/src/locales/default/setting.ts

@@ -224,7 +224,7 @@ export default {
     },
     inputTemplate: {
       desc: '用户最新的一条消息会填充到此模板',
-      placeholder: '预处理模版 {{text}} 将替换为实时输入信息',
+      placeholder: '预处理模版 {{input_template}} 将替换为实时输入信息',
       title: '用户输入预处理',
     },
     submit: '更新聊天偏好',

package/src/services/chat.ts

@@ -41,6 +41,7 @@ import { createErrorResponse } from '@/utils/errorResponse';
 import { FetchSSEOptions, fetchSSE, getMessageError } from '@/utils/fetch';
 import { genToolCallingName } from '@/utils/toolCall';
 import { createTraceHeader, getTraceId } from '@/utils/trace';
+import { parsePlaceholderVariablesMessages } from '@/utils/client/parserPlaceholder';
 
 import { createHeaderWithAuth, createPayloadWithKeyVaults } from './_auth';
 import { API_ENDPOINTS } from './_url';
@@ -172,14 +173,18 @@ class ChatService {
 
     // =================== 0. process search =================== //
     const chatConfig = agentChatConfigSelectors.currentChatConfig(getAgentStoreState());
-
+    const aiInfraStoreState = getAiInfraStoreState();
     const enabledSearch = chatConfig.searchMode !== 'off';
+    const isProviderHasBuiltinSearch = aiProviderSelectors.isProviderHasBuiltinSearch(
+      payload.provider!,
+    )(aiInfraStoreState);
     const isModelHasBuiltinSearch = aiModelSelectors.isModelHasBuiltinSearch(
       payload.model,
       payload.provider!,
-    )(getAiInfraStoreState());
+    )(aiInfraStoreState);
 
-    const useModelSearch = isModelHasBuiltinSearch && chatConfig.useModelBuiltinSearch;
+    const useModelSearch =
+      (isProviderHasBuiltinSearch || isModelHasBuiltinSearch) && chatConfig.useModelBuiltinSearch;
 
     const useApplicationBuiltinSearchTool = enabledSearch && !useModelSearch;
 
@@ -189,11 +194,14 @@ class ChatService {
       pluginIds.push(WebBrowsingManifest.identifier);
     }
 
-    // ============  1. preprocess messages   ============ //
+    // ============  1. preprocess placeholder variables   ============ //
+    const parsedMessages = parsePlaceholderVariablesMessages(messages);
+
+    // ============  2. preprocess messages   ============ //
 
     const oaiMessages = this.processMessages(
       {
-        messages,
+        messages: parsedMessages,
         model: payload.model,
         provider: payload.provider!,
         tools: pluginIds,
@@ -201,28 +209,28 @@ class ChatService {
       options,
     );
 
-    // ============  2. preprocess tools   ============ //
+    // ============  3. preprocess tools   ============ //
 
     const tools = this.prepareTools(pluginIds, {
       model: payload.model,
       provider: payload.provider!,
     });
 
-    // ============  3. process extend params   ============ //
+    // ============  4. process extend params   ============ //
 
     let extendParams: Record<string, any> = {};
 
     const isModelHasExtendParams = aiModelSelectors.isModelHasExtendParams(
       payload.model,
       payload.provider!,
-    )(getAiInfraStoreState());
+    )(aiInfraStoreState);
 
     // model
     if (isModelHasExtendParams) {
       const modelExtendParams = aiModelSelectors.modelExtendParams(
         payload.model,
         payload.provider!,
-      )(getAiInfraStoreState());
+      )(aiInfraStoreState);
       // if model has extended params, then we need to check if the model can use reasoning
 
       if (modelExtendParams!.includes('enableReasoning')) {

package/src/store/chat/slices/aiChat/actions/generateAIChat.ts

@@ -1,7 +1,6 @@
 /* eslint-disable sort-keys-fix/sort-keys-fix, typescript-sort-keys/interface */
 // Disable the auto sort key eslint rule to make the code more logic and readable
 import { produce } from 'immer';
-import { template } from 'lodash-es';
 import { StateCreator } from 'zustand/vanilla';
 
 import { LOADING_FLAT, MESSAGE_CANCEL_FLAT } from '@/const/message';
@@ -13,7 +12,7 @@ import { messageService } from '@/services/message';
 import { useAgentStore } from '@/store/agent';
 import { agentChatConfigSelectors, agentSelectors } from '@/store/agent/selectors';
 import { getAgentStoreState } from '@/store/agent/store';
-import { aiModelSelectors } from '@/store/aiInfra';
+import { aiModelSelectors, aiProviderSelectors } from '@/store/aiInfra';
 import { getAiInfraStoreState } from '@/store/aiInfra/store';
 import { chatHelpers } from '@/store/chat/helpers';
 import { ChatStore } from '@/store/chat/store';
@@ -299,7 +298,8 @@ export const generateAIChat: StateCreator<
     // create a new array to avoid the original messages array change
     const messages = [...originalMessages];
 
-    const { model, provider, chatConfig } = agentSelectors.currentAgentConfig(getAgentStoreState());
+    const agentStoreState = getAgentStoreState();
+    const { model, provider, chatConfig } = agentSelectors.currentAgentConfig(agentStoreState);
 
     let fileChunks: MessageSemanticSearchChunk[] | undefined;
     let ragQueryId;
@@ -323,7 +323,7 @@ export const generateAIChat: StateCreator<
         chunks,
         userQuery: lastMsg.content,
         rewriteQuery,
-        knowledge: agentSelectors.currentEnabledKnowledge(getAgentStoreState()),
+        knowledge: agentSelectors.currentEnabledKnowledge(agentStoreState),
       });
 
       // 3. add the retrieve context messages to the messages history
@@ -355,14 +355,25 @@ export const generateAIChat: StateCreator<
     if (!assistantId) return;
 
     // 3. place a search with the search working model if this model is not support tool use
+    const aiInfraStoreState = getAiInfraStoreState();
     const isModelSupportToolUse = aiModelSelectors.isModelSupportToolUse(
       model,
       provider!,
-    )(getAiInfraStoreState());
-    const isAgentEnableSearch = agentChatConfigSelectors.isAgentEnableSearch(getAgentStoreState());
+    )(aiInfraStoreState);
+    const isProviderHasBuiltinSearch = aiProviderSelectors.isProviderHasBuiltinSearch(provider!)(
+      aiInfraStoreState,
+    );
+    const isModelHasBuiltinSearch = aiModelSelectors.isModelHasBuiltinSearch(
+      model,
+      provider!,
+    )(aiInfraStoreState);
+    const useModelBuiltinSearch = agentChatConfigSelectors.useModelBuiltinSearch(agentStoreState);
+    const useModelSearch =
+      (isProviderHasBuiltinSearch || isModelHasBuiltinSearch) && useModelBuiltinSearch;
+    const isAgentEnableSearch = agentChatConfigSelectors.isAgentEnableSearch(agentStoreState);
 
-    if (isAgentEnableSearch && !isModelSupportToolUse) {
-      const { model, provider } = agentChatConfigSelectors.searchFCModel(getAgentStoreState());
+    if (isAgentEnableSearch && !useModelSearch && !isModelSupportToolUse) {
+      const { model, provider } = agentChatConfigSelectors.searchFCModel(agentStoreState);
 
       let isToolsCalling = false;
       let isError = false;
@@ -460,10 +471,10 @@ export const generateAIChat: StateCreator<
     }
 
     // 6. summary history if context messages is larger than historyCount
-    const historyCount = agentChatConfigSelectors.historyCount(getAgentStoreState());
+    const historyCount = agentChatConfigSelectors.historyCount(agentStoreState);
 
     if (
-      agentChatConfigSelectors.enableHistoryCount(getAgentStoreState()) &&
+      agentChatConfigSelectors.enableHistoryCount(agentStoreState) &&
       chatConfig.enableCompressHistory &&
       originalMessages.length > historyCount
     ) {
@@ -495,8 +506,6 @@ export const generateAIChat: StateCreator<
     const agentConfig = agentSelectors.currentAgentConfig(getAgentStoreState());
     const chatConfig = agentChatConfigSelectors.currentChatConfig(getAgentStoreState());
 
-    const compiler = template(chatConfig.inputTemplate, { interpolate: /{{([\S\s]+?)}}/g });
-
     // ================================== //
     //   messages uniformly preprocess    //
     // ================================== //
@@ -511,29 +520,12 @@ export const generateAIChat: StateCreator<
       historyCount,
     });
 
-    // 2. replace inputMessage template
-    preprocessMsgs = !chatConfig.inputTemplate
-      ? preprocessMsgs
-      : preprocessMsgs.map((m) => {
-          if (m.role === 'user') {
-            try {
-              return { ...m, content: compiler({ text: m.content }) };
-            } catch (error) {
-              console.error(error);
-
-              return m;
-            }
-          }
-
-          return m;
-        });
-
-    // 3. add systemRole
+    // 2. add systemRole
     if (agentConfig.systemRole) {
       preprocessMsgs.unshift({ content: agentConfig.systemRole, role: 'system' } as ChatMessage);
     }
 
-    // 4. handle max_tokens
+    // 3. handle max_tokens
     agentConfig.params.max_tokens = chatConfig.enableMaxTokens
       ? agentConfig.params.max_tokens
       : undefined;

package/src/store/user/slices/auth/selectors.test.ts

@@ -33,6 +33,100 @@ afterEach(() => {
 });
 
 describe('userProfileSelectors', () => {
+  describe('displayUserName', () => {
+    it('should return default username when auth is disabled and not desktop', () => {
+      enableAuth = false;
+      isDesktop = false;
+
+      const store: UserStore = {
+        isSignedIn: false,
+        user: null,
+        enableAuth: () => false,
+      } as unknown as UserStore;
+
+      expect(userProfileSelectors.displayUserName(store)).toBe('LobeChat');
+    });
+
+    it('should return user username when auth is disabled and is desktop', () => {
+      enableAuth = false;
+      isDesktop = true;
+
+      const store: UserStore = {
+        isSignedIn: false,
+        user: { username: 'johndoe' },
+        enableAuth: () => false,
+      } as unknown as UserStore;
+
+      expect(userProfileSelectors.displayUserName(store)).toBe('johndoe');
+    });
+
+    it('should return user username when signed in', () => {
+      const store: UserStore = {
+        isSignedIn: true,
+        user: { username: 'johndoe' },
+        enableAuth: () => true,
+      } as UserStore;
+
+      expect(userProfileSelectors.displayUserName(store)).toBe('johndoe');
+    });
+
+    it('should return email when signed in but username is not existed in UserStore', () => {
+      const store: UserStore = {
+        isSignedIn: true,
+        user: { email: 'demo@lobehub.com' },
+        enableAuth: () => true,
+      } as UserStore;
+
+      expect(userProfileSelectors.displayUserName(store)).toBe('demo@lobehub.com');
+    });
+
+    it('should return "anonymous" when not signed in', () => {
+      const store: UserStore = {
+        enableAuth: () => true,
+        isSignedIn: false,
+        user: null,
+      } as unknown as UserStore;
+
+      expect(userProfileSelectors.displayUserName(store)).toBe('anonymous');
+    });
+  });
+
+  describe('email', () => {
+    it('should return user email if exist', () => {
+      const store: UserStore = {
+        user: { email: 'demo@lobehub.com' },
+      } as UserStore;
+
+      expect(userProfileSelectors.email(store)).toBe('demo@lobehub.com');
+    });
+
+    it('should return empty string if not exist', () => {
+      const store: UserStore = {
+        user: { email: undefined },
+      } as UserStore;
+
+      expect(userProfileSelectors.email(store)).toBe('');
+    });
+  });
+
+  describe('fullName', () => {
+    it('should return user fullName if exist', () => {
+      const store: UserStore = {
+        user: { fullName: 'John Doe' },
+      } as UserStore;
+
+      expect(userProfileSelectors.fullName(store)).toBe('John Doe');
+    });
+
+    it('should return empty string if not exist', () => {
+      const store: UserStore = {
+        user: { fullName: undefined },
+      } as UserStore;
+
+      expect(userProfileSelectors.fullName(store)).toBe('');
+    });
+  });
+
   describe('nickName', () => {
     it('should return default nickname when auth is disabled and not desktop', () => {
       enableAuth = false;

package/src/store/user/slices/auth/selectors.ts

@@ -36,6 +36,9 @@ const username = (s: UserStore) => {
 };
 
 export const userProfileSelectors = {
+  displayUserName: (s: UserStore): string => username(s) || s.user?.email || '',
+  email: (s: UserStore): string => s.user?.email || '',
+  fullName: (s: UserStore): string => s.user?.fullName || '',
   nickName,
   userAvatar: (s: UserStore): string => s.user?.avatar || '',
   userId: (s: UserStore) => s.user?.id,