@lobehub/chat 1.97.15 → 1.97.17

package/src/config/aiModels/zhipu.ts

@@ -1,6 +1,47 @@
 import { AIChatModelCard } from '@/types/aiModel';
 
 const zhipuChatModels: AIChatModelCard[] = [
+  {
+    abilities: {
+      reasoning: true,
+      search: true,
+      vision: true,
+    },
+    contextWindowTokens: 64_000,
+    description: 'GLM-4.1V-Thinking 系列模型是目前已知10B级别的VLM模型中性能最强的视觉模型，融合了同级别SOTA的各项视觉语言任务，包括视频理解、图片问答、学科解题、OCR文字识别、文档和图表解读、GUI Agent、前端网页Coding、Grounding等，多项任务能力甚至超过8倍参数量的Qwen2.5-VL-72B。通过领先的强化学习技术，模型掌握了通过思维链推理的方式提升回答的准确性和丰富度，从最终效果和可解释性等维度都显著超过传统的非thinking模型。',
+    displayName: 'GLM-4.1V-Thinking-FlashX',
+    id: 'glm-4.1v-thinking-flashx',
+    pricing: {
+      currency: 'CNY',
+      input: 2,
+      output: 2,
+    },
+    settings: {
+      searchImpl: 'params',
+    },
+    type: 'chat',
+  },
+  {
+    abilities: {
+      reasoning: true,
+      search: true,
+      vision: true,
+    },
+    contextWindowTokens: 64_000,
+    description: 'GLM-4.1V-Thinking 系列模型是目前已知10B级别的VLM模型中性能最强的视觉模型，融合了同级别SOTA的各项视觉语言任务，包括视频理解、图片问答、学科解题、OCR文字识别、文档和图表解读、GUI Agent、前端网页Coding、Grounding等，多项任务能力甚至超过8倍参数量的Qwen2.5-VL-72B。通过领先的强化学习技术，模型掌握了通过思维链推理的方式提升回答的准确性和丰富度，从最终效果和可解释性等维度都显著超过传统的非thinking模型。',
+    displayName: 'GLM-4.1V-Thinking-Flash',
+    enabled: true,
+    id: 'glm-4.1v-thinking-flash',
+    pricing: {
+      currency: 'CNY',
+      input: 0,
+      output: 0,
+    },
+    settings: {
+      searchImpl: 'params',
+    },
+    type: 'chat',
+  },
   {
     abilities: {
       reasoning: true,
@@ -21,11 +62,11 @@ const zhipuChatModels: AIChatModelCard[] = [
       reasoning: true,
       search: true,
     },
-    contextWindowTokens: 32_000,
+    contextWindowTokens: 128_000,
     description: '推理模型: 具备强大推理能力，适用于需要深度推理的任务。',
     displayName: 'GLM-Z1-Air',
     id: 'glm-z1-air',
-    maxOutput: 30_000,
+    maxOutput: 32_000,
     pricing: {
       currency: 'CNY',
       input: 0.5,
@@ -61,13 +102,34 @@ const zhipuChatModels: AIChatModelCard[] = [
       reasoning: true,
       search: true,
     },
-    contextWindowTokens: 32_000,
+    contextWindowTokens: 128_000,
+    description:
+      '高速低价：Flash增强版本，超快推理速度，更快并发保障。',
+    displayName: 'GLM-Z1-FlashX',
+    id: 'glm-z1-flashx',
+    maxOutput: 32_000,
+    pricing: {
+      currency: 'CNY',
+      input: 0.1,
+      output: 0.1,
+    },
+    settings: {
+      searchImpl: 'params',
+    },
+    type: 'chat',
+  },
+  {
+    abilities: {
+      reasoning: true,
+      search: true,
+    },
+    contextWindowTokens: 128_000,
     description:
       'GLM-Z1 系列具备强大的复杂推理能力，在逻辑推理、数学、编程等领域表现优异。最大上下文长度为32K。',
     displayName: 'GLM-Z1-Flash',
     enabled: true,
     id: 'glm-z1-flash',
-    maxOutput: 30_000,
+    maxOutput: 32_000,
     pricing: {
       currency: 'CNY',
       input: 0,
@@ -88,7 +150,7 @@ const zhipuChatModels: AIChatModelCard[] = [
     displayName: 'GLM-4-Flash-250414',
     enabled: true,
     id: 'glm-4-flash-250414',
-    maxOutput: 4000,
+    maxOutput: 16_000,
     pricing: {
       currency: 'CNY',
       input: 0,
@@ -106,9 +168,9 @@ const zhipuChatModels: AIChatModelCard[] = [
     },
     contextWindowTokens: 128_000,
     description: 'GLM-4-FlashX 是Flash的增强版本，超快推理速度。',
-    displayName: 'GLM-4-FlashX',
+    displayName: 'GLM-4-FlashX-250414',
     id: 'glm-4-flashx',
-    maxOutput: 4000,
+    maxOutput: 16_000,
     pricing: {
       currency: 'CNY',
       input: 0.1,
@@ -144,11 +206,11 @@ const zhipuChatModels: AIChatModelCard[] = [
       functionCall: true,
       search: true,
     },
-    contextWindowTokens: 32_000,
+    contextWindowTokens: 128_000,
     description: 'GLM-4-Air 是性价比高的版本，性能接近GLM-4，提供快速度和实惠的价格。',
     displayName: 'GLM-4-Air-250414',
     id: 'glm-4-air-250414',
-    maxOutput: 4000,
+    maxOutput: 16_000,
     pricing: {
       currency: 'CNY',
       input: 0.5,
@@ -211,8 +273,8 @@ const zhipuChatModels: AIChatModelCard[] = [
     maxOutput: 4000,
     pricing: {
       currency: 'CNY',
-      input: 50,
-      output: 50,
+      input: 5,
+      output: 5,
     },
     settings: {
       searchImpl: 'params',
@@ -279,7 +341,7 @@ const zhipuChatModels: AIChatModelCard[] = [
     abilities: {
       vision: true,
     },
-    contextWindowTokens: 8192,
+    contextWindowTokens: 16_000,
     description: 'GLM-4V-Plus 具备对视频内容及多图片的理解能力，适合多模态任务。',
     displayName: 'GLM-4V-Plus-0111',
     id: 'glm-4v-plus-0111',

package/src/config/modelProviders/zhipu.ts

@@ -9,7 +9,6 @@ const ZhiPu: ModelProviderCard = {
       contextWindowTokens: 16_384,
       description: 'GLM-Zero-Preview具备强大的复杂推理能力，在逻辑推理、数学、编程等领域表现优异。',
       displayName: 'GLM-Zero-Preview',
-      enabled: true,
       id: 'glm-zero-preview',
       pricing: {
         currency: 'CNY',
@@ -207,7 +206,7 @@ const ZhiPu: ModelProviderCard = {
       },
     },
   ],
-  checkModel: 'glm-4-flash',
+  checkModel: 'glm-4-flash-250414',
   description:
     '智谱 AI 提供多模态与语言模型的开放平台，支持广泛的AI应用场景，包括文本处理、图像理解与编程辅助等。',
   id: 'zhipu',

package/src/libs/model-runtime/hunyuan/index.ts

@@ -12,11 +12,12 @@ export const LobeHunyuanAI = createOpenAICompatibleRuntime({
   chatCompletion: {
     handlePayload: (payload) => {
       // eslint-disable-next-line unused-imports/no-unused-vars, @typescript-eslint/no-unused-vars
-      const { enabledSearch, frequency_penalty, presence_penalty, ...rest } = payload;
+      const { enabledSearch, frequency_penalty, model, presence_penalty, thinking, ...rest } = payload;
 
       return {
         ...rest,
         frequency_penalty: undefined,
+        model,
         presence_penalty: undefined,
         stream: true,
         ...(enabledSearch && {
@@ -28,6 +29,13 @@ export const LobeHunyuanAI = createOpenAICompatibleRuntime({
           enable_speed_search: process.env.HUNYUAN_ENABLE_SPEED_SEARCH === '1',
           search_info: true,
         }),
+        ...(model === 'hunyuan-a13b' && {
+          enable_thinking: thinking?.type === 'enabled' 
+            ? true 
+            : thinking?.type === 'disabled' 
+              ? false 
+              : undefined
+        }),
       } as any;
     },
   },