@lobehub/chat 1.19.32 → 1.19.33

package/CHANGELOG.md

@@ -2,6 +2,39 @@
 
 # Changelog
 
+### [Version 1.19.33](https://github.com/lobehub/lobe-chat/compare/v1.19.32...v1.19.33)
+
+<sup>Released on **2024-09-25**</sup>
+
+#### 🐛 Bug Fixes
+
+- **misc**: MiniMax output long content interrupted by non-existent error.
+
+#### 💄 Styles
+
+- **misc**: Update google provider model info.
+
+<br/>
+
+<details>
+<summary><kbd>Improvements and Fixes</kbd></summary>
+
+#### What's fixed
+
+- **misc**: MiniMax output long content interrupted by non-existent error, closes [#4088](https://github.com/lobehub/lobe-chat/issues/4088) ([4f6e20d](https://github.com/lobehub/lobe-chat/commit/4f6e20d))
+
+#### Styles
+
+- **misc**: Update google provider model info, closes [#4129](https://github.com/lobehub/lobe-chat/issues/4129) ([b1442b9](https://github.com/lobehub/lobe-chat/commit/b1442b9))
+
+</details>
+
+<div align="right">
+
+[![](https://img.shields.io/badge/-BACK_TO_TOP-151515?style=flat-square)](#readme-top)
+
+</div>
+
 ### [Version 1.19.32](https://github.com/lobehub/lobe-chat/compare/v1.19.31...v1.19.32)
 
 <sup>Released on **2024-09-25**</sup>

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@lobehub/chat",
-  "version": "1.19.32",
+  "version": "1.19.33",
   "description": "Lobe Chat - an open-source, high-performance chatbot framework that supports speech synthesis, multimodal, and extensible Function Call plugin system. Supports one-click free deployment of your private ChatGPT/LLM web application.",
   "keywords": [
     "framework",

package/src/config/modelProviders/google.ts

@@ -16,78 +16,69 @@ const Google: ModelProviderCard = {
         input: 0.075,
         output: 0.3,
       },
-      tokens: 1_048_576 + 8192,
+      tokens: 1_000_000 + 8192,
       vision: true,
     },
     {
-      description: 'Gemini 1.5 Flash 0827 提供了优化后的多模态处理能力，适用多种复杂任务场景。',
-      displayName: 'Gemini 1.5 Flash 0827',
+      description: 'Gemini 1.5 Flash 002 是一款高效的多模态模型，支持广泛应用的扩展。',
+      displayName: 'Gemini 1.5 Flash 002',
+      enabled: true,
       functionCall: true,
-      id: 'gemini-1.5-flash-exp-0827',
+      id: 'gemini-1.5-flash-002',
       maxOutput: 8192,
       pricing: {
         cachedInput: 0.018_75,
         input: 0.075,
         output: 0.3,
       },
-      releasedAt: '2024-08-27',
-      tokens: 1_048_576 + 8192,
-      vision: true,
-    },
-    {
-      description: 'Gemini 1.5 Flash 8B 0827 专为处理大规模任务场景设计，提供无与伦比的处理速度。',
-      displayName: 'Gemini 1.5 Flash 8B 0827',
-      functionCall: true,
-      id: 'gemini-1.5-flash-8b-exp-0827',
-      maxOutput: 8192,
-      releasedAt: '2024-08-27',
-      tokens: 1_048_576 + 8192,
+      releasedAt: '2024-09-25',
+      tokens: 1_000_000 + 8192,
       vision: true,
     },
     {
-      description:
-        'Gemini 1.5 Flash 8B 0924 是最新的实验性模型，在文本和多模态用例中都有显著的性能提升。',
-      displayName: 'Gemini 1.5 Flash 8B 0924',
-      enabled: true,
+      description: 'Gemini 1.5 Flash 001 是一款高效的多模态模型，支持广泛应用的扩展。',
+      displayName: 'Gemini 1.5 Flash 001',
       functionCall: true,
-      id: 'gemini-1.5-flash-8b-exp-0924',
+      id: 'gemini-1.5-flash-001',
       maxOutput: 8192,
       pricing: {
         cachedInput: 0.018_75,
         input: 0.075,
         output: 0.3,
       },
-      releasedAt: '2024-09-24',
-      tokens: 1_048_576 + 8192,
+      tokens: 1_000_000 + 8192,
       vision: true,
     },
     {
-      description: 'Gemini 1.5 Flash 001 是一款高效的多模态模型，支持广泛应用的扩展。',
-      displayName: 'Gemini 1.5 Flash 001',
+      description: 'Gemini 1.5 Flash 0827 提供了优化后的多模态处理能力，适用多种复杂任务场景。',
+      displayName: 'Gemini 1.5 Flash 0827',
       functionCall: true,
-      id: 'gemini-1.5-flash-001',
+      id: 'gemini-1.5-flash-exp-0827',
       maxOutput: 8192,
       pricing: {
         cachedInput: 0.018_75,
         input: 0.075,
         output: 0.3,
       },
-      tokens: 1_048_576 + 8192,
+      releasedAt: '2024-08-27',
+      tokens: 1_000_000 + 8192,
       vision: true,
     },
+
     {
-      description: 'Gemini 1.5 Flash 002 是一款高效的多模态模型，支持广泛应用的扩展。',
-      displayName: 'Gemini 1.5 Flash 002',
+      description:
+        'Gemini 1.5 Flash 8B 0924 是最新的实验性模型，在文本和多模态用例中都有显著的性能提升。',
+      displayName: 'Gemini 1.5 Flash 8B 0924',
       functionCall: true,
-      id: 'gemini-1.5-flash-002',
+      id: 'gemini-1.5-flash-8b-exp-0924',
       maxOutput: 8192,
       pricing: {
         cachedInput: 0.018_75,
         input: 0.075,
         output: 0.3,
       },
-      releasedAt: '2024-09-25',
-      tokens: 1_048_576 + 8192,
+      releasedAt: '2024-09-24',
+      tokens: 1_000_000 + 8192,
       vision: true,
     },
     {
@@ -104,69 +95,69 @@ const Google: ModelProviderCard = {
         output: 10.5,
       },
       releasedAt: '2024-02-15',
-      tokens: 2_097_152 + 8192,
+      tokens: 2_000_000 + 8192,
       vision: true,
     },
     {
-      description: 'Gemini 1.5 Pro 0827 结合最新优化技术，带来更高效的多模态数据处理能力。',
-      displayName: 'Gemini 1.5 Pro 0827',
+      description:
+        'Gemini 1.5 Pro 002 是最新的生产就绪模型，提供更高质量的输出，特别在数学、长上下文和视觉任务方面有显著提升。',
+      displayName: 'Gemini 1.5 Pro 002',
+      enabled: true,
       functionCall: true,
-      id: 'gemini-1.5-pro-exp-0827',
+      id: 'gemini-1.5-pro-002',
       maxOutput: 8192,
       pricing: {
-        cachedInput: 0.875,
-        input: 3.5,
-        output: 10.5,
+        cachedInput: 0.315,
+        input: 1.25,
+        output: 2.5,
       },
-      releasedAt: '2024-08-27',
-      tokens: 2_097_152 + 8192,
+      releasedAt: '2024-09-24',
+      tokens: 2_000_000 + 8192,
       vision: true,
     },
     {
-      description: 'Gemini 1.5 Pro 0801 提供出色的多模态处理能力，为应用开发带来更大灵活性。',
-      displayName: 'Gemini 1.5 Pro 0801',
+      description: 'Gemini 1.5 Pro 001 是可扩展的多模态AI解决方案，支持广泛的复杂任务。',
+      displayName: 'Gemini 1.5 Pro 001',
       functionCall: true,
-      id: 'gemini-1.5-pro-exp-0801',
+      id: 'gemini-1.5-pro-001',
       maxOutput: 8192,
       pricing: {
         cachedInput: 0.875,
         input: 3.5,
         output: 10.5,
       },
-      releasedAt: '2024-08-01',
-      tokens: 2_097_152 + 8192,
+      releasedAt: '2024-02-15',
+      tokens: 2_000_000 + 8192,
       vision: true,
     },
     {
-      description: 'Gemini 1.5 Pro 001 是可扩展的多模态AI解决方案，支持广泛的复杂任务。',
-      displayName: 'Gemini 1.5 Pro 001',
+      description: 'Gemini 1.5 Pro 0827 结合最新优化技术，带来更高效的多模态数据处理能力。',
+      displayName: 'Gemini 1.5 Pro 0827',
       functionCall: true,
-      id: 'gemini-1.5-pro-001',
+      id: 'gemini-1.5-pro-exp-0827',
       maxOutput: 8192,
       pricing: {
         cachedInput: 0.875,
         input: 3.5,
         output: 10.5,
       },
-      releasedAt: '2024-02-15',
-      tokens: 2_097_152 + 8192,
+      releasedAt: '2024-08-27',
+      tokens: 2_000_000 + 8192,
       vision: true,
     },
     {
-      description:
-        'Gemini 1.5 Pro 002 是最新的生产就绪模型，提供更高质量的输出，特别在数学、长上下文和视觉任务方面有显著提升。',
-      displayName: 'Gemini 1.5 Pro 002',
-      enabled: true,
+      description: 'Gemini 1.5 Pro 0801 提供出色的多模态处理能力，为应用开发带来更大灵活性。',
+      displayName: 'Gemini 1.5 Pro 0801',
       functionCall: true,
-      id: 'gemini-1.5-pro-002',
+      id: 'gemini-1.5-pro-exp-0801',
       maxOutput: 8192,
       pricing: {
-        cachedInput: 0.315,
-        input: 1.25,
-        output: 2.5,
+        cachedInput: 0.875,
+        input: 3.5,
+        output: 10.5,
       },
-      releasedAt: '2024-09-24',
-      tokens: 2_097_152 + 8192,
+      releasedAt: '2024-08-01',
+      tokens: 2_000_000 + 8192,
       vision: true,
     },
     {
@@ -210,7 +201,7 @@ const Google: ModelProviderCard = {
   ],
   checkModel: 'gemini-1.5-flash-latest',
   description:
-    'Google 的 Gemini 系列是其最先进、通用的 A I模型，由 Google DeepMind 打造，专为多模态设计，支持文本、代码、图像、音频和视频的无缝理解与处理。适用于从数据中心到移动设备的多种环境，极大提升了AI模型的效率与应用广泛性。',
+    'Google 的 Gemini 系列是其最先进、通用的 AI模型，由 Google DeepMind 打造，专为多模态设计，支持文本、代码、图像、音频和视频的无缝理解与处理。适用于从数据中心到移动设备的多种环境，极大提升了AI模型的效率与应用广泛性。',
   id: 'google',
   modelsUrl: 'https://ai.google.dev/gemini-api/docs/models/gemini',
   name: 'Google',

package/src/libs/agent-runtime/minimax/index.test.ts

@@ -253,10 +253,10 @@ describe('LobeMinimaxAI', () => {
         });
       });
 
-      it('should include max tokens when model is abab6.5-chat', () => {
+      it('should include max tokens when model is abab6.5t-chat', () => {
         const payload: ChatStreamPayload = {
           messages: [{ content: 'Hello', role: 'user' }],
-          model: 'abab6.5-chat',
+          model: 'abab6.5t-chat',
           temperature: 0,
           top_p: 0,
         };
@@ -265,9 +265,9 @@ describe('LobeMinimaxAI', () => {
 
         expect(result).toEqual({
           messages: [{ content: 'Hello', role: 'user' }],
-          model: 'abab6.5-chat',
+          model: 'abab6.5t-chat',
           stream: true,
-          max_tokens: 2048,
+          max_tokens: 4096,
         });
       });
     });

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

@@ -127,9 +127,14 @@ export class LobeMinimaxAI implements LobeRuntimeAI {
   // https://www.minimaxi.com/document/guides/chat-model/V2
   private getMaxTokens(model: string): number | undefined {
     switch (model) {
-      case 'abab6.5-chat':
+      case 'abab6.5t-chat':
+      case 'abab6.5g-chat':
+      case 'abab5.5s-chat': 
+      case 'abab5.5-chat':{
+        return 4096;
+      }
       case 'abab6.5s-chat': {
-        return 2048;
+        return 8192;
       }
     }
   }
@@ -139,12 +144,17 @@ export class LobeMinimaxAI implements LobeRuntimeAI {
 
     return {
       ...params,
-      max_tokens: this.getMaxTokens(payload.model),
+      frequency_penalty: undefined,
+      max_tokens: 
+        payload.max_tokens !== undefined 
+        ? payload.max_tokens 
+        : this.getMaxTokens(payload.model),
+      presence_penalty: undefined,
       stream: true,
       temperature: 
-            temperature === undefined || temperature <= 0
-            ? undefined
-            : temperature / 2,
+        temperature === undefined || temperature <= 0
+        ? undefined
+        : temperature / 2,
 
       tools: params.tools?.map((tool) => ({
         function: {

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

@@ -0,0 +1,24 @@
+import { describe, it, expect } from 'vitest';
+import { processDoubleData } from './minimax'; // 假设文件名为 minimax.ts
+
+describe('processDoubleData', () => {
+  it('should remove the second "data: {"id": and everything after it when matchCount is 2', () => {
+    const chunkValue = `data: {"id":"first"} some other text 
+    
+    data: {"id":"second"} more text`;
+    const result = processDoubleData(chunkValue);
+    expect(result).toBe('data: {"id":"first"} some other text');
+  });
+
+  it('should not modify chunkValue when matchCount is not 2', () => {
+    const chunkValue = `data: {"id":"first"} some other text`;
+    const result = processDoubleData(chunkValue);
+    expect(result).toBe(chunkValue);
+  });
+
+  it('should not modify chunkValue when matchCount is more than 2', () => {
+    const chunkValue = `data: {"id":"first"} some other text data: {"id":"second"} more text data: {"id":"third"} even more text`;
+    const result = processDoubleData(chunkValue);
+    expect(result).toBe(chunkValue);
+  });
+});

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

@@ -4,6 +4,19 @@ import { ChatStreamCallbacks } from '../../types';
 import { transformOpenAIStream } from './openai';
 import { createCallbacksTransformer, createSSEProtocolTransformer } from './protocol';
 
+export const processDoubleData = (chunkValue: string): string => {
+  const dataPattern = /data: {"id":"/g;
+  const matchCount = (chunkValue.match(dataPattern) || []).length;
+  let modifiedChunkValue = chunkValue;
+  if (matchCount === 2) {
+    const secondDataIdIndex = chunkValue.indexOf('data: {"id":', chunkValue.indexOf('data: {"id":') + 1);
+    if (secondDataIdIndex !== -1) {
+      modifiedChunkValue = chunkValue.slice(0, secondDataIdIndex).trim();
+    }
+  }
+  return modifiedChunkValue;
+};
+
 const unit8ArrayToJSONChunk = (unit8Array: Uint8Array): OpenAI.ChatCompletionChunk => {
   const decoder = new TextDecoder();
 
@@ -12,6 +25,8 @@ const unit8ArrayToJSONChunk = (unit8Array: Uint8Array): OpenAI.ChatCompletionChu
   // chunkValue example:
   // data: {"id":"028a65377137d57aaceeffddf48ae99f","choices":[{"finish_reason":"tool_calls","index":0,"delta":{"role":"assistant","tool_calls":[{"id":"call_function_7371372822","type":"function","function":{"name":"realtime-weather____fetchCurrentWeather","arguments":"{\"city\": [\"杭州\", \"北京\"]}"}}]}}],"created":155511,"model":"abab6.5s-chat","object":"chat.completion.chunk"}
 
+  chunkValue = processDoubleData(chunkValue);
+
   // so we need to remove `data:` prefix and then parse it as JSON
   if (chunkValue.startsWith('data:')) {
     chunkValue = chunkValue.slice(5).trim();