@lobehub/chat 1.84.19 → 1.84.20

package/CHANGELOG.md

@@ -2,6 +2,31 @@
 
 # Changelog
 
+### [Version 1.84.20](https://github.com/lobehub/lobe-chat/compare/v1.84.19...v1.84.20)
+
+<sup>Released on **2025-05-04**</sup>
+
+#### 💄 Styles
+
+- **misc**: Show Aliyun Bailian tokens usage tracking.
+
+<br/>
+
+<details>
+<summary><kbd>Improvements and Fixes</kbd></summary>
+
+#### Styles
+
+- **misc**: Show Aliyun Bailian tokens usage tracking, closes [#7660](https://github.com/lobehub/lobe-chat/issues/7660) ([3ef0542](https://github.com/lobehub/lobe-chat/commit/3ef0542))
+
+</details>
+
+<div align="right">
+
+[![](https://img.shields.io/badge/-BACK_TO_TOP-151515?style=flat-square)](#readme-top)
+
+</div>
+
 ### [Version 1.84.19](https://github.com/lobehub/lobe-chat/compare/v1.84.18...v1.84.19)
 
 <sup>Released on **2025-05-04**</sup>

package/changelog/v1.json

@@ -1,4 +1,13 @@
 [
+  {
+    "children": {
+      "improvements": [
+        "Show Aliyun Bailian tokens usage tracking."
+      ]
+    },
+    "date": "2025-05-04",
+    "version": "1.84.20"
+  },
   {
     "children": {
       "improvements": [

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@lobehub/chat",
-  "version": "1.84.19",
+  "version": "1.84.20",
   "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/libs/agent-runtime/qwen/index.ts

@@ -24,16 +24,19 @@ export const LobeQwenAI = LobeOpenAICompatibleFactory({
   baseURL: 'https://dashscope.aliyuncs.com/compatible-mode/v1',
   chatCompletion: {
     handlePayload: (payload) => {
-      const { model, presence_penalty, temperature, thinking, top_p, enabledSearch, ...rest } = payload;
+      const { model, presence_penalty, temperature, thinking, top_p, enabledSearch, ...rest } =
+        payload;
 
       return {
         ...rest,
-        ...( ['qwen3','qwen-turbo','qwen-plus']
-          .some(keyword => model.toLowerCase().includes(keyword))
-          ? { 
-            enable_thinking: thinking !== undefined ? thinking.type === 'enabled' : false,
-            thinking_budget: thinking?.budget_tokens === 0 ? 0 : (thinking?.budget_tokens || undefined),
-          }
+        ...(['qwen3', 'qwen-turbo', 'qwen-plus'].some((keyword) =>
+          model.toLowerCase().includes(keyword),
+        )
+          ? {
+              enable_thinking: thinking !== undefined ? thinking.type === 'enabled' : false,
+              thinking_budget:
+                thinking?.budget_tokens === 0 ? 0 : thinking?.budget_tokens || undefined,
+            }
           : {}),
         frequency_penalty: undefined,
         model,
@@ -77,7 +80,16 @@ export const LobeQwenAI = LobeOpenAICompatibleFactory({
   models: async ({ client }) => {
     const { LOBE_DEFAULT_MODEL_LIST } = await import('@/config/aiModels');
 
-    const functionCallKeywords = ['qwen-max', 'qwen-plus', 'qwen-turbo', 'qwen-long', 'qwen1.5', 'qwen2', 'qwen2.5', 'qwen3'];
+    const functionCallKeywords = [
+      'qwen-max',
+      'qwen-plus',
+      'qwen-turbo',
+      'qwen-long',
+      'qwen1.5',
+      'qwen2',
+      'qwen2.5',
+      'qwen3',
+    ];
 
     const visionKeywords = ['qvq', 'vl'];
 

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

@@ -1,6 +1,8 @@
+import OpenAI from 'openai';
 import { beforeAll, describe, expect, it, vi } from 'vitest';
 
-import { QwenAIStream } from './qwen';
+import { StreamContext } from './protocol';
+import { QwenAIStream, transformQwenStream } from './qwen';
 
 describe('QwenAIStream', () => {
   beforeAll(() => {});
@@ -349,3 +351,76 @@ describe('QwenAIStream', () => {
     ]);
   });
 });
+
+describe('transformQwenStream', () => {
+  it('should handle usage chunk', () => {
+    const mockChunk: OpenAI.ChatCompletionChunk = {
+      choices: [],
+      id: 'usage-test-id',
+      model: 'qwen-test-model',
+      object: 'chat.completion.chunk',
+      created: Date.now(),
+      usage: {
+        completion_tokens: 50,
+        prompt_tokens: 100,
+        total_tokens: 150,
+        completion_tokens_details: {}, // Ensure these exist even if empty
+        prompt_tokens_details: {},     // Ensure these exist even if empty
+      },
+    };
+
+    const streamContext: StreamContext = { id: '' };
+
+    const result = transformQwenStream(mockChunk, streamContext);
+
+    expect(result).toEqual({
+      id: 'usage-test-id',
+      type: 'usage',
+      data: {
+        inputTextTokens: 100,
+        outputTextTokens: 50,
+        totalInputTokens: 100,
+        totalOutputTokens: 50,
+        totalTokens: 150,
+      },
+    });
+
+    // Verify streamContext is updated
+    expect(streamContext.usage).toEqual({
+      inputTextTokens: 100,
+      outputTextTokens: 50,
+      totalInputTokens: 100,
+      totalOutputTokens: 50,
+      totalTokens: 150,
+    });
+  });
+
+  it('should handle usage chunk without streamContext', () => {
+    const mockChunk: OpenAI.ChatCompletionChunk = {
+      choices: [],
+      id: 'usage-test-id-no-ctx',
+      model: 'qwen-test-model',
+      object: 'chat.completion.chunk',
+      created: Date.now(),
+      usage: {
+        completion_tokens: 55,
+        prompt_tokens: 105,
+        total_tokens: 160,
+      },
+    };
+
+    const result = transformQwenStream(mockChunk); // No streamContext passed
+
+    expect(result).toEqual({
+      id: 'usage-test-id-no-ctx',
+      type: 'usage',
+      data: {
+        inputTextTokens: 105,
+        outputTextTokens: 55,
+        totalInputTokens: 105,
+        totalOutputTokens: 55,
+        totalTokens: 160,
+      },
+    });
+  });
+});

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

@@ -4,17 +4,37 @@ import { ChatCompletionContentPart } from 'openai/resources/index.mjs';
 import type { Stream } from 'openai/streaming';
 
 import { ChatStreamCallbacks } from '../../types';
+import { convertUsage } from '../usageConverter';
 import {
+  StreamContext,
   StreamProtocolChunk,
   StreamProtocolToolCallChunk,
   StreamToolCallChunkData,
   convertIterableToStream,
   createCallbacksTransformer,
   createSSEProtocolTransformer,
+  createTokenSpeedCalculator,
   generateToolCallId,
 } from './protocol';
 
-export const transformQwenStream = (chunk: OpenAI.ChatCompletionChunk): StreamProtocolChunk => {
+export const transformQwenStream = (
+  chunk: OpenAI.ChatCompletionChunk,
+  streamContext?: StreamContext,
+): StreamProtocolChunk | StreamProtocolChunk[] => {
+  if (Array.isArray(chunk.choices) && chunk.choices.length === 0 && chunk.usage) {
+    const usage = convertUsage({
+      ...chunk.usage,
+      completion_tokens_details: chunk.usage.completion_tokens_details || {},
+      prompt_tokens_details: chunk.usage.prompt_tokens_details || {},
+    });
+
+    if (streamContext) {
+      streamContext.usage = usage;
+    }
+
+    return { data: usage, id: chunk.id, type: 'usage' };
+  }
+
   const item = chunk.choices[0];
 
   if (!item) {
@@ -96,10 +116,14 @@ export const QwenAIStream = (
   // eslint-disable-next-line @typescript-eslint/no-unused-vars, unused-imports/no-unused-vars
   { callbacks, inputStartAt }: { callbacks?: ChatStreamCallbacks; inputStartAt?: number } = {},
 ) => {
+  const streamContext: StreamContext = { id: '' };
   const readableStream =
     stream instanceof ReadableStream ? stream : convertIterableToStream(stream);
 
   return readableStream
-    .pipeThrough(createSSEProtocolTransformer(transformQwenStream))
+    .pipeThrough(
+      createTokenSpeedCalculator(transformQwenStream, { inputStartAt, streamStack: streamContext }),
+    )
+    .pipeThrough(createSSEProtocolTransformer((c) => c, streamContext))
     .pipeThrough(createCallbacksTransformer(callbacks));
 };