@lobehub/chat 1.134.7 → 1.135.0

package/CHANGELOG.md

@@ -2,6 +2,31 @@
 
 # Changelog
 
+## [Version 1.135.0](https://github.com/lobehub/lobe-chat/compare/v1.134.7...v1.135.0)
+
+<sup>Released on **2025-10-06**</sup>
+
+#### ✨ Features
+
+- **misc**: Huanyuan text-to-image 3.
+
+<br/>
+
+<details>
+<summary><kbd>Improvements and Fixes</kbd></summary>
+
+#### What's improved
+
+- **misc**: Huanyuan text-to-image 3, closes [#9589](https://github.com/lobehub/lobe-chat/issues/9589) ([1dd0e5e](https://github.com/lobehub/lobe-chat/commit/1dd0e5e))
+
+</details>
+
+<div align="right">
+
+[![](https://img.shields.io/badge/-BACK_TO_TOP-151515?style=flat-square)](#readme-top)
+
+</div>
+
 ### [Version 1.134.7](https://github.com/lobehub/lobe-chat/compare/v1.134.6...v1.134.7)
 
 <sup>Released on **2025-10-06**</sup>

package/changelog/v1.json

@@ -1,4 +1,13 @@
 [
+  {
+    "children": {
+      "features": [
+        "Huanyuan text-to-image 3."
+      ]
+    },
+    "date": "2025-10-06",
+    "version": "1.135.0"
+  },
   {
     "children": {
       "improvements": [

package/locales/ar/models.json

@@ -1253,6 +1253,9 @@
   "fal-ai/flux/schnell": {
     "description": "FLUX.1 [schnell] هو نموذج توليد صور يحتوي على 12 مليار معلمة، يركز على توليد صور عالية الجودة بسرعة."
   },
+  "fal-ai/hunyuan-image/v3": {
+    "description": "نموذج قوي لتوليد الصور متعددة الوسائط الأصلية"
+  },
   "fal-ai/imagen4/preview": {
     "description": "نموذج توليد صور عالي الجودة مقدم من جوجل."
   },

package/locales/bg-BG/models.json

@@ -1253,6 +1253,9 @@
   "fal-ai/flux/schnell": {
     "description": "FLUX.1 [schnell] е модел за генериране на изображения с 12 милиарда параметри, фокусиран върху бързото създаване на висококачествени изображения."
   },
+  "fal-ai/hunyuan-image/v3": {
+    "description": "Мощен оригинален мултимодален модел за генериране на изображения"
+  },
   "fal-ai/imagen4/preview": {
     "description": "Висококачествен модел за генериране на изображения, предоставен от Google."
   },

package/locales/de-DE/models.json

@@ -1253,6 +1253,9 @@
   "fal-ai/flux/schnell": {
     "description": "FLUX.1 [schnell] ist ein bildgenerierendes Modell mit 12 Milliarden Parametern, das sich auf die schnelle Erzeugung hochwertiger Bilder konzentriert."
   },
+  "fal-ai/hunyuan-image/v3": {
+    "description": "Ein leistungsstarkes natives multimodales Bildgenerierungsmodell"
+  },
   "fal-ai/imagen4/preview": {
     "description": "Hochwertiges Bildgenerierungsmodell von Google."
   },

package/locales/en-US/models.json

@@ -1253,6 +1253,9 @@
   "fal-ai/flux/schnell": {
     "description": "FLUX.1 [schnell] is a 12-billion-parameter image generation model focused on fast generation of high-quality images."
   },
+  "fal-ai/hunyuan-image/v3": {
+    "description": "A powerful native multimodal image generation model"
+  },
   "fal-ai/imagen4/preview": {
     "description": "High-quality image generation model provided by Google."
   },

package/locales/es-ES/models.json

@@ -1253,6 +1253,9 @@
   "fal-ai/flux/schnell": {
     "description": "FLUX.1 [schnell] es un modelo generador de imágenes con 12 mil millones de parámetros, enfocado en la generación rápida de imágenes de alta calidad."
   },
+  "fal-ai/hunyuan-image/v3": {
+    "description": "Un potente modelo nativo de generación de imágenes multimodales"
+  },
   "fal-ai/imagen4/preview": {
     "description": "Modelo de generación de imágenes de alta calidad proporcionado por Google."
   },

package/locales/fa-IR/models.json

@@ -1253,6 +1253,9 @@
   "fal-ai/flux/schnell": {
     "description": "FLUX.1 [schnell] یک مدل تولید تصویر با 12 میلیارد پارامتر است که بر تولید سریع تصاویر با کیفیت بالا تمرکز دارد."
   },
+  "fal-ai/hunyuan-image/v3": {
+    "description": "یک مدل قدرتمند بومی تولید تصویر چندوجهی"
+  },
   "fal-ai/imagen4/preview": {
     "description": "مدل تولید تصویر با کیفیت بالا ارائه شده توسط گوگل"
   },

package/locales/fr-FR/models.json

@@ -1253,6 +1253,9 @@
   "fal-ai/flux/schnell": {
     "description": "FLUX.1 [schnell] est un modèle de génération d'images de 12 milliards de paramètres, spécialisé dans la génération rapide d'images de haute qualité."
   },
+  "fal-ai/hunyuan-image/v3": {
+    "description": "Un puissant modèle natif de génération d'images multimodales"
+  },
   "fal-ai/imagen4/preview": {
     "description": "Modèle de génération d'images de haute qualité fourni par Google."
   },

package/locales/it-IT/models.json

@@ -1253,6 +1253,9 @@
   "fal-ai/flux/schnell": {
     "description": "FLUX.1 [schnell] è un modello di generazione immagini con 12 miliardi di parametri, focalizzato sulla generazione rapida di immagini di alta qualità."
   },
+  "fal-ai/hunyuan-image/v3": {
+    "description": "Un potente modello nativo di generazione di immagini multimodali"
+  },
   "fal-ai/imagen4/preview": {
     "description": "Modello di generazione immagini di alta qualità fornito da Google."
   },

package/locales/ja-JP/models.json

@@ -1253,6 +1253,9 @@
   "fal-ai/flux/schnell": {
     "description": "FLUX.1 [schnell] は120億パラメータを持つ画像生成モデルで、高速に高品質な画像生成に特化しています。"
   },
+  "fal-ai/hunyuan-image/v3": {
+    "description": "強力なネイティブマルチモーダル画像生成モデル"
+  },
   "fal-ai/imagen4/preview": {
     "description": "Google 提供の高品質な画像生成モデルです。"
   },

package/locales/ko-KR/models.json

@@ -1253,6 +1253,9 @@
   "fal-ai/flux/schnell": {
     "description": "FLUX.1 [schnell]은 120억 개의 매개변수를 가진 이미지 생성 모델로, 빠른 고품질 이미지 생성을 중점으로 합니다."
   },
+  "fal-ai/hunyuan-image/v3": {
+    "description": "강력한 네이티브 멀티모달 이미지 생성 모델"
+  },
   "fal-ai/imagen4/preview": {
     "description": "Google에서 제공하는 고품질 이미지 생성 모델입니다."
   },

package/locales/nl-NL/models.json

@@ -1253,6 +1253,9 @@
   "fal-ai/flux/schnell": {
     "description": "FLUX.1 [schnell] is een beeldgeneratiemodel met 12 miljard parameters, gericht op het snel genereren van hoogwaardige beelden."
   },
+  "fal-ai/hunyuan-image/v3": {
+    "description": "Een krachtig native multimodaal beeldgeneratiemodel"
+  },
   "fal-ai/imagen4/preview": {
     "description": "Hoogwaardig beeldgeneratiemodel aangeboden door Google."
   },

package/locales/pl-PL/models.json

@@ -1253,6 +1253,9 @@
   "fal-ai/flux/schnell": {
     "description": "FLUX.1 [schnell] to model generowania obrazów z 12 miliardami parametrów, skoncentrowany na szybkim tworzeniu wysokiej jakości obrazów."
   },
+  "fal-ai/hunyuan-image/v3": {
+    "description": "Potężny natywny model generowania obrazów multimodalnych"
+  },
   "fal-ai/imagen4/preview": {
     "description": "Wysokiej jakości model generowania obrazów udostępniony przez Google."
   },

package/locales/pt-BR/models.json

@@ -1253,6 +1253,9 @@
   "fal-ai/flux/schnell": {
     "description": "FLUX.1 [schnell] é um modelo de geração de imagens com 12 bilhões de parâmetros, focado em gerar imagens de alta qualidade rapidamente."
   },
+  "fal-ai/hunyuan-image/v3": {
+    "description": "Um poderoso modelo nativo de geração de imagens multimodais"
+  },
   "fal-ai/imagen4/preview": {
     "description": "Modelo de geração de imagens de alta qualidade fornecido pelo Google."
   },

package/locales/ru-RU/models.json

@@ -1253,6 +1253,9 @@
   "fal-ai/flux/schnell": {
     "description": "FLUX.1 [schnell] — модель генерации изображений с 12 миллиардами параметров, ориентированная на быструю генерацию высококачественных изображений."
   },
+  "fal-ai/hunyuan-image/v3": {
+    "description": "Мощная нативная мультимодальная модель генерации изображений"
+  },
   "fal-ai/imagen4/preview": {
     "description": "Высококачественная модель генерации изображений от Google."
   },

package/locales/tr-TR/models.json

@@ -1253,6 +1253,9 @@
   "fal-ai/flux/schnell": {
     "description": "FLUX.1 [schnell], 12 milyar parametreye sahip bir görüntü oluşturma modelidir ve hızlı yüksek kaliteli görüntü üretimine odaklanır."
   },
+  "fal-ai/hunyuan-image/v3": {
+    "description": "Güçlü bir yerel çok modlu görüntü oluşturma modeli"
+  },
   "fal-ai/imagen4/preview": {
     "description": "Google tarafından sunulan yüksek kaliteli görüntü oluşturma modeli."
   },

package/locales/vi-VN/models.json

@@ -1253,6 +1253,9 @@
   "fal-ai/flux/schnell": {
     "description": "FLUX.1 [schnell] là mô hình tạo ảnh với 12 tỷ tham số, tập trung vào việc tạo ảnh chất lượng cao nhanh chóng."
   },
+  "fal-ai/hunyuan-image/v3": {
+    "description": "Một mô hình tạo hình ảnh đa phương thức gốc mạnh mẽ"
+  },
   "fal-ai/imagen4/preview": {
     "description": "Mô hình tạo ảnh chất lượng cao do Google cung cấp."
   },

package/locales/zh-CN/models.json

@@ -1253,6 +1253,9 @@
   "fal-ai/flux/schnell": {
     "description": "FLUX.1 [schnell] 是一个具有120亿参数的图像生成模型，专注于快速生成高质量图像。"
   },
+  "fal-ai/hunyuan-image/v3": {
+    "description": "一个强大的原生多模态图像生成模型"
+  },
   "fal-ai/imagen4/preview": {
     "description": "Google 提供的高质量的图像生成模型"
   },

package/locales/zh-TW/models.json

@@ -1253,6 +1253,9 @@
   "fal-ai/flux/schnell": {
     "description": "FLUX.1 [schnell] 是一個具有120億參數的圖像生成模型，專注於快速生成高品質圖像。"
   },
+  "fal-ai/hunyuan-image/v3": {
+    "description": "一個強大的原生多模態圖像生成模型"
+  },
   "fal-ai/imagen4/preview": {
     "description": "Google 提供的高品質圖像生成模型"
   },

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@lobehub/chat",
-  "version": "1.134.7",
+  "version": "1.135.0",
   "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/packages/model-bank/src/aiModels/fal.ts

@@ -79,6 +79,34 @@ const falImageModels: AIImageModelCard[] = [
     releasedAt: '2025-09-09',
     type: 'image',
   },
+  {
+    description: '一个强大的原生多模态图像生成模型',
+    displayName: 'HunyuanImage 3.0',
+    enabled: true,
+    id: 'fal-ai/hunyuan-image/v3',
+    parameters: {
+      cfg: { default: 7.5, max: 20, min: 1, step: 0.1 },
+      prompt: { default: '' },
+      seed: { default: null },
+      size: {
+        default: 'square_hd',
+        enum: [
+          'square_hd',
+          'square',
+          'portrait_4_3',
+          'portrait_16_9',
+          'landscape_4_3',
+          'landscape_16_9',
+        ],
+      },
+      steps: { default: 28, max: 50, min: 1, step: 1 },
+    },
+    pricing: {
+      units: [{ name: 'imageGeneration', rate: 0.1, strategy: 'fixed', unit: 'megapixel' }],
+    },
+    releasedAt: '2025-09-28',
+    type: 'image',
+  },
   {
     description: '专注于图像编辑任务的FLUX.1模型，支持文本和图像输入。',
     displayName: 'FLUX.1 Kontext [dev]',

package/packages/model-runtime/src/core/openaiCompatibleFactory/createImage.ts

@@ -1,3 +1,4 @@
+import { cleanObject } from '@lobechat/utils/object';
 import createDebug from 'debug';
 import { RuntimeImageGenParamsValue } from 'model-bank';
 import OpenAI from 'openai';
@@ -34,24 +35,25 @@ async function generateByImageMode(
       value,
     ]),
   );
+  // unify image input to array
+  if (typeof userInput.image === 'string' && userInput.image.trim() !== '') {
+    userInput.image = [userInput.image];
+  }
 
   // https://platform.openai.com/docs/api-reference/images/createEdit
   const isImageEdit = Array.isArray(userInput.image) && userInput.image.length > 0;
+  log('isImageEdit: %O, userInput.image: %O', isImageEdit, userInput.image);
   // If there are imageUrls parameters, convert them to File objects
   if (isImageEdit) {
-    log('Converting imageUrls to File objects: %O', userInput.image);
     try {
       // Convert all image URLs to File objects
       const imageFiles = await Promise.all(
         userInput.image.map((url: string) => convertImageUrlToFile(url)),
       );
 
-      log('Successfully converted %d images to File objects', imageFiles.length);
-
       // According to official docs, if there are multiple images, pass an array; if only one, pass a single File
       userInput.image = imageFiles.length === 1 ? imageFiles[0] : imageFiles;
     } catch (error) {
-      log('Error converting imageUrls to File objects: %O', error);
       throw new Error(`Failed to convert image URLs to File objects: ${error}`);
     }
   } else {
@@ -68,11 +70,11 @@ async function generateByImageMode(
     ...(isImageEdit ? { input_fidelity: 'high' } : {}),
   };
 
-  const options = {
+  const options = cleanObject({
     model,
     ...defaultInput,
     ...userInput,
-  };
+  });
 
   log('options: %O', options);
 
@@ -83,13 +85,11 @@ async function generateByImageMode(
 
   // Check the integrity of response data
   if (!img || !img.data || !Array.isArray(img.data) || img.data.length === 0) {
-    log('Invalid image response: missing data array');
     throw new Error('Invalid image response: missing or empty data array');
   }
 
   const imageData = img.data[0];
   if (!imageData) {
-    log('Invalid image response: first data item is null/undefined');
     throw new Error('Invalid image response: first data item is null or undefined');
   }
 
@@ -111,12 +111,9 @@ async function generateByImageMode(
   }
   // If neither format exists, throw error
   else {
-    log('Invalid image response: missing both b64_json and url fields');
     throw new Error('Invalid image response: missing both b64_json and url fields');
   }
 
-  log('provider: %s', provider);
-
   return {
     imageUrl,
     ...(img.usage
@@ -180,7 +177,6 @@ async function generateByChatModel(
       });
       log('Successfully processed image URL for chat input');
     } catch (error) {
-      log('Error processing image URL: %O', error);
       throw new Error(`Failed to process image URL: ${error}`);
     }
   }
@@ -218,7 +214,6 @@ async function generateByChatModel(
   }
 
   // If no images found, throw error
-  log('No images found in chat completion response');
   throw new Error('No image generated in chat completion response');
 }
 
@@ -228,21 +223,15 @@ async function generateByChatModel(
 export async function createOpenAICompatibleImage(
   client: OpenAI,
   payload: CreateImagePayload,
-  provider: string, // eslint-disable-line @typescript-eslint/no-unused-vars
+  provider: string,
 ): Promise<CreateImageResponse> {
-  try {
-    const { model } = payload;
-
-    // Check if it's a chat model for image generation (via :image suffix)
-    if (model.endsWith(':image')) {
-      return await generateByChatModel(client, payload);
-    }
+  const { model } = payload;
 
-    // Default to traditional images API
-    return await generateByImageMode(client, payload, provider);
-  } catch (error) {
-    const err = error as Error;
-    log('Error in createImage: %O', err);
-    throw err;
+  // Check if it's a chat model for image generation (via :image suffix)
+  if (model.endsWith(':image')) {
+    return await generateByChatModel(client, payload);
   }
+
+  // Default to traditional images API
+  return await generateByImageMode(client, payload, provider);
 }

package/packages/model-runtime/src/core/openaiCompatibleFactory/index.test.ts

@@ -296,6 +296,8 @@ describe('LobeOpenAICompatibleFactory', () => {
       });
 
       it('should transform non-streaming response to stream correctly', async () => {
+        vi.useFakeTimers();
+
         const mockResponse = {
           id: 'a',
           object: 'chat.completion',
@@ -319,13 +321,18 @@ describe('LobeOpenAICompatibleFactory', () => {
           mockResponse as any,
         );
 
-        const result = await instance.chat({
+        const chatPromise = instance.chat({
           messages: [{ content: 'Hello', role: 'user' }],
           model: 'mistralai/mistral-7b-instruct:free',
           temperature: 0,
           stream: false,
         });
 
+        // Advance time to simulate processing delay
+        vi.advanceTimersByTime(10);
+
+        const result = await chatPromise;
+
         const decoder = new TextDecoder();
         const reader = result.body!.getReader();
         const stream: string[] = [];
@@ -345,16 +352,20 @@ describe('LobeOpenAICompatibleFactory', () => {
           'data: {"inputTextTokens":5,"outputTextTokens":5,"totalInputTokens":5,"totalOutputTokens":5,"totalTokens":10}\n\n',
           'id: output_speed\n',
           'event: speed\n',
-          expect.stringMatching(/^data: \{.*"tps":.*,"ttft":.*}\n\n$/), // tps ttft 测试结果不一样
+          expect.stringMatching(/^data: \{.*"tps":.*,"ttft":.*}\n\n$/), // tps ttft should be calculated with elapsed time
           'id: a\n',
           'event: stop\n',
           'data: "stop"\n\n',
         ]);
 
         expect((await reader.read()).done).toBe(true);
+
+        vi.useRealTimers();
       });
 
       it('should transform non-streaming response to stream correctly with reasoning content', async () => {
+        vi.useFakeTimers();
+
         const mockResponse = {
           id: 'a',
           object: 'chat.completion',
@@ -382,13 +393,18 @@ describe('LobeOpenAICompatibleFactory', () => {
           mockResponse as any,
         );
 
-        const result = await instance.chat({
+        const chatPromise = instance.chat({
           messages: [{ content: 'Hello', role: 'user' }],
           model: 'deepseek/deepseek-reasoner',
           temperature: 0,
           stream: false,
         });
 
+        // Advance time to simulate processing delay
+        vi.advanceTimersByTime(10);
+
+        const result = await chatPromise;
+
         const decoder = new TextDecoder();
         const reader = result.body!.getReader();
         const stream: string[] = [];
@@ -411,13 +427,15 @@ describe('LobeOpenAICompatibleFactory', () => {
           'data: {"inputTextTokens":5,"outputTextTokens":5,"totalInputTokens":5,"totalOutputTokens":5,"totalTokens":10}\n\n',
           'id: output_speed\n',
           'event: speed\n',
-          expect.stringMatching(/^data: \{.*"tps":.*,"ttft":.*}\n\n$/), // tps ttft 测试结果不一样
+          expect.stringMatching(/^data: \{.*"tps":.*,"ttft":.*}\n\n$/), // tps ttft should be calculated with elapsed time
           'id: a\n',
           'event: stop\n',
           'data: "stop"\n\n',
         ]);
 
         expect((await reader.read()).done).toBe(true);
+
+        vi.useRealTimers();
       });
     });
 
@@ -974,7 +992,11 @@ describe('LobeOpenAICompatibleFactory', () => {
           .spyOn(inst['client'].responses, 'create')
           .mockResolvedValue({ tee: () => [prod, debug] } as any);
 
-        await inst.chat({ messages: [{ content: 'hi', role: 'user' }], model: 'any-model', temperature: 0 });
+        await inst.chat({
+          messages: [{ content: 'hi', role: 'user' }],
+          model: 'any-model',
+          temperature: 0,
+        });
 
         expect(mockResponsesCreate).toHaveBeenCalled();
       });
@@ -990,20 +1012,38 @@ describe('LobeOpenAICompatibleFactory', () => {
         const inst = new LobeMockProviderUseResponseModels({ apiKey: 'test' });
         const spy = vi.spyOn(inst['client'].responses, 'create');
         // Prevent hanging by mocking normal chat completion stream
-        vi.spyOn(inst['client'].chat.completions, 'create').mockResolvedValue(new ReadableStream() as any);
+        vi.spyOn(inst['client'].chat.completions, 'create').mockResolvedValue(
+          new ReadableStream() as any,
+        );
 
         // First invocation: model contains the string
-        spy.mockResolvedValueOnce({ tee: () => [new ReadableStream(), new ReadableStream()] } as any);
-        await inst.chat({ messages: [{ content: 'hi', role: 'user' }], model: 'prefix-special-model-suffix', temperature: 0 });
+        spy.mockResolvedValueOnce({
+          tee: () => [new ReadableStream(), new ReadableStream()],
+        } as any);
+        await inst.chat({
+          messages: [{ content: 'hi', role: 'user' }],
+          model: 'prefix-special-model-suffix',
+          temperature: 0,
+        });
         expect(spy).toHaveBeenCalledTimes(1);
 
         // Second invocation: model matches the RegExp
-        spy.mockResolvedValueOnce({ tee: () => [new ReadableStream(), new ReadableStream()] } as any);
-        await inst.chat({ messages: [{ content: 'hi', role: 'user' }], model: 'special-xyz', temperature: 0 });
+        spy.mockResolvedValueOnce({
+          tee: () => [new ReadableStream(), new ReadableStream()],
+        } as any);
+        await inst.chat({
+          messages: [{ content: 'hi', role: 'user' }],
+          model: 'special-xyz',
+          temperature: 0,
+        });
         expect(spy).toHaveBeenCalledTimes(2);
 
         // Third invocation: model does not match any useResponseModels patterns
-        await inst.chat({ messages: [{ content: 'hi', role: 'user' }], model: 'unrelated-model', temperature: 0 });
+        await inst.chat({
+          messages: [{ content: 'hi', role: 'user' }],
+          model: 'unrelated-model',
+          temperature: 0,
+        });
         expect(spy).toHaveBeenCalledTimes(2); // Ensure no additional calls were made
       });
     });

package/packages/model-runtime/src/core/streams/protocol.ts

@@ -384,7 +384,6 @@ export const createTokenSpeedCalculator = (
   }: { enableStreaming?: boolean; inputStartAt?: number; streamStack?: StreamContext } = {},
 ) => {
   let outputStartAt: number | undefined;
-  let outputThinking: boolean | undefined;
 
   const process = (chunk: StreamProtocolChunk) => {
     let result = [chunk];
@@ -393,24 +392,12 @@ export const createTokenSpeedCalculator = (
       outputStartAt = Date.now();
     }
 
-    /**
-     * 部分 provider 在正式输出 reasoning 前，可能会先输出 content 为空字符串的 chunk，
-     * 其中 reasoning 可能为 null，会导致判断是否输出思考内容错误，所以过滤掉 null 或者空字符串。
-     * 也可能是某些特殊 token，所以不修改 outputStartAt 的逻辑。
-     */
-    if (
-      outputThinking === undefined &&
-      (chunk.type === 'text' || chunk.type === 'reasoning') &&
-      typeof chunk.data === 'string' &&
-      chunk.data.length > 0
-    ) {
-      outputThinking = chunk.type === 'reasoning';
-    }
     // if the chunk is the stop chunk, set as output finish
     if (inputStartAt && outputStartAt && chunk.type === 'usage') {
       // TPS should always include all generated tokens (including reasoning tokens)
       // because it measures generation speed, not just visible content
-      const outputTokens = chunk.data?.totalOutputTokens ?? 0;
+      const usage = chunk.data as ModelUsage;
+      const outputTokens = usage?.totalOutputTokens ?? 0;
       const now = Date.now();
       const elapsed = now - (enableStreaming ? outputStartAt : inputStartAt);
       const duration = now - outputStartAt;

package/packages/model-runtime/src/providers/fal/index.ts

@@ -33,6 +33,7 @@ export class LobeFalAI implements LobeRuntimeAI {
       ['cfg', 'guidance_scale'],
       ['imageUrl', 'image_url'],
       ['imageUrls', 'image_urls'],
+      ['size', 'image_size'],
     ]);
 
     const defaultInput: Record<string, unknown> = {
@@ -50,12 +51,16 @@ export class LobeFalAI implements LobeRuntimeAI {
     );
 
     if ('width' in userInput && 'height' in userInput) {
-      userInput.image_size = {
-        height: userInput.height,
-        width: userInput.width,
-      };
-      delete userInput.width;
-      delete userInput.height;
+      if (userInput.size) {
+        throw new Error('width/height and size are not supported at the same time');
+      } else {
+        userInput.image_size = {
+          height: userInput.height,
+          width: userInput.width,
+        };
+        delete userInput.width;
+        delete userInput.height;
+      }
     }
 
     const modelsAcceleratedByDefault = new Set<string>(['flux/krea']);
@@ -66,7 +71,7 @@ export class LobeFalAI implements LobeRuntimeAI {
     // Ensure model has fal-ai/ prefix
     let endpoint = model.startsWith('fal-ai/') ? model : `fal-ai/${model}`;
     const hasImageUrls = (params.imageUrls?.length ?? 0) > 0;
-    if (endpoint === 'fal-ai/bytedance/seedream/v4') {
+    if (['fal-ai/bytedance/seedream/v4', 'fal-ai/hunyuan-image/v3'].includes(endpoint)) {
       endpoint += hasImageUrls ? '/edit' : '/text-to-image';
     } else if (endpoint === 'fal-ai/nano-banana' && hasImageUrls) {
       endpoint += '/edit';

package/packages/model-runtime/src/providers/newapi/index.test.ts

@@ -563,7 +563,22 @@ describe('NewAPI Runtime - 100% Branch Coverage', () => {
 
           if (inputPrice !== undefined) {
             const outputPrice = inputPrice * (pricing.completion_ratio || 1);
-            enhancedModel.pricing = { input: inputPrice, output: outputPrice };
+            enhancedModel.pricing = {
+              units: [
+                {
+                  name: 'textInput',
+                  unit: 'millionTokens',
+                  strategy: 'fixed',
+                  rate: inputPrice,
+                },
+                {
+                  name: 'textOutput',
+                  unit: 'millionTokens',
+                  strategy: 'fixed',
+                  rate: outputPrice,
+                },
+              ],
+            };
           }
         }
 
@@ -582,8 +597,18 @@ describe('NewAPI Runtime - 100% Branch Coverage', () => {
       });
 
       // Verify pricing results
-      expect(enrichedModels[0].pricing).toEqual({ input: 40, output: 120 }); // model_price * 2, input * completion_ratio
-      expect(enrichedModels[1].pricing).toEqual({ input: 10, output: 10 }); // model_ratio * 2, input * 1 (default)
+      expect(enrichedModels[0].pricing).toEqual({
+        units: [
+          { name: 'textInput', unit: 'millionTokens', strategy: 'fixed', rate: 40 },
+          { name: 'textOutput', unit: 'millionTokens', strategy: 'fixed', rate: 120 },
+        ],
+      }); // model_price * 2, input * completion_ratio
+      expect(enrichedModels[1].pricing).toEqual({
+        units: [
+          { name: 'textInput', unit: 'millionTokens', strategy: 'fixed', rate: 10 },
+          { name: 'textOutput', unit: 'millionTokens', strategy: 'fixed', rate: 10 },
+        ],
+      }); // model_ratio * 2, input * 1 (default)
       expect(enrichedModels[2].pricing).toBeUndefined(); // quota_type = 1, skipped
 
       // Verify provider detection

package/packages/model-runtime/src/providers/newapi/index.ts

@@ -20,41 +20,21 @@ export interface NewAPIPricing {
   model_name: string;
   model_price?: number;
   model_ratio?: number;
-  quota_type: number; // 0: 按量计费, 1: 按次计费
+  /** 0: Pay-per-token, 1: Pay-per-call */
+  quota_type: number;
   supported_endpoint_types?: string[];
 }
 
 const handlePayload = (payload: ChatStreamPayload) => {
-  // 处理 OpenAI responses API 模式
+  // Handle OpenAI responses API mode
   if (
     responsesAPIModels.has(payload.model) ||
     payload.model.includes('gpt-') ||
     /^o\d/.test(payload.model)
   ) {
-    return { ...payload, apiMode: 'responses' } as any;
+    return { ...payload, apiMode: 'responses' };
   }
-  return payload as any;
-};
-
-// 根据 owned_by 字段判断提供商（基于 NewAPI 的 channel name）
-const getProviderFromOwnedBy = (ownedBy: string): string => {
-  const normalizedOwnedBy = ownedBy.toLowerCase();
-
-  if (normalizedOwnedBy.includes('claude') || normalizedOwnedBy.includes('anthropic')) {
-    return 'anthropic';
-  }
-  if (normalizedOwnedBy.includes('google') || normalizedOwnedBy.includes('gemini')) {
-    return 'google';
-  }
-  if (normalizedOwnedBy.includes('xai') || normalizedOwnedBy.includes('grok')) {
-    return 'xai';
-  }
-  if (normalizedOwnedBy.includes('ali') || normalizedOwnedBy.includes('qwen')) {
-    return 'qwen';
-  }
-
-  // 默认为 openai
-  return 'openai';
+  return payload;
 };
 
 export const LobeNewAPIAI = createRouterRuntime({
@@ -66,16 +46,16 @@ export const LobeNewAPIAI = createRouterRuntime({
   },
   id: ModelProvider.NewAPI,
   models: async ({ client: openAIClient }) => {
-    // 获取基础 URL（移除末尾的 API 版本路径如 /v1、/v1beta 等）
+    // Get base URL (remove trailing API version paths like /v1, /v1beta, etc.)
     const baseURL = openAIClient.baseURL.replace(/\/v\d+[a-z]*\/?$/, '');
 
     const modelsPage = (await openAIClient.models.list()) as any;
     const modelList: NewAPIModelCard[] = modelsPage.data || [];
 
-    // 尝试获取 pricing 信息以补充模型详细信息
+    // Try to get pricing information to enrich model details
     let pricingMap: Map<string, NewAPIPricing> = new Map();
     try {
-      // 使用保存的 baseURL
+      // Use saved baseURL
       const pricingResponse = await fetch(`${baseURL}/api/pricing`, {
         headers: {
           Authorization: `Bearer ${openAIClient.apiKey}`,
@@ -99,22 +79,22 @@ export const LobeNewAPIAI = createRouterRuntime({
     const enrichedModelList = modelList.map((model) => {
       let enhancedModel: any = { ...model };
 
-      // 1. 添加 pricing 信息
+      // add pricing info
       const pricing = pricingMap.get(model.id);
       if (pricing) {
-        // NewAPI 的价格计算逻辑：
-        // - quota_type: 0 表示按量计费（按 token），1 表示按次计费
-        // - model_ratio: 相对于基础价格的倍率（基础价格 = $0.002/1K tokens）
-        // - model_price: 直接指定的价格（优先使用）
-        // - completion_ratio: 输出价格相对于输入价格的倍率
+        // NewAPI pricing calculation logic:
+        // - quota_type: 0 means pay-per-token, 1 means pay-per-call
+        // - model_ratio: multiplier relative to base price (base price = $0.002/1K tokens)
+        // - model_price: directly specified price (takes priority)
+        // - completion_ratio: output price multiplier relative to input price
         //
-        // LobeChat 需要的格式：美元/百万 token
+        // LobeChat required format: USD per million tokens
 
         let inputPrice: number | undefined;
         let outputPrice: number | undefined;
 
         if (pricing.quota_type === 0) {
-          // 按量计费
+          // Pay-per-token
           if (pricing.model_price && pricing.model_price > 0) {
             // model_price is a direct price value; need to confirm its unit.
             // Assumption: model_price is the price per 1,000 tokens (i.e., $/1K tokens).
@@ -124,62 +104,38 @@ export const LobeNewAPIAI = createRouterRuntime({
             inputPrice = pricing.model_price * 2;
           } else if (pricing.model_ratio) {
             // model_ratio × $0.002/1K = model_ratio × $2/1M
-            inputPrice = pricing.model_ratio * 2; // 转换为 $/1M tokens
+            inputPrice = pricing.model_ratio * 2; // Convert to $/1M tokens
           }
 
           if (inputPrice !== undefined) {
-            // 计算输出价格
+            // Calculate output price
             outputPrice = inputPrice * (pricing.completion_ratio || 1);
 
             enhancedModel.pricing = {
-              input: inputPrice,
-              output: outputPrice,
+              units: [
+                {
+                  name: 'textInput',
+                  rate: inputPrice,
+                  strategy: 'fixed',
+                  unit: 'millionTokens',
+                },
+                {
+                  name: 'textOutput',
+                  rate: outputPrice,
+                  strategy: 'fixed',
+                  unit: 'millionTokens',
+                },
+              ],
             };
           }
         }
-        // quota_type === 1 按次计费暂不支持
-      }
-
-      // 2. 根据优先级处理 provider 信息并缓存路由
-      let detectedProvider = 'openai'; // 默认
-
-      // 优先级1：使用 supported_endpoint_types
-      if (model.supported_endpoint_types && model.supported_endpoint_types.length > 0) {
-        if (model.supported_endpoint_types.includes('anthropic')) {
-          detectedProvider = 'anthropic';
-        } else if (model.supported_endpoint_types.includes('gemini')) {
-          detectedProvider = 'google';
-        } else if (model.supported_endpoint_types.includes('xai')) {
-          detectedProvider = 'xai';
-        } else if (model.supported_endpoint_types.includes('qwen')) {
-          detectedProvider = 'qwen';
-        }
-      }
-      // 优先级2：使用 owned_by 字段
-      else if (model.owned_by) {
-        detectedProvider = getProviderFromOwnedBy(model.owned_by);
-      }
-      // 优先级3：基于模型名称检测
-      else {
-        detectedProvider = detectModelProvider(model.id);
+        // quota_type === 1 pay-per-call is not currently supported
       }
 
-      // 将检测到的 provider 信息附加到模型上
-      enhancedModel._detectedProvider = detectedProvider;
-
       return enhancedModel;
     });
 
-    // 使用 processMultiProviderModelList 处理模型能力
-    const processedModels = await processMultiProviderModelList(enrichedModelList, 'newapi');
-
-    // 清理临时字段
-    return processedModels.map((model: any) => {
-      if (model._detectedProvider) {
-        delete model._detectedProvider;
-      }
-      return model;
-    });
+    return processMultiProviderModelList(enrichedModelList, 'newapi');
   },
   routers: (options) => {
     const userBaseURL = options.baseURL?.replace(/\/v\d+[a-z]*\/?$/, '') || '';
@@ -215,16 +171,6 @@ export const LobeNewAPIAI = createRouterRuntime({
           baseURL: urlJoin(userBaseURL, '/v1'),
         },
       },
-      {
-        apiType: 'qwen',
-        models: LOBE_DEFAULT_MODEL_LIST.map((m) => m.id).filter(
-          (id) => detectModelProvider(id) === 'qwen',
-        ),
-        options: {
-          ...options,
-          baseURL: urlJoin(userBaseURL, '/v1'),
-        },
-      },
       {
         apiType: 'openai',
         options: {

package/packages/model-runtime/src/types/index.ts

@@ -8,4 +8,3 @@ export * from './textToImage';
 export * from './toolsCalling';
 export * from './tts';
 export * from './type';
-export * from './usage';

package/packages/types/src/message/base.ts

@@ -61,6 +61,7 @@ export interface ModelTokensUsage {
   rejectedPredictionTokens?: number;
 
   // Total tokens
+  // TODO: make all following fields required
   totalInputTokens?: number;
   totalOutputTokens?: number;
   totalTokens?: number;

package/packages/utils/package.json

@@ -5,7 +5,8 @@
   "exports": {
     ".": "./src/index.ts",
     "./server": "./src/server/index.ts",
-    "./client": "./src/client/index.ts"
+    "./client": "./src/client/index.ts",
+    "./object": "./src/object.ts"
   },
   "scripts": {
     "test": "vitest",

package/src/app/[variants]/(main)/image/@menu/components/SizeSelect/index.tsx

@@ -1,6 +1,6 @@
 'use client';
 
-import { Block, Grid, GridProps, Text } from '@lobehub/ui';
+import { Block, Grid, GridProps, Select, Text } from '@lobehub/ui';
 import { useTheme } from 'antd-style';
 import { ReactNode, memo } from 'react';
 import { Center } from 'react-layout-kit';
@@ -13,6 +13,19 @@ export interface SizeSelectProps extends Omit<GridProps, 'children' | 'onChange'
   value?: 'auto' | string;
 }
 
+/**
+ * Check if a size value can be parsed as valid aspect ratio
+ */
+const canParseAsRatio = (value: string): boolean => {
+  if (value === 'auto') return true;
+
+  const parts = value.split('x');
+  if (parts.length !== 2) return false;
+
+  const [width, height] = parts.map(Number);
+  return !isNaN(width) && !isNaN(height) && width > 0 && height > 0;
+};
+
 const SizeSelect = memo<SizeSelectProps>(({ options, onChange, value, defaultValue, ...rest }) => {
   const theme = useTheme();
   const [active, setActive] = useMergeState('auto', {
@@ -20,6 +33,16 @@ const SizeSelect = memo<SizeSelectProps>(({ options, onChange, value, defaultVal
     onChange,
     value,
   });
+
+  // Check if all options can be parsed as valid ratios
+  const hasInvalidRatio = options?.some((item) => !canParseAsRatio(item.value));
+
+  // If any option cannot be parsed as ratio, fallback to regular Select
+  if (hasInvalidRatio) {
+    return (
+      <Select onChange={onChange} options={options} style={{ width: '100%' }} value={active} />
+    );
+  }
   return (
     <Block padding={4} variant={'filled'} {...rest}>
       <Grid gap={4} maxItemWidth={72} rows={16}>

package/src/server/modules/EdgeConfig/index.ts

@@ -3,22 +3,9 @@ import createDebug from 'debug';
 
 import { appEnv } from '@/envs/app';
 
-const debug = createDebug('lobe-server:edge-config');
+import { EdgeConfigData, EdgeConfigKeys } from './types';
 
-const EdgeConfigKeys = {
-  /**
-   * Assistant whitelist
-   */
-  AssistantBlacklist: 'assistant_blacklist',
-  /**
-   * Assistant whitelist
-   */
-  AssistantWhitelist: 'assistant_whitelist',
-  /**
-   * Feature flags configuration
-   */
-  FeatureFlags: 'feature_flags',
-};
+const debug = createDebug('lobe-server:edge-config');
 
 export class EdgeConfig {
   get client(): EdgeConfigClient {
@@ -38,29 +25,24 @@ export class EdgeConfig {
     return isEnabled;
   }
 
-  getAgentRestrictions = async () => {
-    const { assistant_blacklist: blacklist, assistant_whitelist: whitelist } =
-      await this.client.getAll([
-        EdgeConfigKeys.AssistantWhitelist,
-        EdgeConfigKeys.AssistantBlacklist,
-      ]);
+  private async getValue<K extends EdgeConfigKeys>(key: K) {
+    return this.client.get<EdgeConfigData[K]>(key);
+  }
 
-    return { blacklist, whitelist } as {
-      blacklist: string[] | undefined;
-      whitelist: string[] | undefined;
-    };
-  };
+  private async getValues<const K extends EdgeConfigKeys>(keys: K[]) {
+    return this.client.getAll<Pick<EdgeConfigData, K>>(keys);
+  }
 
-  getFlagByKey = async (key: string) => {
-    const value = await this.client.get(key);
-    return value;
+  getAgentRestrictions = async () => {
+    const { assistant_blacklist: blacklist, assistant_whitelist: whitelist } = await this.getValues(
+      ['assistant_blacklist', 'assistant_whitelist'],
+    );
+    return { blacklist, whitelist };
   };
 
   getFeatureFlags = async () => {
-    const featureFlags = await this.client.get(EdgeConfigKeys.FeatureFlags);
+    const featureFlags = await this.getValue('feature_flags');
     debug('Feature flags retrieved: %O', featureFlags);
-    return featureFlags as Record<string, boolean | string[]> | undefined;
+    return featureFlags;
   };
 }
-
-export { EdgeConfigKeys };

package/src/server/modules/EdgeConfig/types.ts

@@ -4,6 +4,19 @@
  * EdgeConfig 完整配置类型
  */
 export interface EdgeConfigData {
+  /**
+   * Assistant blacklist
+   */
   assistant_blacklist?: string[];
+  /**
+   * Assistant whitelist
+   */
   assistant_whitelist?: string[];
+
+  /**
+   * Feature flags configuration
+   */
+  feature_flags?: Record<string, boolean | string[]>;
 }
+
+export type EdgeConfigKeys = keyof EdgeConfigData;

package/packages/model-runtime/src/types/usage.ts

@@ -1,27 +0,0 @@
-export interface ModelTokensUsage {
-  acceptedPredictionTokens?: number;
-  inputAudioTokens?: number;
-  inputCacheMissTokens?: number;
-  inputCachedTokens?: number;
-  /**
-   * currently only pplx has citation_tokens
-   */
-  inputCitationTokens?: number;
-  /**
-   * user prompt image
-   */
-  inputImageTokens?: number;
-  /**
-   * user prompt input
-   */
-  inputTextTokens?: number;
-  inputWriteCacheTokens?: number;
-  outputAudioTokens?: number;
-  outputImageTokens?: number;
-  outputReasoningTokens?: number;
-  outputTextTokens?: number;
-  rejectedPredictionTokens?: number;
-  totalInputTokens?: number;
-  totalOutputTokens?: number;
-  totalTokens?: number;
-}