feat: add support for qwenlm and image upload (#726)

* feat: add support for qwenlm and image upload

* fix: qwenlm return

* feat: add provider config

src/main/ipc.ts

@@ -117,6 +117,7 @@ export function registerIpc(mainWindow: BrowserWindow, app: Electron.App) {
   ipcMain.handle('file:base64Image', fileManager.base64Image)
   ipcMain.handle('file:download', fileManager.downloadFile)
   ipcMain.handle('file:copy', fileManager.copyFile)
+  ipcMain.handle('file:binaryFile', fileManager.binaryFile)
 
   // minapp
   ipcMain.handle('minapp', (_, args) => {

src/main/services/FileStorage.ts

@@ -263,6 +263,13 @@ class FileStorage {
     }
   }
 
+  public binaryFile = async (_: Electron.IpcMainInvokeEvent, id: string): Promise<{ data: Buffer; mime: string }> => {
+    const filePath = path.join(this.storageDir, id)
+    const data = await fs.promises.readFile(filePath)
+    const mime = `image/${path.extname(filePath).slice(1)}`
+    return { data, mime }
+  }
+
   public clear = async (): Promise<void> => {
     await fs.promises.rmdir(this.storageDir, { recursive: true })
     await this.initStorageDir()

src/preload/index.d.ts

@@ -53,6 +53,7 @@ declare global {
         base64Image: (fileId: string) => Promise<{ mime: string; base64: string; data: string }>
         download: (url: string) => Promise<FileType | null>
         copy: (fileId: string, destPath: string) => Promise<void>
+        binaryFile: (fileId: string) => Promise<{ data: Buffer; mime: string }>
       }
       export: {
         toWord: (markdown: string, fileName: string) => Promise<void>

src/preload/index.ts

@@ -43,7 +43,8 @@ const api = {
     saveImage: (name: string, data: string) => ipcRenderer.invoke('file:saveImage', name, data),
     base64Image: (fileId: string) => ipcRenderer.invoke('file:base64Image', fileId),
     download: (url: string) => ipcRenderer.invoke('file:download', url),
-    copy: (fileId: string, destPath: string) => ipcRenderer.invoke('file:copy', fileId, destPath)
+    copy: (fileId: string, destPath: string) => ipcRenderer.invoke('file:copy', fileId, destPath),
+    binaryFile: (fileId: string) => ipcRenderer.invoke('file:binaryFile', fileId)
   },
   export: {
     toWord: (markdown: string, fileName: string) => ipcRenderer.invoke('export:word', markdown, fileName)

src/renderer/src/config/models.ts

@@ -264,6 +264,56 @@ export function getModelLogo(modelId: string) {
 }
 
 export const SYSTEM_MODELS: Record<string, Model[]> = {
+  qwenlm: [
+    {
+      id: 'qwen-plus-latest',
+      provider: 'qwenlm',
+      name: 'Qwen2.5-Plus',
+      group: 'Qwen'
+    },
+    {
+      id: 'qvq-72b-preview',
+      provider: 'qwenlm',
+      name: 'QVQ-72B-Preview',
+      group: 'Qwen'
+    },
+    {
+      id: 'qwq-32b-preview',
+      provider: 'qwenlm',
+      name: 'QwQ-32B-Preview',
+      group: 'Qwen'
+    },
+    {
+      id: 'qwen2.5-coder-32b-instruct',
+      provider: 'qwenlm',
+      name: 'Qwen2.5-Coder-32B-Instruct',
+      group: 'Qwen'
+    },
+    {
+      id: 'qwen-vl-max-latest',
+      provider: 'qwenlm',
+      name: 'Qwen2-VL-Max',
+      group: 'Qwen'
+    },
+    {
+      id: 'qwen-turbo-latest',
+      provider: 'qwenlm',
+      name: 'Qwen2.5-Turbo',
+      group: 'Qwen'
+    },
+    {
+      id: 'qwen2.5-72b-instruct',
+      provider: 'qwenlm',
+      name: 'Qwen2.5-72B-Instruct',
+      group: 'Qwen'
+    },
+    {
+      id: 'qwen2.5-32b-instruct',
+      provider: 'qwenlm',
+      name: 'Qwen2.5-32B-Instruct',
+      group: 'Qwen'
+    }
+  ],
   aihubmix: [
     {
       id: 'gpt-4o',

src/renderer/src/config/providers.ts

@@ -28,6 +28,7 @@ import StepProviderLogo from '@renderer/assets/images/providers/step.png'
 import TogetherProviderLogo from '@renderer/assets/images/providers/together.png'
 import ZeroOneProviderLogo from '@renderer/assets/images/providers/zero-one.png'
 import ZhipuProviderLogo from '@renderer/assets/images/providers/zhipu.png'
+import QwenLMProviderLogo from '@renderer/assets/images/providers/qwenlm.png'
 
 export function getProviderLogo(providerId: string) {
   switch (providerId) {
@@ -91,6 +92,8 @@ export function getProviderLogo(providerId: string) {
       return MistralProviderLogo
     case 'jina':
       return JinaProviderLogo
+    case 'qwenlm':
+      return QwenLMProviderLogo
     default:
       return undefined
   }
@@ -418,5 +421,16 @@ export const PROVIDER_CONFIG = {
       docs: 'https://learn.microsoft.com/en-us/azure/ai-services/openai/',
       models: 'https://learn.microsoft.com/en-us/azure/ai-services/openai/concepts/models'
     }
+  },
+  qwenlm: {
+    api: {
+      url: 'https://chat.qwenlm.ai/api/'
+    },
+    websites: {
+      official: 'https://chat.qwenlm.ai',
+      apiKey: 'https://chat.qwenlm.ai',
+      docs: 'https://chat.qwenlm.ai',
+      models: 'https://chat.qwenlm.ai'
+    }
   }
 }

src/renderer/src/i18n/locales/en-us.json

@@ -326,7 +326,8 @@
       "together": "Together",
       "yi": "Yi",
       "zhinao": "360AI",
-      "zhipu": "ZHIPU AI"
+      "zhipu": "ZHIPU AI",
+      "qwenlm": "QwenLM"
     },
     "settings": {
       "about": "About & Feedback",

src/renderer/src/i18n/locales/ja-jp.json

@@ -324,7 +324,8 @@
       "together": "Together",
       "yi": "零一万物",
       "zhinao": "360智脳",
-      "zhipu": "智譜AI"
+      "zhipu": "智譜AI",
+      "qwenlm": "QwenLM"
     },
     "settings": {
       "about": "について",

src/renderer/src/i18n/locales/ru-ru.json

@@ -326,7 +326,8 @@
       "together": "Together",
       "yi": "Yi",
       "zhinao": "360AI",
-      "zhipu": "ZHIPU AI"
+      "zhipu": "ZHIPU AI",
+      "qwenlm": "QwenLM"
     },
     "settings": {
       "about": "О программе и обратная связь",

src/renderer/src/i18n/locales/zh-cn.json

@@ -327,7 +327,8 @@
       "together": "Together",
       "yi": "零一万物",
       "zhinao": "360智脑",
-      "zhipu": "智谱AI"
+      "zhipu": "智谱AI",
+      "qwenlm": "QwenLM"
     },
     "settings": {
       "about": "关于我们",

src/renderer/src/i18n/locales/zh-tw.json

@@ -326,7 +326,8 @@
       "together": "Together",
       "yi": "零一萬物",
       "zhinao": "360智腦",
-      "zhipu": "智譜AI"
+      "zhipu": "智譜AI",
+      "qwenlm": "QwenLM"
     },
     "settings": {
       "about": "關於與回饋",

src/renderer/src/providers/OpenAIProvider.ts

@@ -46,6 +46,33 @@ export default class OpenAIProvider extends BaseProvider {
     return providers.includes(this.provider.id)
   }
 
+  private async uploadImageToQwenLM(image_file: Buffer, file_name: string, mime: string): Promise<string> {
+    try {
+      // 创建 FormData
+      const formData = new FormData()
+      formData.append('file', new Blob([image_file], { type: mime }), file_name)
+
+      // 发送上传请求
+      const response = await fetch(`${this.provider.apiHost}v1/files/`, {
+        method: 'POST',
+        headers: {
+          Authorization: `Bearer ${this.apiKey}`
+        },
+        body: formData
+      })
+
+      if (!response.ok) {
+        throw new Error('Failed to upload image to QwenLM')
+      }
+
+      const data = await response.json()
+      return data.id
+    } catch (error) {
+      console.error('Error uploading image to QwenLM:', error)
+      throw error
+    }
+  }
+
   private async getMessageParam(
     message: Message,
     model: Model
@@ -94,20 +121,48 @@ export default class OpenAIProvider extends BaseProvider {
       }
     ]
 
-    for (const file of message.files || []) {
-      if (file.type === FileTypes.IMAGE && isVision) {
-        const image = await window.api.file.base64Image(file.id + file.ext)
-        parts.push({
-          type: 'image_url',
-          image_url: { url: image.data }
-        })
+    //QwenLM上传图片
+    if (this.provider.id === 'qwenlm') {
+      let qwenlm_image_url: { type: string; image: string }[] = []
+
+      for (const file of message.files || []) {
+        if (file.type === FileTypes.IMAGE && isVision) {
+          const image = await window.api.file.binaryFile(file.id + file.ext)
+
+          const imageId = await this.uploadImageToQwenLM(image.data, file.origin_name, image.mime)
+          qwenlm_image_url.push({
+            type: 'image',
+            image: imageId
+          })
+        }
+        if ([FileTypes.TEXT, FileTypes.DOCUMENT].includes(file.type)) {
+          const fileContent = await (await window.api.file.read(file.id + file.ext)).trim()
+          parts.push({
+            type: 'text',
+            text: file.origin_name + '\n' + fileContent
+          })
+        }
       }
-      if ([FileTypes.TEXT, FileTypes.DOCUMENT].includes(file.type)) {
-        const fileContent = await (await window.api.file.read(file.id + file.ext)).trim()
-        parts.push({
-          type: 'text',
-          text: file.origin_name + '\n' + fileContent
-        })
+      return {
+        role: message.role,
+        content: [...parts, ...qwenlm_image_url]
+      } as ChatCompletionMessageParam
+    } else {
+      for (const file of message.files || []) {
+        if (file.type === FileTypes.IMAGE && isVision) {
+          const image = await window.api.file.base64Image(file.id + file.ext)
+          parts.push({
+            type: 'image_url',
+            image_url: { url: image.data }
+          })
+        }
+        if ([FileTypes.TEXT, FileTypes.DOCUMENT].includes(file.type)) {
+          const fileContent = await (await window.api.file.read(file.id + file.ext)).trim()
+          parts.push({
+            type: 'text',
+            text: file.origin_name + '\n' + fileContent
+          })
+        }
       }
     }
 
@@ -172,6 +227,40 @@ export default class OpenAIProvider extends BaseProvider {
       })
     }
 
+    // 处理QwenLM的流式输出
+    if (this.provider.id === 'qwenlm') {
+      let accumulatedText = ''
+      for await (const chunk of stream) {
+        if (window.keyv.get(EVENT_NAMES.CHAT_COMPLETION_PAUSED)) {
+          break
+        }
+        if (time_first_token_millsec == 0) {
+          time_first_token_millsec = new Date().getTime() - start_time_millsec
+        }
+
+        // 获取当前块的完整内容
+        const currentContent = chunk.choices[0]?.delta?.content || ''
+
+        // 如果内容与累积的内容不同，则只发送增量部分
+        if (currentContent !== accumulatedText) {
+          const deltaText = currentContent.slice(accumulatedText.length)
+          accumulatedText = currentContent // 更新累积的文本
+
+          const time_completion_millsec = new Date().getTime() - start_time_millsec
+          onChunk({
+            text: deltaText,
+            usage: chunk.usage,
+            metrics: {
+              completion_tokens: chunk.usage?.completion_tokens,
+              time_completion_millsec,
+              time_first_token_millsec
+            }
+          })
+        }
+      }
+      return
+    }
+
     for await (const chunk of stream) {
       if (window.keyv.get(EVENT_NAMES.CHAT_COMPLETION_PAUSED)) {
         break

src/renderer/src/store/index.ts

@@ -30,7 +30,7 @@ const persistedReducer = persistReducer(
   {
     key: 'cherry-studio',
     storage,
-    version: 55,
+    version: 56,
     blacklist: ['runtime'],
     migrate
   },

src/renderer/src/store/llm.ts

@@ -323,6 +323,16 @@ const initialState: LlmState = {
       models: SYSTEM_MODELS.jina,
       isSystem: true,
       enabled: false
+    },
+    {
+      id: 'qwenlm',
+      name: 'QwenLM',
+      type: 'openai',
+      apiKey: '',
+      apiHost: 'https://chat.qwenlm.ai/api/',
+      models: SYSTEM_MODELS.qwenlm,
+      isSystem: true,
+      enabled: false
     }
   ],
   settings: {

src/renderer/src/store/migrate.ts

@@ -799,6 +799,19 @@ const migrateConfig = {
       }
     }
     return state
+  },
+  '56': (state: RootState) => {
+    state.llm.providers.push({
+      id: 'qwenlm',
+      name: 'QwenLM',
+      type: 'openai',
+      apiKey: '',
+      apiHost: 'https://chat.qwenlm.ai/api/',
+      models: SYSTEM_MODELS.qwenlm,
+      isSystem: true,
+      enabled: false
+    })
+    return state
   }
 }