feat: new build-in OCR provider -> intel OV(NPU) OCR (#10737)

* new build-in ocr provider intel ov

Signed-off-by: Ma, Kejiang <kj.ma@intel.com>
Signed-off-by: Kejiang Ma <kj.ma@intel.com>

* updated base on PR's commnets

Signed-off-by: Kejiang Ma <kj.ma@intel.com>

* feat(OcrImageSettings): use swr to fetch available providers

Add loading state and error handling when fetching available OCR providers. Display an alert when provider loading fails, showing the error message. Also optimize provider filtering logic using useMemo.

* refactor(ocr): rename providers to listProviders for consistency

Update method name to better reflect its functionality and maintain naming consistency across the codebase

---------

Signed-off-by: Ma, Kejiang <kj.ma@intel.com>
Signed-off-by: Kejiang Ma <kj.ma@intel.com>
Co-authored-by: icarus <eurfelux@gmail.com>

packages/shared/IpcChannel.ts

@@ -337,6 +337,7 @@ export enum IpcChannel {
 
   // OCR
   OCR_ocr = 'ocr:ocr',
+  OCR_ListProviders = 'ocr:list-providers',
 
   // OVMS
   Ovms_AddModel = 'ovms:add-model',

src/main/ipc.ts

@@ -875,6 +875,7 @@ export function registerIpc(mainWindow: BrowserWindow, app: Electron.App) {
   ipcMain.handle(IpcChannel.OCR_ocr, (_, file: SupportedOcrFile, provider: OcrProvider) =>
     ocrService.ocr(file, provider)
   )
+  ipcMain.handle(IpcChannel.OCR_ListProviders, () => ocrService.listProviderIds())
 
   // OVMS
   ipcMain.handle(IpcChannel.Ovms_AddModel, (_, modelName: string, modelId: string, modelSource: string, task: string) =>

src/main/services/ocr/OcrService.ts

@@ -2,6 +2,7 @@ import { loggerService } from '@logger'
 import { isLinux } from '@main/constant'
 import { BuiltinOcrProviderIds, OcrHandler, OcrProvider, OcrResult, SupportedOcrFile } from '@types'
 
+import { ovOcrService } from './builtin/OvOcrService'
 import { ppocrService } from './builtin/PpocrService'
 import { systemOcrService } from './builtin/SystemOcrService'
 import { tesseractService } from './builtin/TesseractService'
@@ -22,6 +23,10 @@ export class OcrService {
     this.registry.delete(providerId)
   }
 
+  public listProviderIds(): string[] {
+    return Array.from(this.registry.keys())
+  }
+
   public async ocr(file: SupportedOcrFile, provider: OcrProvider): Promise<OcrResult> {
     const handler = this.registry.get(provider.id)
     if (!handler) {
@@ -39,3 +44,5 @@ ocrService.register(BuiltinOcrProviderIds.tesseract, tesseractService.ocr.bind(t
 !isLinux && ocrService.register(BuiltinOcrProviderIds.system, systemOcrService.ocr.bind(systemOcrService))
 
 ocrService.register(BuiltinOcrProviderIds.paddleocr, ppocrService.ocr.bind(ppocrService))
+
+ovOcrService.isAvailable() && ocrService.register(BuiltinOcrProviderIds.ovocr, ovOcrService.ocr.bind(ovOcrService))

src/main/services/ocr/builtin/OvOcrService.ts

@@ -0,0 +1,128 @@
+import { loggerService } from '@logger'
+import { isWin } from '@main/constant'
+import { isImageFileMetadata, OcrOvConfig, OcrResult, SupportedOcrFile } from '@types'
+import { exec } from 'child_process'
+import * as fs from 'fs'
+import * as os from 'os'
+import * as path from 'path'
+import { promisify } from 'util'
+
+import { OcrBaseService } from './OcrBaseService'
+
+const logger = loggerService.withContext('OvOcrService')
+const execAsync = promisify(exec)
+
+const PATH_BAT_FILE = path.join(os.homedir(), '.cherrystudio', 'ovms', 'ovocr', 'run.npu.bat')
+
+export class OvOcrService extends OcrBaseService {
+  constructor() {
+    super()
+  }
+
+  public isAvailable(): boolean {
+    return (
+      isWin &&
+      os.cpus()[0].model.toLowerCase().includes('intel') &&
+      os.cpus()[0].model.toLowerCase().includes('ultra') &&
+      fs.existsSync(PATH_BAT_FILE)
+    )
+  }
+
+  private getOvOcrPath(): string {
+    return path.join(os.homedir(), '.cherrystudio', 'ovms', 'ovocr')
+  }
+
+  private getImgDir(): string {
+    return path.join(this.getOvOcrPath(), 'img')
+  }
+
+  private getOutputDir(): string {
+    return path.join(this.getOvOcrPath(), 'output')
+  }
+
+  private async clearDirectory(dirPath: string): Promise<void> {
+    if (fs.existsSync(dirPath)) {
+      const files = await fs.promises.readdir(dirPath)
+      for (const file of files) {
+        const filePath = path.join(dirPath, file)
+        const stats = await fs.promises.stat(filePath)
+        if (stats.isDirectory()) {
+          await this.clearDirectory(filePath)
+          await fs.promises.rmdir(filePath)
+        } else {
+          await fs.promises.unlink(filePath)
+        }
+      }
+    } else {
+      // If the directory does not exist, create it
+      await fs.promises.mkdir(dirPath, { recursive: true })
+    }
+  }
+
+  private async copyFileToImgDir(sourceFilePath: string, targetFileName: string): Promise<void> {
+    const imgDir = this.getImgDir()
+    const targetFilePath = path.join(imgDir, targetFileName)
+    await fs.promises.copyFile(sourceFilePath, targetFilePath)
+  }
+
+  private async runOcrBatch(): Promise<void> {
+    const ovOcrPath = this.getOvOcrPath()
+
+    try {
+      // Execute run.bat in the ov-ocr directory
+      await execAsync(`"${PATH_BAT_FILE}"`, {
+        cwd: ovOcrPath,
+        timeout: 60000 // 60 second timeout
+      })
+    } catch (error) {
+      logger.error(`Error running ovocr batch: ${error}`)
+      throw new Error(`Failed to run OCR batch: ${error}`)
+    }
+  }
+
+  private async ocrImage(filePath: string, options?: OcrOvConfig): Promise<OcrResult> {
+    logger.info(`OV OCR called on ${filePath} with options ${JSON.stringify(options)}`)
+
+    try {
+      // 1. Clear img directory and output directory
+      await this.clearDirectory(this.getImgDir())
+      await this.clearDirectory(this.getOutputDir())
+
+      // 2. Copy file to img directory
+      const fileName = path.basename(filePath)
+      await this.copyFileToImgDir(filePath, fileName)
+      logger.info(`File copied to img directory: ${fileName}`)
+
+      // 3. Run run.bat
+      logger.info('Running OV OCR batch process...')
+      await this.runOcrBatch()
+
+      // 4. Check that output/[basename].txt file exists
+      const baseNameWithoutExt = path.basename(fileName, path.extname(fileName))
+      const outputFilePath = path.join(this.getOutputDir(), `${baseNameWithoutExt}.txt`)
+      if (!fs.existsSync(outputFilePath)) {
+        throw new Error(`OV OCR output file not found at: ${outputFilePath}`)
+      }
+
+      // 5. Read output/[basename].txt file content
+      const ocrText = await fs.promises.readFile(outputFilePath, 'utf-8')
+      logger.info(`OV OCR text extracted: ${ocrText.substring(0, 100)}...`)
+
+      // 6. Return result
+      return { text: ocrText }
+    } catch (error) {
+      logger.error(`Error during OV OCR process: ${error}`)
+      throw error
+    }
+  }
+
+  public ocr = async (file: SupportedOcrFile, options?: OcrOvConfig): Promise<OcrResult> => {
+    if (isImageFileMetadata(file)) {
+      return this.ocrImage(file.path, options)
+    } else {
+      throw new Error('Unsupported file type, currently only image files are supported')
+    }
+  }
+}
+
+export const ovOcrService = new OvOcrService()

src/preload/index.ts

@@ -480,7 +480,8 @@ const api = {
   },
   ocr: {
     ocr: (file: SupportedOcrFile, provider: OcrProvider): Promise<OcrResult> =>
-      ipcRenderer.invoke(IpcChannel.OCR_ocr, file, provider)
+      ipcRenderer.invoke(IpcChannel.OCR_ocr, file, provider),
+    listProviders: (): Promise<string[]> => ipcRenderer.invoke(IpcChannel.OCR_ListProviders)
   },
   cherryai: {
     generateSignature: (params: { method: string; path: string; query: string; body: Record<string, any> }) =>

src/renderer/src/config/ocr.ts

@@ -1,6 +1,7 @@
 import {
   BuiltinOcrProvider,
   BuiltinOcrProviderId,
+  OcrOvProvider,
   OcrPpocrProvider,
   OcrProviderCapability,
   OcrSystemProvider,
@@ -50,10 +51,23 @@ const ppocrOcr: OcrPpocrProvider = {
   }
 } as const
 
+const ovOcr: OcrOvProvider = {
+  id: 'ovocr',
+  name: 'Intel OV(NPU) OCR',
+  config: {
+    langs: isWin ? ['en-us', 'zh-cn'] : undefined
+  },
+  capabilities: {
+    image: true
+    // pdf: true
+  }
+} as const satisfies OcrOvProvider
+
 export const BUILTIN_OCR_PROVIDERS_MAP = {
   tesseract,
   system: systemOcr,
-  paddleocr: ppocrOcr
+  paddleocr: ppocrOcr,
+  ovocr: ovOcr
 } as const satisfies Record<BuiltinOcrProviderId, BuiltinOcrProvider>
 
 export const BUILTIN_OCR_PROVIDERS: BuiltinOcrProvider[] = Object.values(BUILTIN_OCR_PROVIDERS_MAP)

src/renderer/src/hooks/useOcrProvider.tsx

@@ -1,4 +1,5 @@
 import { loggerService } from '@logger'
+import IntelLogo from '@renderer/assets/images/providers/intel.png'
 import PaddleocrLogo from '@renderer/assets/images/providers/paddleocr.png'
 import TesseractLogo from '@renderer/assets/images/providers/Tesseract.js.png'
 import { BUILTIN_OCR_PROVIDERS_MAP, DEFAULT_OCR_PROVIDER } from '@renderer/config/ocr'
@@ -83,6 +84,8 @@ export const useOcrProviders = () => {
           return <MonitorIcon size={size} />
         case 'paddleocr':
           return <Avatar size={size} src={PaddleocrLogo} />
+        case 'ovocr':
+          return <Avatar size={size} src={IntelLogo} />
       }
     }
     return <FileQuestionMarkIcon size={size} />

src/renderer/src/i18n/label.ts

@@ -340,12 +340,14 @@ export const getBuiltInMcpServerDescriptionLabel = (key: string): string => {
 const builtinOcrProviderKeyMap = {
   system: 'ocr.builtin.system',
   tesseract: '',
-  paddleocr: ''
+  paddleocr: '',
+  ovocr: ''
 } as const satisfies Record<BuiltinOcrProviderId, string>
 
 export const getBuiltinOcrProviderLabel = (key: BuiltinOcrProviderId) => {
   if (key === 'tesseract') return 'Tesseract'
   else if (key == 'paddleocr') return 'PaddleOCR'
+  else if (key == 'ovocr') return 'Intel OV(NPU) OCR'
   else return getLabel(builtinOcrProviderKeyMap, key)
 }
 

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

@@ -2049,6 +2049,7 @@
       "provider": {
         "cannot_remove_builtin": "Cannot delete built-in provider",
         "existing": "The provider already exists",
+        "get_providers": "Failed to get available providers",
         "not_found": "OCR provider does not exist",
         "update_failed": "Failed to update configuration"
       },

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

@@ -2049,6 +2049,7 @@
       "provider": {
         "cannot_remove_builtin": "不能删除内置提供商",
         "existing": "提供商已存在",
+        "get_providers": "获取可用提供商失败",
         "not_found": "OCR 提供商不存在",
         "update_failed": "更新配置失败"
       },

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

@@ -2048,8 +2048,9 @@
     "error": {
       "provider": {
         "cannot_remove_builtin": "不能刪除內建提供者",
-        "existing": "提供商已存在",
+        "existing": "提供者已存在",
-        "not_found": "OCR 提供商不存在",
+        "get_providers": "取得可用提供者失敗",
+        "not_found": "OCR 提供者不存在",
         "update_failed": "更新配置失敗"
       },
       "unknown": "OCR過程發生錯誤"

src/renderer/src/i18n/translate/el-gr.json

@@ -2041,6 +2041,7 @@
       "provider": {
         "cannot_remove_builtin": "Δεν είναι δυνατή η διαγραφή του ενσωματωμένου παρόχου",
         "existing": "Ο πάροχος υπηρεσιών υπάρχει ήδη",
+        "get_providers": "Αποτυχία λήψης διαθέσιμων παρόχων",
         "not_found": "Ο πάροχος OCR δεν υπάρχει",
         "update_failed": "Αποτυχία ενημέρωσης της διαμόρφωσης"
       },

src/renderer/src/i18n/translate/es-es.json

@@ -2041,6 +2041,7 @@
       "provider": {
         "cannot_remove_builtin": "No se puede eliminar el proveedor integrado",
         "existing": "El proveedor ya existe",
+        "get_providers": "Error al obtener proveedores disponibles",
         "not_found": "El proveedor de OCR no existe",
         "update_failed": "Actualización de la configuración fallida"
       },

src/renderer/src/i18n/translate/fr-fr.json

@@ -2041,6 +2041,7 @@
       "provider": {
         "cannot_remove_builtin": "Impossible de supprimer le fournisseur intégré",
         "existing": "Le fournisseur existe déjà",
+        "get_providers": "Échec de l'obtention des fournisseurs disponibles",
         "not_found": "Le fournisseur OCR n'existe pas",
         "update_failed": "Échec de la mise à jour de la configuration"
       },

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

@@ -2041,6 +2041,7 @@
       "provider": {
         "cannot_remove_builtin": "組み込みプロバイダーは削除できません",
         "existing": "プロバイダーはすでに存在します",
+        "get_providers": "利用可能なプロバイダーの取得に失敗しました",
         "not_found": "OCRプロバイダーが存在しません",
         "update_failed": "更新構成に失敗しました"
       },

src/renderer/src/i18n/translate/pt-pt.json

@@ -2041,6 +2041,7 @@
       "provider": {
         "cannot_remove_builtin": "Não é possível excluir o provedor integrado",
         "existing": "O provedor já existe",
+        "get_providers": "Falha ao obter provedores disponíveis",
         "not_found": "O provedor OCR não existe",
         "update_failed": "Falha ao atualizar a configuração"
       },

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

@@ -2041,6 +2041,7 @@
       "provider": {
         "cannot_remove_builtin": "Не удается удалить встроенного поставщика",
         "existing": "Поставщик уже существует",
+        "get_providers": "Не удалось получить доступных поставщиков",
         "not_found": "Поставщик OCR отсутствует",
         "update_failed": "Обновление конфигурации не удалось"
       },

src/renderer/src/pages/settings/DocProcessSettings/OcrImageSettings.tsx

@@ -1,11 +1,14 @@
+import { Alert, Skeleton } from '@heroui/react'
 import { loggerService } from '@logger'
 import { ErrorTag } from '@renderer/components/Tags/ErrorTag'
 import { isMac, isWin } from '@renderer/config/constant'
 import { useOcrProviders } from '@renderer/hooks/useOcrProvider'
 import { BuiltinOcrProviderIds, ImageOcrProvider, isImageOcrProvider, OcrProvider } from '@renderer/types'
+import { getErrorMessage } from '@renderer/utils'
 import { Select } from 'antd'
-import { useEffect, useMemo } from 'react'
+import { useCallback, useEffect, useMemo } from 'react'
 import { useTranslation } from 'react-i18next'
+import useSWRImmutable from 'swr/immutable'
 
 import { SettingRow, SettingRowTitle } from '..'
 
@@ -18,10 +21,16 @@ type Props = {
 const OcrImageSettings = ({ setProvider }: Props) => {
   const { t } = useTranslation()
   const { providers, imageProvider, getOcrProviderName, setImageProviderId } = useOcrProviders()
+  const fetcher = useCallback(() => {
+    return window.api.ocr.listProviders()
+  }, [])
+
+  const { data: validProviders, isLoading, error } = useSWRImmutable('ocr/providers', fetcher)
 
   const imageProviders = providers.filter((p) => isImageOcrProvider(p))
 
   // 挂载时更新外部状态
+  // FIXME: Just keep the imageProvider always valid, so we don't need update it in this component.
   useEffect(() => {
     setProvider(imageProvider)
   }, [imageProvider, setProvider])
@@ -40,12 +49,17 @@ const OcrImageSettings = ({ setProvider }: Props) => {
 
   const platformSupport = isMac || isWin
   const options = useMemo(() => {
+    if (!validProviders) return []
     const platformFilter = platformSupport ? () => true : (p: ImageOcrProvider) => p.id !== BuiltinOcrProviderIds.system
-    return imageProviders.filter(platformFilter).map((p) => ({
+    const validFilter = (p: ImageOcrProvider) => validProviders.includes(p.id)
-      value: p.id,
+    return imageProviders
-      label: getOcrProviderName(p)
+      .filter(platformFilter)
-    }))
+      .filter(validFilter)
-  }, [getOcrProviderName, imageProviders, platformSupport])
+      .map((p) => ({
+        value: p.id,
+        label: getOcrProviderName(p)
+      }))
+  }, [getOcrProviderName, imageProviders, platformSupport, validProviders])
 
   const isSystem = imageProvider.id === BuiltinOcrProviderIds.system
 
@@ -55,12 +69,23 @@ const OcrImageSettings = ({ setProvider }: Props) => {
         <SettingRowTitle>{t('settings.tool.ocr.image_provider')}</SettingRowTitle>
         <div style={{ display: 'flex', gap: '8px', alignItems: 'center' }}>
           {!platformSupport && isSystem && <ErrorTag message={t('settings.tool.ocr.error.not_system')} />}
-          <Select
+          <Skeleton isLoaded={!isLoading}>
-            value={imageProvider.id}
+            {!error && (
-            style={{ width: '200px' }}
+              <Select
-            onChange={(id: string) => setImageProvider(id)}
+                value={imageProvider.id}
-            options={options}
+                style={{ width: '200px' }}
-          />
+                onChange={(id: string) => setImageProvider(id)}
+                options={options}
+              />
+            )}
+            {error && (
+              <Alert
+                color="danger"
+                title={t('ocr.error.provider.get_providers')}
+                description={getErrorMessage(error)}
+              />
+            )}
+          </Skeleton>
         </div>
       </SettingRow>
     </>

src/renderer/src/pages/settings/DocProcessSettings/OcrOVSettings.tsx

@@ -0,0 +1,32 @@
+import { useOcrProvider } from '@renderer/hooks/useOcrProvider'
+import { BuiltinOcrProviderIds, isOcrOVProvider } from '@renderer/types'
+import { Flex, Tag } from 'antd'
+import { useTranslation } from 'react-i18next'
+
+import { SettingRow, SettingRowTitle } from '..'
+
+export const OcrOVSettings = () => {
+  const { t } = useTranslation()
+  const { provider } = useOcrProvider(BuiltinOcrProviderIds.ovocr)
+
+  if (!isOcrOVProvider(provider)) {
+    throw new Error('Not OV OCR provider.')
+  }
+
+  return (
+    <>
+      <SettingRow>
+        <SettingRowTitle>
+          <Flex align="center" gap={4}>
+            {t('settings.tool.ocr.common.langs')}
+          </Flex>
+        </SettingRowTitle>
+        <div style={{ display: 'flex', gap: '8px' }}>
+          <Tag>🇬🇧 {t('languages.english')}</Tag>
+          <Tag>🇨🇳 {t('languages.chinese')}</Tag>
+          <Tag>🇭🇰 {t('languages.chinese-traditional')}</Tag>
+        </div>
+      </SettingRow>
+    </>
+  )
+}

src/renderer/src/pages/settings/DocProcessSettings/OcrProviderSettings.tsx

@@ -8,6 +8,7 @@ import { Divider, Flex } from 'antd'
 import styled from 'styled-components'
 
 import { SettingGroup, SettingTitle } from '..'
+import { OcrOVSettings } from './OcrOVSettings'
 import { OcrPpocrSettings } from './OcrPpocrSettings'
 import { OcrSystemSettings } from './OcrSystemSettings'
 import { OcrTesseractSettings } from './OcrTesseractSettings'
@@ -35,6 +36,8 @@ const OcrProviderSettings = ({ provider }: Props) => {
           return <OcrSystemSettings />
         case 'paddleocr':
           return <OcrPpocrSettings />
+        case 'ovocr':
+          return <OcrOVSettings />
         default:
           return null
       }

src/renderer/src/store/migrate.ts

@@ -2667,6 +2667,15 @@ const migrateConfig = {
       logger.error('migrate 162 error', error as Error)
       return state
     }
+  },
+  '163': (state: RootState) => {
+    try {
+      addOcrProvider(state, BUILTIN_OCR_PROVIDERS_MAP.ovocr)
+      return state
+    } catch (error) {
+      logger.error('migrate 163 error', error as Error)
+      return state
+    }
   }
 }
 

src/renderer/src/types/ocr.ts

@@ -5,7 +5,8 @@ import { FileMetadata, ImageFileMetadata, isImageFileMetadata, TranslateLanguage
 export const BuiltinOcrProviderIds = {
   tesseract: 'tesseract',
   system: 'system',
-  paddleocr: 'paddleocr'
+  paddleocr: 'paddleocr',
+  ovocr: 'ovocr'
 } as const
 
 export type BuiltinOcrProviderId = keyof typeof BuiltinOcrProviderIds
@@ -188,3 +189,19 @@ export type OcrPpocrProvider = {
 export const isOcrPpocrProvider = (p: OcrProvider): p is OcrPpocrProvider => {
   return p.id === BuiltinOcrProviderIds.paddleocr
 }
+
+// OV OCR Types
+export type OcrOvConfig = OcrProviderBaseConfig & {
+  langs?: TranslateLanguageCode[]
+}
+
+export type OcrOvProvider = {
+  id: 'ovocr'
+  config: OcrOvConfig
+} & ImageOcrProvider &
+  // PdfOcrProvider &
+  BuiltinOcrProvider
+
+export const isOcrOVProvider = (p: OcrProvider): p is OcrOvProvider => {
+  return p.id === BuiltinOcrProviderIds.ovocr
+}