Merge branch 'feat/ocr' into feat/ocr-translate

packages/shared/config/constant.ts

@@ -197,12 +197,6 @@ export enum FeedUrl {
   GITHUB_LATEST = 'https://github.com/CherryHQ/cherry-studio/releases/latest/download'
 }
 
-export const tesseractLangs = ['chi_sim', 'chi_tra', 'eng']
-export enum TesseractLangsDownloadUrl {
-  CN = 'https://gitcode.com/beyondkmp/tessdata/releases/download/4.1.0/',
-  GLOBAL = 'https://github.com/tesseract-ocr/tessdata/raw/main/'
-}
-
 export enum UpgradeChannel {
   LATEST = 'latest', // 最新稳定版本
   RC = 'rc', // 公测版本

src/main/services/ocr/OcrService.ts

@@ -1,13 +1,19 @@
-import { BuiltinOcrProviderIds, FileMetadata, OcrProvider, OcrResult, SupportedOcrFile } from '@types'
+import { loggerService } from '@logger'
+import { BuiltinOcrProviderIds, OcrProvider, OcrResult, SupportedOcrFile } from '@types'
 
 import { tesseractService } from './tesseract/TesseractService'
 
-type OcrHandler = (file: FileMetadata) => Promise<OcrResult>
+type OcrHandler = (file: SupportedOcrFile) => Promise<OcrResult>
+
+const logger = loggerService.withContext('OcrService')
 
 export class OcrService {
   private registry: Map<string, OcrHandler> = new Map()
 
   register(providerId: string, handler: OcrHandler): void {
+    if (this.registry.has(providerId)) {
+      logger.warn(`Provider ${providerId} has existing handler. Overwrited.`)
+    }
     this.registry.set(providerId, handler)
   }
 

src/main/services/ocr/tesseract/TesseractService.ts

@@ -1,7 +1,7 @@
 import { loggerService } from '@logger'
 import { getIpCountry } from '@main/utils/ipService'
-import { MB, TesseractLangsDownloadUrl } from '@shared/config/constant'
+import { MB } from '@shared/config/constant'
-import { FileMetadata, ImageFileMetadata, isImageFile, OcrResult } from '@types'
+import { ImageFileMetadata, isImageFile, OcrResult, SupportedOcrFile } from '@types'
 import { app } from 'electron'
 import fs from 'fs'
 import path from 'path'
@@ -114,13 +114,21 @@ const logger = loggerService.withContext('TesseractService')
 //   'yi-us': 'yid'
 // }
 
+// config
+const MB_SIZE_THRESHOLD = 50
+const tesseractLangs = ['chi_sim', 'chi_tra', 'eng']
+enum TesseractLangsDownloadUrl {
+  CN = 'https://gitcode.com/beyondkmp/tessdata/releases/download/4.1.0/',
+  GLOBAL = 'https://github.com/tesseract-ocr/tessdata/raw/main/'
+}
+
 export class TesseractService {
   private worker: Tesseract.Worker | null = null
 
   async getWorker(): Promise<Tesseract.Worker> {
     if (!this.worker) {
       // for now, only support limited languages
-      this.worker = await createWorker(['chi_sim', 'chi_tra', 'eng'], undefined, {
+      this.worker = await createWorker(tesseractLangs, undefined, {
         langPath: await this._getLangPath(),
         cachePath: await this._getCacheDir(),
         gzip: false,
@@ -133,15 +141,15 @@ export class TesseractService {
   async imageOcr(file: ImageFileMetadata): Promise<OcrResult> {
     const worker = await this.getWorker()
     const stat = await fs.promises.stat(file.path)
-    if (stat.size > 50 * MB) {
+    if (stat.size > MB_SIZE_THRESHOLD * MB) {
-      throw new Error('This image is too large (max 50MB)')
+      throw new Error(`This image is too large (max ${MB_SIZE_THRESHOLD}MB)`)
     }
     const buffer = await fs.promises.readFile(file.path)
     const result = await worker.recognize(buffer)
     return { text: result.data.text }
   }
 
-  async ocr(file: FileMetadata): Promise<OcrResult> {
+  async ocr(file: SupportedOcrFile): Promise<OcrResult> {
     if (!isImageFile(file)) {
       throw new Error('Only image files are supported currently')
     }

src/renderer/src/types/ocr.ts

@@ -21,7 +21,7 @@ export const isOcrProviderCapability = (cap: string): cap is OcrProviderCapabili
   return Object.hasOwn(OcrProviderCapabilities, cap)
 }
 
-export type OcrProviderCapabilityRecord = Record<OcrProviderCapability, boolean>
+export type OcrProviderCapabilityRecord = Partial<Record<OcrProviderCapability, boolean>>
 
 export type OcrProvider = {
   id: string
@@ -59,7 +59,7 @@ export type ImageOcrProvider = OcrProvider & {
 }
 
 export const isImageOcrProvider = (p: OcrProvider): p is ImageOcrProvider => {
-  return p.capabilities.image
+  return p.capabilities.image === true
 }
 
 export type SupportedOcrFile = ImageFileMetadata