mirror of
https://github.com/CherryHQ/cherry-studio.git
synced 2025-12-19 14:41:24 +08:00
feat(ocr): enhance OCR service for Linux compatibility and update image processing
- Added conditional registration of the system OCR service for non-Linux platforms. - Updated SystemOcrService to handle Linux by returning an empty text response. - Modified image preprocessing to dynamically require the 'sharp' library, improving compatibility and performance. - Included additional files in the electron-builder configuration for packaging.
This commit is contained in:
parent
5ab98c9d05
commit
80f49aecd7
@ -62,6 +62,7 @@ files:
|
||||
asarUnpack:
|
||||
- resources/**
|
||||
- '**/*.{metal,exp,lib}'
|
||||
- 'node_modules/@img/sharp-libvips-*/**'
|
||||
win:
|
||||
executableName: Cherry Studio
|
||||
artifactName: ${productName}-${version}-${arch}-setup.${ext}
|
||||
|
||||
@ -1,7 +1,7 @@
|
||||
import { loggerService } from '@logger'
|
||||
import { isLinux } from '@main/constant'
|
||||
import { BuiltinOcrProviderIds, OcrHandler, OcrProvider, OcrResult, SupportedOcrFile } from '@types'
|
||||
|
||||
import { systemOcrService } from './builtin/SystemOcrService'
|
||||
import { tesseractService } from './builtin/TesseractService'
|
||||
|
||||
const logger = loggerService.withContext('OcrService')
|
||||
@ -33,4 +33,8 @@ export const ocrService = new OcrService()
|
||||
|
||||
// Register built-in providers
|
||||
ocrService.register(BuiltinOcrProviderIds.tesseract, tesseractService.ocr.bind(tesseractService))
|
||||
ocrService.register(BuiltinOcrProviderIds.system, systemOcrService.ocr.bind(systemOcrService))
|
||||
|
||||
if (!isLinux) {
|
||||
const { systemOcrService } = require('./builtin/SystemOcrService')
|
||||
ocrService.register(BuiltinOcrProviderIds.system, systemOcrService.ocr.bind(systemOcrService))
|
||||
}
|
||||
|
||||
@ -1,6 +1,5 @@
|
||||
import { isMac, isWin } from '@main/constant'
|
||||
import { isLinux, isWin } from '@main/constant'
|
||||
import { loadOcrImage } from '@main/utils/ocr'
|
||||
import { OcrAccuracy, recognize } from '@napi-rs/system-ocr'
|
||||
import {
|
||||
ImageFileMetadata,
|
||||
isImageFileMetadata as isImageFileMetadata,
|
||||
@ -15,12 +14,14 @@ import { OcrBaseService } from './OcrBaseService'
|
||||
export class SystemOcrService extends OcrBaseService {
|
||||
constructor() {
|
||||
super()
|
||||
if (!isWin && !isMac) {
|
||||
throw new Error('System OCR is only supported on Windows and macOS')
|
||||
}
|
||||
}
|
||||
|
||||
private async ocrImage(file: ImageFileMetadata, options?: OcrSystemConfig): Promise<OcrResult> {
|
||||
if (isLinux) {
|
||||
return { text: '' }
|
||||
}
|
||||
|
||||
const { OcrAccuracy, recognize } = require('@napi-rs/system-ocr')
|
||||
const buffer = await loadOcrImage(file)
|
||||
const langs = isWin ? options?.langs : undefined
|
||||
const result = await recognize(buffer, OcrAccuracy.Accurate, langs)
|
||||
|
||||
@ -1,8 +1,8 @@
|
||||
import { ImageFileMetadata } from '@types'
|
||||
import { readFile } from 'fs/promises'
|
||||
import sharp from 'sharp'
|
||||
|
||||
const preprocessImage = async (buffer: Buffer): Promise<Buffer> => {
|
||||
const sharp = require('sharp')
|
||||
return sharp(buffer)
|
||||
.grayscale() // 转为灰度
|
||||
.normalize()
|
||||
|
||||
Loading…
Reference in New Issue
Block a user