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

src/main/services/ocr/OcrService.ts

@@ -1,10 +1,8 @@
 import { loggerService } from '@logger'
-import { BuiltinOcrProviderIds, OcrProvider, OcrResult, SupportedOcrFile } from '@types'
+import { BuiltinOcrProviderIds, OcrHandler, OcrProvider, OcrResult, SupportedOcrFile } from '@types'
 
 import { tesseractService } from './tesseract/TesseractService'
 
-type OcrHandler = (file: SupportedOcrFile) => Promise<OcrResult>
-
 const logger = loggerService.withContext('OcrService')
 
 export class OcrService {

src/renderer/src/hooks/useOcr.ts

@@ -1,5 +1,5 @@
 import { loggerService } from '@logger'
-import * as OcrService from '@renderer/services/OcrService'
+import * as OcrService from '@renderer/services/ocr/OcrService'
 import { useAppSelector } from '@renderer/store'
 import { ImageFileMetadata, isImageFile, SupportedOcrFile } from '@renderer/types'
 import { uuid } from '@renderer/utils'

src/renderer/src/services/OcrService.ts

@@ -1,14 +0,0 @@
-import { OcrProvider, OcrResult, SupportedOcrFile } from '@renderer/types'
-
-// const logger = loggerService.withContext('renderer:OcrService')
-
-/**
- * ocr a file
- * @param file any supported file
- * @param provider ocr provider
- * @returns ocr result
- * @throws {Error}
- */
-export const ocr = async (file: SupportedOcrFile, provider: OcrProvider): Promise<OcrResult> => {
-  return window.api.ocr.ocr(file, provider)
-}

src/renderer/src/services/ocr/OcrService.ts

@@ -0,0 +1,23 @@
+import { loggerService } from '@logger'
+import { isOcrApiProvider, OcrProvider, OcrResult, SupportedOcrFile } from '@renderer/types'
+
+import { OcrApiClientFactory } from './clients/OcrApiClientFactory'
+
+const logger = loggerService.withContext('renderer:OcrService')
+
+/**
+ * ocr a file
+ * @param file any supported file
+ * @param provider ocr provider
+ * @returns ocr result
+ * @throws {Error}
+ */
+export const ocr = async (file: SupportedOcrFile, provider: OcrProvider): Promise<OcrResult> => {
+  logger.info(`ocr file ${file.path}`)
+  if (isOcrApiProvider(provider)) {
+    const client = OcrApiClientFactory.create(provider)
+    return client.ocr(file)
+  } else {
+    return window.api.ocr.ocr(file, provider)
+  }
+}

src/renderer/src/services/ocr/clients/OcrApiClientFactory.ts

@@ -0,0 +1,28 @@
+import { loggerService } from '@logger'
+import { OcrApiProvider } from '@renderer/types'
+
+import { OcrBaseApiClient } from './OcrBaseApiClient'
+import { OcrExampleApiClient } from './OcrExampleApiClient'
+
+const logger = loggerService.withContext('OcrApiClientFactory')
+
+export class OcrApiClientFactory {
+  /**
+   * Create an ApiClient instance for the given provider
+   * 为给定的提供者创建ApiClient实例
+   */
+  static create(provider: OcrApiProvider): OcrBaseApiClient {
+    logger.debug(`Creating ApiClient for provider:`, {
+      id: provider.id,
+      config: provider.config
+    })
+
+    let instance: OcrBaseApiClient
+
+    // Extend other clients here
+    // eslint-disable-next-line prefer-const
+    instance = new OcrExampleApiClient(provider)
+
+    return instance
+  }
+}

src/renderer/src/services/ocr/clients/OcrBaseApiClient.ts

@@ -0,0 +1,43 @@
+import { OcrApiProvider, OcrHandler } from '@renderer/types'
+
+export abstract class OcrBaseApiClient {
+  public provider: OcrApiProvider
+  protected host: string
+  protected apiKey: string
+
+  constructor(provider: OcrApiProvider) {
+    this.provider = provider
+    this.host = this.getHost()
+    this.apiKey = this.getApiKey()
+  }
+
+  abstract ocr: OcrHandler
+
+  // copy from BaseApiClient
+  public getHost(): string {
+    return this.provider.config.api.apiHost
+  }
+
+  // copy from BaseApiClient
+  public getApiKey() {
+    const keys = this.provider.config.api.apiKey.split(',').map((key) => key.trim())
+    const keyName = `ocr_provider:${this.provider.id}:last_used_key`
+
+    if (keys.length === 1) {
+      return keys[0]
+    }
+
+    const lastUsedKey = window.keyv.get(keyName)
+    if (!lastUsedKey) {
+      window.keyv.set(keyName, keys[0])
+      return keys[0]
+    }
+
+    const currentIndex = keys.indexOf(lastUsedKey)
+    const nextIndex = (currentIndex + 1) % keys.length
+    const nextKey = keys[nextIndex]
+    window.keyv.set(keyName, nextKey)
+
+    return nextKey
+  }
+}

src/renderer/src/services/ocr/clients/OcrExampleApiClient.ts

@@ -0,0 +1,15 @@
+import { OcrApiProvider, SupportedOcrFile } from '@renderer/types'
+
+import { OcrBaseApiClient } from './OcrBaseApiClient'
+
+export type OcrExampleProvider = OcrApiProvider
+
+export class OcrExampleApiClient extends OcrBaseApiClient {
+  constructor(provider: OcrApiProvider) {
+    super(provider)
+  }
+
+  public ocr = async (file: SupportedOcrFile) => {
+    return { text: `Example output: ${file.path}` }
+  }
+}

src/renderer/src/types/ocr.ts

@@ -1,4 +1,4 @@
-import { FileMetadata, ImageFileMetadata, isImageFile, Model } from '.'
+import { FileMetadata, ImageFileMetadata, isImageFile } from '.'
 
 export const BuiltinOcrProviderIds = {
   tesseract: 'tesseract'
@@ -23,22 +23,70 @@ export const isOcrProviderCapability = (cap: string): cap is OcrProviderCapabili
 
 export type OcrProviderCapabilityRecord = Partial<Record<OcrProviderCapability, boolean>>
 
+// OCR models and providers share the same type definition.
+// A provider can offer capabilities to process multiple file types,
+// while a model belonging to that provider may be limited to processing only one specific file type.
+export type OcrModelCapabilityRecord = OcrProviderCapabilityRecord
+
+export interface OcrModel {
+  id: string
+  name: string
+  providerId: string
+  capabilities: OcrModelCapabilityRecord
+}
+
+/**
+ * Extend this type to define provider-specefic config types.
+ */
+export type OcrProviderApiConfig = {
+  apiKey: string
+  apiHost: string
+  apiVersion?: string
+}
+
+export const isOcrProviderApiConfig = (config: unknown): config is OcrProviderApiConfig => {
+  return (
+    typeof config === 'object' &&
+    config !== null &&
+    'apiKey' in config &&
+    typeof config.apiKey === 'string' &&
+    'apiHost' in config &&
+    typeof config.apiHost === 'string' &&
+    (!('apiVersion' in config) || typeof config.apiVersion === 'string')
+  )
+}
+
+/**
+ * For future. Model based ocr, api based ocr. May different api client.
+ *
+ * Extend this type to define provider-specific config types.
+ */
+export type OcrProviderConfig = {
+  /** Not used for now. Could safely remove. */
+  api?: OcrProviderApiConfig
+  /** Not used for now. Could safely remove. */
+  models?: OcrModel[]
+  /** Not used for now. Could safely remove. */
+  enabled?: boolean
+}
+
 export type OcrProvider = {
   id: string
   name: string
   capabilities: OcrProviderCapabilityRecord
-  config?: {
-    // for future. Model based ocr, api based ocr. May different api client.
-    api?: {
-      apiKey: string
-      apiHost: string
-      apiVersion?: string
-    }
-    models?: Model[]
-    enabled?: boolean
+  config?: OcrProviderConfig
+}
+
+export type OcrApiProvider = OcrProvider & {
+  config: OcrProviderConfig & {
+    api: OcrProviderApiConfig
   }
 }
 
+export const isOcrApiProvider = (p: OcrProvider): p is OcrApiProvider => {
+  return !!(p.config && p.config.api && isOcrProviderApiConfig(p.config.api))
+}
+
 export type BuiltinOcrProvider = OcrProvider & {
   id: BuiltinOcrProviderId
 }
@@ -71,3 +119,7 @@ export const isSupportedOcrFile = (file: FileMetadata): file is SupportedOcrFile
 export type OcrResult = {
   text: string
 }
+
+export type OcrHandler = (file: SupportedOcrFile) => Promise<OcrResult>
+
+export type OcrImageHandler = (file: ImageFileMetadata) => Promise<OcrResult>