feat(lang): create frontend components to support multiple ocr languages

This commit is contained in:
perf3ct 2025-07-14 19:45:35 +00:00
parent 7317fd5ebb
commit bd3f35cf38
6 changed files with 469 additions and 25 deletions

View File

@ -3,6 +3,7 @@ import { useDropzone } from 'react-dropzone'
import { DocumentArrowUpIcon } from '@heroicons/react/24/outline'
import { Document, documentService } from '../services/api'
import { useNotifications } from '../contexts/NotificationContext'
import LanguageSelector from './LanguageSelector'
interface FileUploadProps {
onUploadSuccess: (document: Document) => void
@ -11,6 +12,8 @@ interface FileUploadProps {
function FileUpload({ onUploadSuccess }: FileUploadProps) {
const [uploading, setUploading] = useState(false)
const [error, setError] = useState<string | null>(null)
const [selectedLanguages, setSelectedLanguages] = useState<string[]>(['eng'])
const [primaryLanguage, setPrimaryLanguage] = useState<string>('eng')
const { addBatchNotification } = useNotifications()
const onDrop = useCallback(async (acceptedFiles: File[]) => {
@ -21,7 +24,7 @@ function FileUpload({ onUploadSuccess }: FileUploadProps) {
setError(null)
try {
const response = await documentService.upload(file)
const response = await documentService.upload(file, selectedLanguages.length > 0 ? selectedLanguages : undefined)
onUploadSuccess(response.data)
// Trigger success notification
@ -34,7 +37,16 @@ function FileUpload({ onUploadSuccess }: FileUploadProps) {
} finally {
setUploading(false)
}
}, [onUploadSuccess, addBatchNotification])
}, [onUploadSuccess, addBatchNotification, selectedLanguages])
const handleLanguagesChange = (languages: string[], primary?: string) => {
setSelectedLanguages(languages)
if (primary) {
setPrimaryLanguage(primary)
} else if (languages.length > 0) {
setPrimaryLanguage(languages[0])
}
}
const { getRootProps, getInputProps, isDragActive } = useDropzone({
onDrop,
@ -49,7 +61,16 @@ function FileUpload({ onUploadSuccess }: FileUploadProps) {
})
return (
<div className="w-full">
<div className="w-full space-y-4">
{/* Language Selector */}
<LanguageSelector
selectedLanguages={selectedLanguages}
primaryLanguage={primaryLanguage}
onLanguagesChange={handleLanguagesChange}
disabled={uploading}
/>
{/* File Upload Area */}
<div
{...getRootProps()}
className={`border-2 border-dashed rounded-lg p-6 text-center cursor-pointer transition-colors ${

View File

@ -0,0 +1,265 @@
import React, { useState, useEffect } from 'react'
import { CheckIcon, XMarkIcon } from '@heroicons/react/24/outline'
import { LanguageInfo } from '../services/api'
interface LanguageSelectorProps {
selectedLanguages: string[]
primaryLanguage?: string
onLanguagesChange: (languages: string[], primary?: string) => void
maxLanguages?: number
disabled?: boolean
showPrimarySelector?: boolean
className?: string
}
// Common languages with display names
const COMMON_LANGUAGES: LanguageInfo[] = [
{ code: 'eng', name: 'English', installed: true },
{ code: 'spa', name: 'Spanish', installed: true },
{ code: 'fra', name: 'French', installed: true },
{ code: 'deu', name: 'German', installed: true },
{ code: 'ita', name: 'Italian', installed: true },
{ code: 'por', name: 'Portuguese', installed: true },
{ code: 'rus', name: 'Russian', installed: true },
{ code: 'chi_sim', name: 'Chinese (Simplified)', installed: true },
{ code: 'chi_tra', name: 'Chinese (Traditional)', installed: true },
{ code: 'jpn', name: 'Japanese', installed: true },
{ code: 'kor', name: 'Korean', installed: true },
{ code: 'ara', name: 'Arabic', installed: true },
{ code: 'hin', name: 'Hindi', installed: true },
{ code: 'nld', name: 'Dutch', installed: true },
{ code: 'swe', name: 'Swedish', installed: true },
{ code: 'nor', name: 'Norwegian', installed: true },
{ code: 'dan', name: 'Danish', installed: true },
{ code: 'fin', name: 'Finnish', installed: true },
{ code: 'pol', name: 'Polish', installed: true },
{ code: 'ces', name: 'Czech', installed: true },
{ code: 'hun', name: 'Hungarian', installed: true },
{ code: 'tur', name: 'Turkish', installed: true },
{ code: 'tha', name: 'Thai', installed: true },
{ code: 'vie', name: 'Vietnamese', installed: true },
]
function LanguageSelector({
selectedLanguages,
primaryLanguage,
onLanguagesChange,
maxLanguages = 4,
disabled = false,
showPrimarySelector = true,
className = '',
}: LanguageSelectorProps) {
const [availableLanguages, setAvailableLanguages] = useState<LanguageInfo[]>(COMMON_LANGUAGES)
const [isOpen, setIsOpen] = useState(false)
// Auto-set primary language to first selected if not specified
const effectivePrimary = primaryLanguage || selectedLanguages[0] || ''
const handleLanguageToggle = (languageCode: string) => {
if (disabled) return
let newLanguages: string[]
let newPrimary = effectivePrimary
if (selectedLanguages.includes(languageCode)) {
// Remove language
newLanguages = selectedLanguages.filter(lang => lang !== languageCode)
// If removing the primary language, set new primary to first remaining language
if (languageCode === effectivePrimary && newLanguages.length > 0) {
newPrimary = newLanguages[0]
} else if (newLanguages.length === 0) {
newPrimary = ''
}
} else {
// Add language (check max limit)
if (selectedLanguages.length >= maxLanguages) {
return
}
newLanguages = [...selectedLanguages, languageCode]
// If this is the first language, make it primary
if (newLanguages.length === 1) {
newPrimary = languageCode
}
}
onLanguagesChange(newLanguages, newPrimary)
}
const handlePrimaryChange = (languageCode: string) => {
if (disabled || !selectedLanguages.includes(languageCode)) return
onLanguagesChange(selectedLanguages, languageCode)
}
const removeLanguage = (languageCode: string) => {
handleLanguageToggle(languageCode)
}
const getLanguageName = (code: string) => {
const language = availableLanguages.find(lang => lang.code === code)
return language?.name || code
}
return (
<div className={`relative ${className}`}>
{/* Selected Languages Display */}
<div className="mb-3">
<label className="block text-sm font-medium text-gray-700 mb-2">
OCR Languages {selectedLanguages.length > 0 && `(${selectedLanguages.length}/${maxLanguages})`}
</label>
{selectedLanguages.length > 0 ? (
<div className="flex flex-wrap gap-2">
{selectedLanguages.map((langCode) => (
<span
key={langCode}
className={`inline-flex items-center px-3 py-1 rounded-full text-sm font-medium ${
langCode === effectivePrimary
? 'bg-blue-100 text-blue-800 border-2 border-blue-300'
: 'bg-gray-100 text-gray-800 border border-gray-300'
}`}
>
{getLanguageName(langCode)}
{langCode === effectivePrimary && (
<span className="ml-1 text-xs font-bold text-blue-600">(Primary)</span>
)}
{!disabled && (
<button
type="button"
onClick={() => removeLanguage(langCode)}
className="ml-2 text-gray-400 hover:text-gray-600"
>
<XMarkIcon className="h-4 w-4" />
</button>
)}
</span>
))}
</div>
) : (
<div className="text-sm text-gray-500 italic">
No languages selected. Documents will use default OCR language.
</div>
)}
</div>
{/* Language Selector Button */}
{!disabled && (
<button
type="button"
onClick={() => setIsOpen(!isOpen)}
className="w-full px-4 py-2 text-left border border-gray-300 rounded-lg bg-white hover:bg-gray-50 focus:ring-2 focus:ring-blue-500 focus:border-transparent"
>
<span className="text-gray-600">
{selectedLanguages.length === 0
? 'Select OCR languages...'
: `Add more languages (${maxLanguages - selectedLanguages.length} remaining)`
}
</span>
</button>
)}
{/* Dropdown Panel */}
{isOpen && !disabled && (
<div className="absolute z-10 mt-1 w-full bg-white border border-gray-300 rounded-lg shadow-lg max-h-64 overflow-y-auto">
<div className="p-3">
<div className="text-xs text-gray-500 mb-2 uppercase tracking-wide font-semibold">
Available Languages
</div>
<div className="space-y-1">
{availableLanguages
.filter(lang => lang.installed)
.map((language) => {
const isSelected = selectedLanguages.includes(language.code)
const isPrimary = language.code === effectivePrimary
const canSelect = !isSelected && selectedLanguages.length < maxLanguages
return (
<div
key={language.code}
className={`flex items-center justify-between p-2 rounded ${
isSelected
? 'bg-blue-50 border border-blue-200'
: canSelect
? 'hover:bg-gray-50 cursor-pointer'
: 'opacity-50 cursor-not-allowed'
}`}
>
<div className="flex items-center">
<button
type="button"
onClick={() => handleLanguageToggle(language.code)}
disabled={!canSelect && !isSelected}
className={`flex items-center space-x-2 ${
canSelect || isSelected ? 'cursor-pointer' : 'cursor-not-allowed'
}`}
>
<div className={`w-5 h-5 border-2 rounded flex items-center justify-center ${
isSelected
? 'border-blue-500 bg-blue-500'
: 'border-gray-300'
}`}>
{isSelected && <CheckIcon className="h-3 w-3 text-white" />}
</div>
<span className={`text-sm ${isSelected ? 'font-medium text-blue-900' : 'text-gray-700'}`}>
{language.name}
</span>
{isPrimary && (
<span className="text-xs bg-blue-600 text-white px-2 py-0.5 rounded font-bold">
PRIMARY
</span>
)}
</button>
</div>
{/* Primary selector */}
{isSelected && showPrimarySelector && selectedLanguages.length > 1 && (
<button
type="button"
onClick={() => handlePrimaryChange(language.code)}
className={`text-xs px-2 py-1 rounded font-medium ${
isPrimary
? 'bg-blue-600 text-white cursor-default'
: 'bg-gray-200 text-gray-700 hover:bg-gray-300'
}`}
>
{isPrimary ? 'Primary' : 'Set Primary'}
</button>
)}
</div>
)
})}
</div>
{selectedLanguages.length >= maxLanguages && (
<div className="mt-3 p-2 bg-amber-50 border border-amber-200 rounded text-xs text-amber-800">
Maximum {maxLanguages} languages allowed for optimal performance.
</div>
)}
</div>
<div className="border-t border-gray-200 p-3">
<button
type="button"
onClick={() => setIsOpen(false)}
className="w-full text-center text-sm text-gray-600 hover:text-gray-800"
>
Close
</button>
</div>
</div>
)}
{/* Help Text */}
{selectedLanguages.length > 1 && (
<div className="mt-2 text-xs text-gray-500">
<p>
<strong>Primary language</strong> is processed first for better accuracy.
Multiple languages help with mixed-language documents.
</p>
</div>
)}
</div>
)
}
export default LanguageSelector

View File

@ -13,6 +13,7 @@ import {
} from '@mui/material';
import { Refresh as RefreshIcon, Language as LanguageIcon } from '@mui/icons-material';
import OcrLanguageSelector from '../OcrLanguageSelector';
import LanguageSelector from '../LanguageSelector';
import { ocrService } from '../../services/api';
interface OcrRetryDialogProps {
@ -38,8 +39,16 @@ const OcrRetryDialog: React.FC<OcrRetryDialogProps> = ({
onRetryError,
}) => {
const [selectedLanguage, setSelectedLanguage] = useState<string>('');
const [selectedLanguages, setSelectedLanguages] = useState<string[]>([]);
const [primaryLanguage, setPrimaryLanguage] = useState<string>('');
const [useMultiLanguage, setUseMultiLanguage] = useState<boolean>(false);
const [retrying, setRetrying] = useState<boolean>(false);
const handleLanguagesChange = (languages: string[], primary?: string) => {
setSelectedLanguages(languages);
setPrimaryLanguage(primary || languages[0] || '');
};
// Simple language code to name mapping for display
const getLanguageDisplayName = (langCode: string): string => {
const languageNames: Record<string, string> = {
@ -95,15 +104,32 @@ const OcrRetryDialog: React.FC<OcrRetryDialogProps> = ({
try {
setRetrying(true);
// Use multi-language if enabled and languages are selected, otherwise use single language
const languagesToUse = useMultiLanguage && selectedLanguages.length > 0
? selectedLanguages
: undefined;
const singleLanguageToUse = !useMultiLanguage && selectedLanguage
? selectedLanguage
: undefined;
const response = await ocrService.retryWithLanguage(
document.id,
selectedLanguage || undefined
singleLanguageToUse,
languagesToUse
);
if (response.data.success) {
const waitTime = response.data.estimated_wait_minutes || 'Unknown';
const languageInfo = selectedLanguage ?
` with language "${getLanguageDisplayName(selectedLanguage)}"` : '';
let languageInfo = '';
if (languagesToUse && languagesToUse.length > 0) {
const langNames = languagesToUse.map(lang => getLanguageDisplayName(lang));
languageInfo = ` with languages: ${langNames.join(', ')} (Primary: ${getLanguageDisplayName(primaryLanguage)})`;
} else if (singleLanguageToUse) {
languageInfo = ` with language "${getLanguageDisplayName(singleLanguageToUse)}"`;
}
onRetrySuccess(
`OCR retry queued for "${document.filename}"${languageInfo}. Estimated wait time: ${waitTime} minutes.`
);
@ -124,6 +150,9 @@ const OcrRetryDialog: React.FC<OcrRetryDialogProps> = ({
const handleClose = () => {
if (!retrying) {
setSelectedLanguage('');
setSelectedLanguages([]);
setPrimaryLanguage('');
setUseMultiLanguage(false);
onClose();
}
};
@ -170,16 +199,51 @@ const OcrRetryDialog: React.FC<OcrRetryDialogProps> = ({
OCR Language Selection
</Typography>
<Typography variant="body2" color="text.secondary" sx={{ mb: 2 }}>
Choose a different language if the previous OCR attempt used the wrong language for this document.
Choose a different language or language combination if the previous OCR attempt failed due to incorrect language settings.
</Typography>
<OcrLanguageSelector
value={selectedLanguage}
onChange={setSelectedLanguage}
label="OCR Language (Optional)"
size="medium"
helperText="Leave empty to use your default language setting"
showCurrentIndicator={true}
/>
{/* Toggle between single and multi-language */}
<Box sx={{ mb: 2 }}>
<Button
variant={!useMultiLanguage ? "contained" : "outlined"}
size="small"
onClick={() => setUseMultiLanguage(false)}
sx={{ mr: 1 }}
>
Single Language
</Button>
<Button
variant={useMultiLanguage ? "contained" : "outlined"}
size="small"
onClick={() => setUseMultiLanguage(true)}
>
Multiple Languages
</Button>
</Box>
{/* Single Language Selector */}
{!useMultiLanguage && (
<OcrLanguageSelector
value={selectedLanguage}
onChange={setSelectedLanguage}
label="OCR Language (Optional)"
size="medium"
helperText="Leave empty to use your default language setting"
showCurrentIndicator={true}
/>
)}
{/* Multi-Language Selector */}
{useMultiLanguage && (
<LanguageSelector
selectedLanguages={selectedLanguages}
primaryLanguage={primaryLanguage}
onLanguagesChange={handleLanguagesChange}
disabled={retrying}
showPrimarySelector={true}
className="w-full"
/>
)}
</Box>
<Alert severity="info" sx={{ mt: 2 }}>

View File

@ -32,6 +32,7 @@ import api from '../../services/api';
import { useNotifications } from '../../contexts/NotificationContext';
import LabelSelector from '../Labels/LabelSelector';
import { type LabelData } from '../Labels/Label';
import LanguageSelector from '../LanguageSelector';
interface UploadedDocument {
id: string;
@ -65,6 +66,8 @@ const UploadZone: React.FC<UploadZoneProps> = ({ onUploadComplete }) => {
const [selectedLabels, setSelectedLabels] = useState<LabelData[]>([]);
const [availableLabels, setAvailableLabels] = useState<LabelData[]>([]);
const [labelsLoading, setLabelsLoading] = useState<boolean>(false);
const [selectedLanguages, setSelectedLanguages] = useState<string[]>(['eng']);
const [primaryLanguage, setPrimaryLanguage] = useState<string>('eng');
useEffect(() => {
fetchLabels();
@ -99,6 +102,15 @@ const UploadZone: React.FC<UploadZoneProps> = ({ onUploadComplete }) => {
}
};
const handleLanguagesChange = (languages: string[], primary?: string) => {
setSelectedLanguages(languages);
if (primary) {
setPrimaryLanguage(primary);
} else if (languages.length > 0) {
setPrimaryLanguage(languages[0]);
}
};
const onDrop = useCallback((acceptedFiles: File[], rejectedFiles: FileRejection[]) => {
setError('');
@ -151,6 +163,13 @@ const UploadZone: React.FC<UploadZoneProps> = ({ onUploadComplete }) => {
formData.append('label_ids', JSON.stringify(labelIds));
}
// Add selected languages to the form data
if (selectedLanguages.length > 0) {
selectedLanguages.forEach((lang, index) => {
formData.append(`ocr_languages[${index}]`, lang);
});
}
try {
setFiles(prev => prev.map(f =>
f.id === fileItem.id
@ -338,6 +357,26 @@ const UploadZone: React.FC<UploadZoneProps> = ({ onUploadComplete }) => {
</Alert>
)}
{/* Language Selection */}
<Card elevation={0} sx={{ mb: 3 }}>
<CardContent>
<Typography variant="h6" sx={{ fontWeight: 600, mb: 2 }}>
🌐 OCR Language Settings
</Typography>
<Typography variant="body2" color="text.secondary" sx={{ mb: 2 }}>
Select languages for optimal OCR text recognition
</Typography>
<Box sx={{ '& > div': { width: '100%' } }}>
<LanguageSelector
selectedLanguages={selectedLanguages}
primaryLanguage={primaryLanguage}
onLanguagesChange={handleLanguagesChange}
disabled={uploading}
/>
</Box>
</CardContent>
</Card>
{/* Label Selection */}
<Card elevation={0} sx={{ mb: 3 }}>
<CardContent>

View File

@ -43,6 +43,7 @@ import { Edit as EditIcon, Delete as DeleteIcon, Add as AddIcon,
import { useAuth } from '../contexts/AuthContext';
import api, { queueService } from '../services/api';
import OcrLanguageSelector from '../components/OcrLanguageSelector';
import LanguageSelector from '../components/LanguageSelector';
interface User {
id: string;
@ -53,6 +54,9 @@ interface User {
interface Settings {
ocrLanguage: string;
preferredLanguages: string[];
primaryLanguage: string;
autoDetectLanguageCombination: boolean;
concurrentOcrJobs: number;
ocrTimeoutSeconds: number;
maxFileSizeMb: number;
@ -188,6 +192,9 @@ const SettingsPage: React.FC = () => {
const [tabValue, setTabValue] = useState<number>(0);
const [settings, setSettings] = useState<Settings>({
ocrLanguage: 'eng',
preferredLanguages: ['eng'],
primaryLanguage: 'eng',
autoDetectLanguageCombination: false,
concurrentOcrJobs: 4,
ocrTimeoutSeconds: 300,
maxFileSizeMb: 50,
@ -268,6 +275,9 @@ const SettingsPage: React.FC = () => {
const response = await api.get('/settings');
setSettings({
ocrLanguage: response.data.ocr_language || 'eng',
preferredLanguages: response.data.preferred_languages || ['eng'],
primaryLanguage: response.data.primary_language || 'eng',
autoDetectLanguageCombination: response.data.auto_detect_language_combination || false,
concurrentOcrJobs: response.data.concurrent_ocr_jobs || 4,
ocrTimeoutSeconds: response.data.ocr_timeout_seconds || 300,
maxFileSizeMb: response.data.max_file_size_mb || 50,
@ -430,6 +440,20 @@ const SettingsPage: React.FC = () => {
handleSettingsChange('searchResultsPerPage', event.target.value);
};
const handleLanguagesChange = (languages: string[], primary?: string) => {
// Update multiple fields at once
const updates = {
preferredLanguages: languages,
primaryLanguage: primary || languages[0] || 'eng',
ocrLanguage: primary || languages[0] || 'eng', // Backward compatibility
};
// Update all language-related settings
Object.entries(updates).forEach(([key, value]) => {
handleSettingsChange(key as keyof Settings, value);
});
};
const fetchOcrStatus = async (): Promise<void> => {
try {
const response = await queueService.getOcrStatus();
@ -521,14 +545,34 @@ const SettingsPage: React.FC = () => {
</Typography>
<Divider sx={{ mb: 2 }} />
<Grid container spacing={2}>
<Grid item xs={12} md={6}>
<OcrLanguageSelector
value={settings.ocrLanguage}
onChange={(language) => handleSettingsChange('ocrLanguage', language)}
disabled={loading}
showCurrentIndicator={false}
helperText="Default language for OCR text extraction from your documents"
<Grid item xs={12}>
<Typography variant="body2" color="text.secondary" sx={{ mb: 2 }}>
Configure languages for OCR text extraction. Multiple languages help with mixed-language documents.
</Typography>
<Box sx={{ '& > div': { width: '100%' } }}>
<LanguageSelector
selectedLanguages={settings.preferredLanguages}
primaryLanguage={settings.primaryLanguage}
onLanguagesChange={handleLanguagesChange}
disabled={loading}
showPrimarySelector={true}
/>
</Box>
</Grid>
<Grid item xs={12}>
<FormControlLabel
control={
<Switch
checked={settings.autoDetectLanguageCombination}
onChange={(e) => handleSettingsChange('autoDetectLanguageCombination', e.target.checked)}
disabled={loading}
/>
}
label="Auto-detect language combinations"
/>
<Typography variant="caption" color="text.secondary" sx={{ display: 'block', mt: 0.5 }}>
Automatically suggest optimal language combinations based on document content analysis
</Typography>
</Grid>
<Grid item xs={12} md={6}>
<TextField

View File

@ -230,9 +230,17 @@ export interface OcrResponse {
}
export const documentService = {
upload: (file: File) => {
upload: (file: File, languages?: string[]) => {
const formData = new FormData()
formData.append('file', file)
// Add multiple languages if provided
if (languages && languages.length > 0) {
languages.forEach((lang, index) => {
formData.append(`ocr_languages[${index}]`, lang)
})
}
return api.post('/documents', formData, {
headers: {
'Content-Type': 'multipart/form-data',
@ -417,6 +425,7 @@ export interface AvailableLanguagesResponse {
export interface RetryOcrRequest {
language?: string
languages?: string[]
}
export const queueService = {
@ -450,9 +459,11 @@ export const ocrService = {
return api.get('/ocr/health')
},
retryWithLanguage: (documentId: string, language?: string) => {
retryWithLanguage: (documentId: string, language?: string, languages?: string[]) => {
const data: RetryOcrRequest = {}
if (language) {
if (languages && languages.length > 0) {
data.languages = languages
} else if (language) {
data.language = language
}
return api.post(`/documents/${documentId}/retry-ocr`, data)