Feature/mineru improvements (#11938)

我已在下面的评论中用中文重复说明。

### What problem does this PR solve?

## Summary
This PR enhances the MinerU document parser with additional
configuration options, giving users more control over PDF parsing
behavior and improving support for multilingual documents.

## Changes

### Backend (`deepdoc/parser/mineru_parser.py`)
- Added configurable parsing options:
- **Parse Method**: `auto`, `txt`, or `ocr` — allows users to choose the
extraction strategy
- **Formula Recognition**: Toggle for enabling/disabling formula
extraction (useful to disable for Cyrillic documents where it may cause
issues)
- **Table Recognition**: Toggle for enabling/disabling table extraction
- Added language code mapping (`LANGUAGE_TO_MINERU_MAP`) to translate
RAGFlow language settings to MinerU-compatible language codes for better
OCR accuracy
- Improved parser configuration handling to pass these options through
the processing pipeline

### Frontend (`web/`)
- Created new `MinerUOptionsFormField` component that conditionally
renders when MinerU is selected as the layout recognition engine
- Added UI controls for:
  - Parse method selection (dropdown)
  - Formula recognition toggle (switch)
  - Table recognition toggle (switch)
- Added i18n translations for English and Chinese
- Integrated the options into both the dataset creation dialog and
dataset settings page

### Integration
- Updated `rag/app/naive.py` to forward MinerU options to the parser
- Updated task service to handle the new configuration parameters

## Why
MinerU is a powerful document parser, but the default settings don't
work well for all document types. This PR allows users to:
1. Choose the best parsing method for their documents
2. Disable formula recognition for Cyrillic/non-Latin scripts where it
causes issues
3. Control table extraction based on document needs
4. Benefit from automatic language detection for better OCR results

## Testing
- [x] Tested MinerU parsing with different parse methods
- [x] Verified UI renders correctly when MinerU is selected/deselected
- [x] Confirmed settings persist correctly in dataset configuration

### Type of change

- [x] Bug Fix (non-breaking change which fixes an issue)
- [x] New Feature (non-breaking change which adds functionality)
- [ ] Documentation Update
- [x] Refactoring
- [ ] Performance Improvement
- [ ] Other (please describe):

---------

Co-authored-by: user210 <user210@rt>
Co-authored-by: Kevin Hu <kevinhu.sh@gmail.com>
This commit is contained in:
concertdictate
2025-12-16 07:15:25 +02:00
committed by GitHub
parent 1112b6291b
commit 49c74d08e8
13 changed files with 371 additions and 46 deletions

View File

@ -0,0 +1,97 @@
import { RAGFlowFormItem } from '@/components/ragflow-form';
import { RAGFlowSelect } from '@/components/ui/select';
import { Switch } from '@/components/ui/switch';
import { LLMFactory } from '@/constants/llm';
import { buildOptions } from '@/utils/form';
import { useFormContext, useWatch } from 'react-hook-form';
import { useTranslation } from 'react-i18next';
const parseMethodOptions = buildOptions(['auto', 'txt', 'ocr']);
export function MinerUOptionsFormField() {
const form = useFormContext();
const { t } = useTranslation();
const layoutRecognize = useWatch({
control: form.control,
name: 'parser_config.layout_recognize',
});
// Check if MinerU is selected (the value contains 'MinerU' or matches the factory name)
const isMinerUSelected =
layoutRecognize?.includes(LLMFactory.MinerU) ||
layoutRecognize?.toLowerCase()?.includes('mineru');
if (!isMinerUSelected) {
return null;
}
return (
<div className="space-y-4 border-l-2 border-primary/30 pl-4 ml-2">
<div className="text-sm font-medium text-text-secondary">
{t('knowledgeConfiguration.mineruOptions', 'MinerU Options')}
</div>
<RAGFlowFormItem
name="parser_config.mineru_parse_method"
label={t('knowledgeConfiguration.mineruParseMethod', 'Parse Method')}
tooltip={t(
'knowledgeConfiguration.mineruParseMethodTip',
'Method for parsing PDF: auto (automatic detection), txt (text extraction), ocr (optical character recognition)',
)}
horizontal={true}
>
{(field) => (
<RAGFlowSelect
value={field.value || 'auto'}
onChange={field.onChange}
options={parseMethodOptions}
placeholder={t('common.selectPlaceholder', 'Select value')}
/>
)}
</RAGFlowFormItem>
<RAGFlowFormItem
name="parser_config.mineru_formula_enable"
label={t(
'knowledgeConfiguration.mineruFormulaEnable',
'Formula Recognition',
)}
tooltip={t(
'knowledgeConfiguration.mineruFormulaEnableTip',
'Enable formula recognition. Note: This may not work correctly for Cyrillic documents.',
)}
horizontal={true}
labelClassName="!mb-0"
>
{(field) => (
<Switch
checked={field.value ?? true}
onCheckedChange={field.onChange}
/>
)}
</RAGFlowFormItem>
<RAGFlowFormItem
name="parser_config.mineru_table_enable"
label={t(
'knowledgeConfiguration.mineruTableEnable',
'Table Recognition',
)}
tooltip={t(
'knowledgeConfiguration.mineruTableEnableTip',
'Enable table recognition and extraction.',
)}
horizontal={true}
labelClassName="!mb-0"
>
{(field) => (
<Switch
checked={field.value ?? true}
onCheckedChange={field.onChange}
/>
)}
</RAGFlowFormItem>
</div>
);
}

View File

@ -103,14 +103,22 @@ export const LanguageTranslationMap = {
Chinese: 'zh',
'Traditional Chinese': 'zh-TRADITIONAL',
Russian: 'ru',
Indonesia: 'id',
Indonesian: 'id',
Spanish: 'es',
Vietnamese: 'vi',
Japanese: 'ja',
Korean: 'ko',
'Portuguese BR': 'pt-br',
German: 'de',
French: 'fr',
Italian: 'it',
Tamil: 'ta',
Telugu: 'te',
Kannada: 'ka',
Thai: 'th',
Greek: 'el',
Hindi: 'hi',
Ukrainian: 'uk',
};
export enum FileMimeType {

View File

@ -330,6 +330,15 @@ export default {
reRankModelWaring: 'Re-rank model is very time consuming.',
},
knowledgeConfiguration: {
mineruOptions: 'MinerU Options',
mineruParseMethod: 'Parse Method',
mineruParseMethodTip:
'Method for parsing PDF: auto (automatic detection), txt (text extraction), ocr (optical character recognition)',
mineruFormulaEnable: 'Formula Recognition',
mineruFormulaEnableTip:
'Enable formula recognition. Note: This may not work correctly for Cyrillic documents.',
mineruTableEnable: 'Table Recognition',
mineruTableEnableTip: 'Enable table recognition and extraction.',
overlappedPercent: 'Overlapped percent(%)',
generationScopeTip:
'Determines whether RAPTOR is generated for the entire dataset or for a single file.',

View File

@ -265,6 +265,15 @@ export default {
theDocumentBeingParsedCannotBeDeleted: '正在解析的文档不能被删除',
},
knowledgeConfiguration: {
mineruOptions: 'MinerU 选项',
mineruParseMethod: '解析方法',
mineruParseMethodTip:
'PDF 解析方法auto自动检测、txt文本提取、ocr光学字符识别',
mineruFormulaEnable: '公式识别',
mineruFormulaEnableTip:
'启用公式识别。注意:对于西里尔文档可能无法正常工作。',
mineruTableEnable: '表格识别',
mineruTableEnableTip: '启用表格识别和提取。',
generationScopeTip: '选择 RAPTOR 的生成范围:整个知识库或单个文件。',
generationScope: '生成范围',
scopeSingleFile: '单文件',

View File

@ -6,6 +6,7 @@ import { DelimiterFormField } from '@/components/delimiter-form-field';
import { ExcelToHtmlFormField } from '@/components/excel-to-html-form-field';
import { LayoutRecognizeFormField } from '@/components/layout-recognize-form-field';
import { MaxTokenNumberFormField } from '@/components/max-token-number-from-field';
import { MinerUOptionsFormField } from '@/components/mineru-options-form-field';
import {
ConfigurationFormContainer,
MainContainer,
@ -17,6 +18,7 @@ export function NaiveConfiguration() {
<MainContainer>
<ConfigurationFormContainer>
<LayoutRecognizeFormField></LayoutRecognizeFormField>
<MinerUOptionsFormField></MinerUOptionsFormField>
<MaxTokenNumberFormField initialValue={512}></MaxTokenNumberFormField>
<DelimiterFormField></DelimiterFormField>
<EnableTocToggle />

View File

@ -13,6 +13,7 @@ export const formSchema = z
// avatar: z.instanceof(File),
avatar: z.any().nullish(),
permission: z.string().optional(),
language: z.string().optional(),
parser_id: z.string(),
pipeline_id: z.string().optional(),
pipeline_name: z.string().optional(),
@ -30,6 +31,10 @@ export const formSchema = z
topn_tags: z.number().optional(),
toc_extraction: z.boolean().optional(),
overlapped_percent: z.number().optional(),
// MinerU-specific options
mineru_parse_method: z.enum(['auto', 'txt', 'ocr']).optional(),
mineru_formula_enable: z.boolean().optional(),
mineru_table_enable: z.boolean().optional(),
raptor: z
.object({
use_raptor: z.boolean().optional(),

View File

@ -1,5 +1,7 @@
import { AvatarUpload } from '@/components/avatar-upload';
import { SelectWithSearch } from '@/components/originui/select-with-search';
import PageRankFormField from '@/components/page-rank-form-field';
import { RAGFlowFormItem } from '@/components/ragflow-form';
import {
FormControl,
FormField,
@ -8,6 +10,8 @@ import {
FormMessage,
} from '@/components/ui/form';
import { Input } from '@/components/ui/input';
import { LanguageTranslationMap } from '@/constants/common';
import { useMemo } from 'react';
import { useFormContext } from 'react-hook-form';
import { useTranslation } from 'react-i18next';
import { TagItems } from './components/tag-item';
@ -18,6 +22,13 @@ export function GeneralForm() {
const form = useFormContext();
const { t } = useTranslation();
const languageOptions = useMemo(() => {
return Object.keys(LanguageTranslationMap).map((x) => ({
label: x,
value: x,
}));
}, []);
return (
<>
<FormField
@ -41,6 +52,18 @@ export function GeneralForm() {
</FormItem>
)}
/>
<div className="items-center">
<RAGFlowFormItem
name="language"
label={t('common.language')}
horizontal={true}
>
<SelectWithSearch
options={languageOptions}
triggerClassName="w-full"
></SelectWithSearch>
</RAGFlowFormItem>
</div>
<FormField
control={form.control}
name="avatar"

View File

@ -58,6 +58,7 @@ export default function DatasetSettings() {
name: '',
parser_id: DocumentParserType.Naive,
permission: PermissionRole.Me,
language: 'English',
parser_config: {
layout_recognize: DocumentType.DeepDOC,
chunk_token_num: 512,
@ -68,6 +69,10 @@ export default function DatasetSettings() {
topn_tags: 3,
toc_extraction: false,
overlapped_percent: 0,
// MinerU-specific defaults
mineru_parse_method: 'auto',
mineru_formula_enable: true,
mineru_table_enable: true,
raptor: {
use_raptor: true,
max_token: 256,

View File

@ -16,10 +16,18 @@ import {
FormMessage,
} from '@/components/ui/form';
import { Input } from '@/components/ui/input';
import {
Select,
SelectContent,
SelectItem,
SelectTrigger,
SelectValue,
} from '@/components/ui/select';
import { LanguageTranslationMap } from '@/constants/common';
import { FormLayout } from '@/constants/form';
import { IModalProps } from '@/interfaces/common';
import { zodResolver } from '@hookform/resolvers/zod';
import { useEffect } from 'react';
import { useEffect, useMemo } from 'react';
import { useForm, useWatch } from 'react-hook-form';
import { useTranslation } from 'react-i18next';
import { z } from 'zod';
@ -34,6 +42,13 @@ const FormId = 'dataset-creating-form';
export function InputForm({ onOk }: IModalProps<any>) {
const { t } = useTranslation();
const languageOptions = useMemo(() => {
return Object.keys(LanguageTranslationMap).map((x) => ({
label: x,
value: x,
}));
}, []);
const FormSchema = z
.object({
name: z
@ -51,6 +66,7 @@ export function InputForm({ onOk }: IModalProps<any>) {
.trim(),
parser_id: z.string().optional(),
pipeline_id: z.string().optional(),
language: z.string().optional(),
})
.superRefine((data, ctx) => {
// When parseType === 1, parser_id is required
@ -83,6 +99,7 @@ export function InputForm({ onOk }: IModalProps<any>) {
parseType: 1,
parser_id: '',
embd_id: '',
language: 'English',
},
});
@ -130,6 +147,33 @@ export function InputForm({ onOk }: IModalProps<any>) {
)}
/>
<FormField
control={form.control}
name="language"
render={({ field }) => (
<FormItem>
<FormLabel>{t('common.language')}</FormLabel>
<Select onValueChange={field.onChange} defaultValue={field.value}>
<FormControl>
<SelectTrigger>
<SelectValue
placeholder={t('common.languagePlaceholder')}
/>
</SelectTrigger>
</FormControl>
<SelectContent>
{languageOptions.map((option) => (
<SelectItem key={option.value} value={option.value}>
{option.label}
</SelectItem>
))}
</SelectContent>
</Select>
<FormMessage />
</FormItem>
)}
/>
<EmbeddingModelItem line={2} isEdit={false} />
<ParseTypeItem />
{parseType === 1 && <ChunkMethodItem></ChunkMethodItem>}