Refactor: enhance graphrag - part 2 (#14972)

### What problem does this PR solve?
1. expose batch_chunk_token_size for configuration
2. retrieve chunks when build subgraph for the doc, not retreive all
docs chunks at the begining
3. get all chunks for a document, used to be hard coded 10000
4. delete not used method run_graphrag

### Type of change

- [x] New Feature (non-breaking change which adds functionality)
- [x] Refactoring

Follow on: #14617

web/src/components/parse-configuration/graph-rag-form-fields.tsx

@@ -1,3 +1,4 @@
+import { FormLayout } from '@/constants/form';
 import { DocumentParserType } from '@/constants/knowledge';
 import { useTranslate } from '@/hooks/common-hooks';
 import { cn } from '@/lib/utils';
@@ -12,6 +13,7 @@ import { useCallback, useMemo } from 'react';
 import { useFormContext, useWatch } from 'react-hook-form';
 import { EntityTypesFormField } from '../entity-types-form-field';
 import { FormContainer } from '../form-container';
+import { SliderInputFormField } from '../slider-input-form-field';
 import {
   FormControl,
   FormField,
@@ -191,6 +193,19 @@ const GraphRagItems = ({
             )}
           />
 
+          <SliderInputFormField
+            name="parser_config.graphrag.batch_chunk_token_size"
+            label={t('graphRagBatchChunkTokenSize')}
+            tooltip={t('graphRagBatchChunkTokenSizeTip')}
+            max={8196}
+            min={512}
+            step={1}
+            defaultValue={4096}
+            layout={FormLayout.Horizontal}
+            sliderTestId="ds-settings-graph-batch-chunk-token-size-slider"
+            numberInputTestId="ds-settings-graph-batch-chunk-token-size-input"
+          ></SliderInputFormField>
+
           <FormField
             control={form.control}
             name="parser_config.graphrag.resolution"

web/src/interfaces/database/dataset.ts

@@ -89,6 +89,7 @@ interface Parentchild {
 }
 
 interface Graphrag {
+  batch_chunk_token_size?: number;
   entity_types: string[];
   method: string;
   use_graphrag: boolean;

web/src/interfaces/database/document.ts

@@ -67,6 +67,7 @@ interface Raptor {
 }
 
 interface GraphRag {
+  batch_chunk_token_size?: number;
   community?: boolean;
   entity_types?: string[];
   method?: string;

web/src/locales/en.ts

@@ -903,6 +903,9 @@ This auto-tagging feature enhances retrieval by adding another layer of domain-s
       Light: (Default) Use prompts provided by github.com/HKUDS/LightRAG to extract entities and relationships. This option consumes fewer tokens, less memory, and fewer computational resources.</br>
       General: Use prompts provided by github.com/microsoft/graphrag to extract entities and relationships.</br>
       NER: Use spaCy NER and rule-based keyword extraction to extract entities and relationships. No LLM is required for extraction itself, making it fast and resource-efficient.`,
+      graphRagBatchChunkTokenSize: 'Batch chunk token size',
+      graphRagBatchChunkTokenSizeTip:
+        'The token limit for each batch of chunks sent to the LLM for knowledge graph entity and relation extraction. Not applied to NER.',
       resolution: 'Entity resolution',
       resolutionTip: `An entity deduplication switch. When enabled, the LLM will combine similar entities - e.g., '2025' and 'the year of 2025', or 'IT' and 'Information Technology' - to construct a more accurate graph`,
       community: 'Community reports',

web/src/locales/zh.ts

@@ -818,6 +818,9 @@ export default {
       graphRagMethodTip: `Light：实体和关系提取提示来自 GitHub - HKUDS/LightRAG：“LightRAG：简单快速的检索增强生成”<br>
 General：实体和关系提取提示来自 GitHub - microsoft/graphrag：基于图的模块化检索增强生成 (RAG) 系统<br>
 NER：使用 spaCy NER 和基于规则的关键词提取来抽取实体和关系，无需 LLM 参与提取过程，速度快且资源消耗低`,
+      graphRagBatchChunkTokenSize: '批量chunk token 大小',
+      graphRagBatchChunkTokenSizeTip:
+        '发送给 LLM 进行知识图谱实体和关系抽取时，每批文本块的 token 上限。NER 不适用。',
       resolution: '实体归一化',
       resolutionTip: `解析过程会将具有相同含义的实体合并在一起，从而使知识图谱更简洁、更准确。应合并以下实体：特朗普总统、唐纳德·特朗普、唐纳德·J·特朗普、唐纳德·约翰·特朗普`,
       community: '社区报告生成',

web/src/pages/dataset/dataset-setting/form-schema.ts

@@ -70,6 +70,12 @@ export const formSchema = z
             method: z.string().optional(),
             resolution: z.boolean().optional(),
             community: z.boolean().optional(),
+            batch_chunk_token_size: z
+              .number()
+              .int()
+              .min(512)
+              .max(8196)
+              .optional(),
           })
           .refine(
             (data) => {

web/src/pages/dataset/dataset-setting/index.tsx

@@ -103,6 +103,7 @@ export default function DatasetSettings() {
           use_graphrag: true,
           entity_types: initialEntityTypes,
           method: MethodValue.Light,
+          batch_chunk_token_size: 4096,
         },
         metadata: {
           type: 'object',