remove token in Indexing estimate

api/.idea/vcs.xml

@@ -12,5 +12,6 @@
   </component>
   <component name="VcsDirectoryMappings">
     <mapping directory="" vcs="Git" />
+    <mapping directory="$PROJECT_DIR$/.." vcs="Git" />
   </component>
 </project>

api/controllers/console/datasets/datasets_document.py

@@ -302,6 +302,10 @@ class DatasetInitApi(Resource):
                             location='json')
         parser.add_argument('retrieval_model', type=dict, required=False, nullable=False,
                             location='json')
+        parser.add_argument('embedding_model', type=str,  required=False, nullable=True,
+                            location='json')
+        parser.add_argument('embedding_model_provider', type=str, required=False, nullable=True,
+                            location='json')
         args = parser.parse_args()
 
         # The role of the current user in the ta table must be admin, owner, or editor, or dataset_operator
@@ -309,6 +313,8 @@ class DatasetInitApi(Resource):
             raise Forbidden()
 
         if args['indexing_technique'] == 'high_quality':
+            if args['embedding_model'] is None or args['embedding_model_provider'] is None:
+                raise ValueError('embedding model and embedding model provider are required for high quality indexing.')
             try:
                 model_manager = ModelManager()
                 model_manager.get_default_model_instance(
@@ -419,9 +425,6 @@ class DocumentBatchIndexingEstimateApi(DocumentResource):
         batch = str(batch)
         documents = self.get_batch_documents(dataset_id, batch)
         response = {
-            "tokens": 0,
-            "total_price": 0,
-            "currency": "USD",
             "total_segments": 0,
             "preview": []
         }

api/core/indexing_runner.py

@@ -255,11 +255,8 @@ class IndexingRunner:
                     tenant_id=tenant_id,
                     model_type=ModelType.TEXT_EMBEDDING,
                 )
-        tokens = 0
         preview_texts = []
         total_segments = 0
-        total_price = 0
-        currency = 'USD'
         index_type = doc_form
         index_processor = IndexProcessorFactory(index_type).init_index_processor()
         all_text_docs = []
@@ -286,54 +283,22 @@ class IndexingRunner:
             for document in documents:
                 if len(preview_texts) < 5:
                     preview_texts.append(document.page_content)
-                if indexing_technique == 'high_quality' or embedding_model_instance:
-                    tokens += embedding_model_instance.get_text_embedding_num_tokens(
-                        texts=[self.filter_string(document.page_content)]
-                    )
 
         if doc_form and doc_form == 'qa_model':
-            model_instance = self.model_manager.get_default_model_instance(
-                tenant_id=tenant_id,
-                model_type=ModelType.LLM
-            )
-
-            model_type_instance = model_instance.model_type_instance
-            model_type_instance = cast(LargeLanguageModel, model_type_instance)
 
             if len(preview_texts) > 0:
                 # qa model document
                 response = LLMGenerator.generate_qa_document(current_user.current_tenant_id, preview_texts[0],
                                                              doc_language)
                 document_qa_list = self.format_split_text(response)
-                price_info = model_type_instance.get_price(
-                    model=model_instance.model,
-                    credentials=model_instance.credentials,
-                    price_type=PriceType.INPUT,
-                    tokens=total_segments * 2000,
-                )
+
                 return {
                     "total_segments": total_segments * 20,
-                    "tokens": total_segments * 2000,
-                    "total_price": '{:f}'.format(price_info.total_amount),
-                    "currency": price_info.currency,
                     "qa_preview": document_qa_list,
                     "preview": preview_texts
                 }
-        if embedding_model_instance:
-            embedding_model_type_instance = cast(TextEmbeddingModel, embedding_model_instance.model_type_instance)
-            embedding_price_info = embedding_model_type_instance.get_price(
-                model=embedding_model_instance.model,
-                credentials=embedding_model_instance.credentials,
-                price_type=PriceType.INPUT,
-                tokens=tokens
-            )
-            total_price = '{:f}'.format(embedding_price_info.total_amount)
-            currency = embedding_price_info.currency
         return {
             "total_segments": total_segments,
-            "tokens": tokens,
-            "total_price": total_price,
-            "currency": currency,
             "preview": preview_texts
         }
 

api/core/rag/splitter/text_splitter.py

@@ -12,7 +12,7 @@ from typing import (
     Optional,
     TypedDict,
     TypeVar,
-    Union,
+    Union, List,
 )
 
 from core.rag.models.document import BaseDocumentTransformer, Document
@@ -109,7 +109,7 @@ class TextSplitter(BaseDocumentTransformer, ABC):
         else:
             return text
 
-    def _merge_splits(self, splits: Iterable[str], separator: str) -> list[str]:
+    def _merge_splits(self, splits: Iterable[str], separator: str, lengths: List[int]) -> list[str]:
         # We now want to combine these smaller pieces into medium size
         # chunks to send to the LLM.
         separator_len = self._length_function(separator)
@@ -117,8 +117,9 @@ class TextSplitter(BaseDocumentTransformer, ABC):
         docs = []
         current_doc: list[str] = []
         total = 0
+        index = 0
         for d in splits:
-            _len = self._length_function(d)
+            _len = lengths[index]
             if (
                     total + _len + (separator_len if len(current_doc) > 0 else 0)
                     > self._chunk_size
@@ -146,6 +147,7 @@ class TextSplitter(BaseDocumentTransformer, ABC):
                         current_doc = current_doc[1:]
             current_doc.append(d)
             total += _len + (separator_len if len(current_doc) > 1 else 0)
+            index += 1
         doc = self._join_docs(current_doc, separator)
         if doc is not None:
             docs.append(doc)
@@ -494,11 +496,10 @@ class RecursiveCharacterTextSplitter(TextSplitter):
         self._separators = separators or ["\n\n", "\n", " ", ""]
 
     def _split_text(self, text: str, separators: list[str]) -> list[str]:
-        """Split incoming text and return chunks."""
         final_chunks = []
-        # Get appropriate separator to use
         separator = separators[-1]
         new_separators = []
+
         for i, _s in enumerate(separators):
             if _s == "":
                 separator = _s
@@ -509,25 +510,31 @@ class RecursiveCharacterTextSplitter(TextSplitter):
                 break
 
         splits = _split_text_with_regex(text, separator, self._keep_separator)
-        # Now go merging things, recursively splitting longer texts.
         _good_splits = []
+        _good_splits_lengths = []  # cache the lengths of the splits
         _separator = "" if self._keep_separator else separator
+
         for s in splits:
-            if self._length_function(s) < self._chunk_size:
+            s_len = self._length_function(s)
+            if s_len < self._chunk_size:
                 _good_splits.append(s)
+                _good_splits_lengths.append(s_len)
             else:
                 if _good_splits:
-                    merged_text = self._merge_splits(_good_splits, _separator)
+                    merged_text = self._merge_splits(_good_splits, _separator, _good_splits_lengths)
                     final_chunks.extend(merged_text)
                     _good_splits = []
+                    _good_splits_lengths = []
                 if not new_separators:
                     final_chunks.append(s)
                 else:
                     other_info = self._split_text(s, new_separators)
                     final_chunks.extend(other_info)
+
         if _good_splits:
-            merged_text = self._merge_splits(_good_splits, _separator)
+            merged_text = self._merge_splits(_good_splits, _separator, _good_splits_lengths)
             final_chunks.extend(merged_text)
+
         return final_chunks
 
     def split_text(self, text: str) -> list[str]:

api/services/dataset_service.py

@@ -1067,18 +1067,12 @@ class DocumentService:
 
             DocumentService.check_documents_upload_quota(count, features)
 
-        embedding_model = None
         dataset_collection_binding_id = None
         retrieval_model = None
         if document_data['indexing_technique'] == 'high_quality':
-            model_manager = ModelManager()
-            embedding_model = model_manager.get_default_model_instance(
-                tenant_id=current_user.current_tenant_id,
-                model_type=ModelType.TEXT_EMBEDDING
-            )
             dataset_collection_binding = DatasetCollectionBindingService.get_dataset_collection_binding(
-                embedding_model.provider,
-                embedding_model.model
+                document_data['embedding_model_provider'],
+                document_data['embedding_model']
             )
             dataset_collection_binding_id = dataset_collection_binding.id
             if document_data.get('retrieval_model'):
@@ -1100,10 +1094,10 @@ class DocumentService:
             tenant_id=tenant_id,
             name='',
             data_source_type=document_data["data_source"]["type"],
-            indexing_technique=document_data["indexing_technique"],
+            indexing_technique=document_data.get('indexing_technique', 'high_quality'),
             created_by=account.id,
-            embedding_model=embedding_model.model if embedding_model else None,
-            embedding_model_provider=embedding_model.provider if embedding_model else None,
+            embedding_model=document_data.get('embedding_model', None),
+            embedding_model_provider=document_data.get('embedding_model_provider', None),
             collection_binding_id=dataset_collection_binding_id,
             retrieval_model=retrieval_model
         )

web/app/components/datasets/create/embedding-process/index.tsx

@@ -13,8 +13,7 @@ import cn from '@/utils/classnames'
 import { FieldInfo } from '@/app/components/datasets/documents/detail/metadata'
 import Button from '@/app/components/base/button'
 import type { FullDocumentDetail, IndexingStatusResponse, ProcessRuleResponse } from '@/models/datasets'
-import { formatNumber } from '@/utils/format'
-import { fetchIndexingStatusBatch as doFetchIndexingStatus, fetchIndexingEstimateBatch, fetchProcessRule } from '@/service/datasets'
+import { fetchIndexingStatusBatch as doFetchIndexingStatus, fetchProcessRule } from '@/service/datasets'
 import { DataSourceType } from '@/models/datasets'
 import NotionIcon from '@/app/components/base/notion-icon'
 import PriorityLabel from '@/app/components/billing/priority-label'
@@ -142,14 +141,6 @@ const EmbeddingProcess: FC<Props> = ({ datasetId, batchId, documents = [], index
   }, apiParams => fetchProcessRule(omit(apiParams, 'action')), {
     revalidateOnFocus: false,
   })
-  // get cost
-  const { data: indexingEstimateDetail } = useSWR({
-    action: 'fetchIndexingEstimateBatch',
-    datasetId,
-    batchId,
-  }, apiParams => fetchIndexingEstimateBatch(omit(apiParams, 'action')), {
-    revalidateOnFocus: false,
-  })
 
   const router = useRouter()
   const navToDocumentList = () => {
@@ -190,28 +181,11 @@ const EmbeddingProcess: FC<Props> = ({ datasetId, batchId, documents = [], index
 
   return (
     <>
-      <div className='h-5 flex justify-between items-center mb-5'>
+      <div className='h-5 flex items-center mb-5'>
         <div className={s.embeddingStatus}>
           {isEmbedding && t('datasetDocuments.embedding.processing')}
           {isEmbeddingCompleted && t('datasetDocuments.embedding.completed')}
         </div>
-        <div className={s.cost}>
-          {indexingType === 'high_quality' && (
-            <div className='flex items-center'>
-              <div className={cn(s.commonIcon, s.highIcon)} />
-              {t('datasetDocuments.embedding.highQuality')} · {t('datasetDocuments.embedding.estimate')}
-              <span className={s.tokens}>{formatNumber(indexingEstimateDetail?.tokens || 0)}</span>tokens
-              (<span className={s.price}>${formatNumber(indexingEstimateDetail?.total_price || 0)}</span>)
-            </div>
-          )}
-          {indexingType === 'economy' && (
-            <div className='flex items-center'>
-              <div className={cn(s.commonIcon, s.economyIcon)} />
-              {t('datasetDocuments.embedding.economy')} · {t('datasetDocuments.embedding.estimate')}
-              <span className={s.tokens}>0</span>tokens
-            </div>
-          )}
-        </div>
       </div>
       {
         enableBilling && plan.type !== Plan.team && (

web/app/components/datasets/create/step-two/index.module.css

@@ -30,7 +30,7 @@
 }
 
 .indexItem {
-  min-height: 146px;
+  min-height: 126px;
 }
 
 .indexItem .disableMask {
@@ -121,10 +121,6 @@
   @apply pb-1;
 }
 
-.radioItem.indexItem .typeHeader .tip {
-  @apply pb-3;
-}
-
 .radioItem .typeIcon {
   position: absolute;
   top: 18px;
@@ -264,7 +260,7 @@
 }
 
 .input {
-  @apply inline-flex h-9 w-full py-1 px-2 rounded-lg text-xs leading-normal;
+  @apply inline-flex h-9 w-full py-1 px-2 pr-14 rounded-lg text-xs leading-normal;
   @apply bg-gray-100 caret-primary-600 hover:bg-gray-100 focus:ring-1 focus:ring-inset focus:ring-gray-200 focus-visible:outline-none focus:bg-white placeholder:text-gray-400;
 }
 

web/app/components/datasets/create/step-two/index.tsx

@@ -16,7 +16,7 @@ import PreviewItem, { PreviewType } from './preview-item'
 import LanguageSelect from './language-select'
 import s from './index.module.css'
 import cn from '@/utils/classnames'
-import type { CrawlOptions, CrawlResultItem, CreateDocumentReq, CustomFile, FileIndexingEstimateResponse, FullDocumentDetail, IndexingEstimateParams, IndexingEstimateResponse, NotionInfo, PreProcessingRule, ProcessRule, Rules, createDocumentResponse } from '@/models/datasets'
+import type { CrawlOptions, CrawlResultItem, CreateDocumentReq, CustomFile, FileIndexingEstimateResponse, FullDocumentDetail, IndexingEstimateParams, NotionInfo, PreProcessingRule, ProcessRule, Rules, createDocumentResponse } from '@/models/datasets'
 import {
   createDocument,
   createFirstDocument,
@@ -44,8 +44,10 @@ import { RETRIEVE_METHOD } from '@/types/app'
 import useBreakpoints, { MediaType } from '@/hooks/use-breakpoints'
 import Tooltip from '@/app/components/base/tooltip'
 import TooltipPlus from '@/app/components/base/tooltip-plus'
-import { useModelListAndDefaultModelAndCurrentProviderAndModel } from '@/app/components/header/account-setting/model-provider-page/hooks'
+import { useDefaultModel, useModelList, useModelListAndDefaultModelAndCurrentProviderAndModel } from '@/app/components/header/account-setting/model-provider-page/hooks'
 import { LanguagesSupported } from '@/i18n/language'
+import ModelSelector from '@/app/components/header/account-setting/model-provider-page/model-selector'
+import type { DefaultModel } from '@/app/components/header/account-setting/model-provider-page/declarations'
 import { ModelTypeEnum } from '@/app/components/header/account-setting/model-provider-page/declarations'
 import { Globe01 } from '@/app/components/base/icons/src/vender/line/mapsAndTravel'
 
@@ -112,7 +114,7 @@ const StepTwo = ({
   const [previewScrolled, setPreviewScrolled] = useState(false)
   const [segmentationType, setSegmentationType] = useState<SegmentType>(SegmentType.AUTO)
   const [segmentIdentifier, setSegmentIdentifier] = useState('\\n')
-  const [max, setMax] = useState(500)
+  const [max, setMax] = useState(5000) // default chunk length
   const [overlap, setOverlap] = useState(50)
   const [rules, setRules] = useState<PreProcessingRule[]>([])
   const [defaultConfig, setDefaultConfig] = useState<Rules>()
@@ -132,7 +134,6 @@ const StepTwo = ({
   const [showPreview, { setTrue: setShowPreview, setFalse: hidePreview }] = useBoolean()
   const [customFileIndexingEstimate, setCustomFileIndexingEstimate] = useState<FileIndexingEstimateResponse | null>(null)
   const [automaticFileIndexingEstimate, setAutomaticFileIndexingEstimate] = useState<FileIndexingEstimateResponse | null>(null)
-  const [estimateTokes, setEstimateTokes] = useState<Pick<IndexingEstimateResponse, 'tokens' | 'total_price'> | null>(null)
 
   const fileIndexingEstimate = (() => {
     return segmentationType === SegmentType.AUTO ? automaticFileIndexingEstimate : customFileIndexingEstimate
@@ -193,13 +194,10 @@ const StepTwo = ({
   const fetchFileIndexingEstimate = async (docForm = DocForm.TEXT) => {
     // eslint-disable-next-line @typescript-eslint/no-use-before-define
     const res = await didFetchFileIndexingEstimate(getFileIndexingEstimateParams(docForm)!)
-    if (segmentationType === SegmentType.CUSTOM) {
+    if (segmentationType === SegmentType.CUSTOM)
       setCustomFileIndexingEstimate(res)
-    }
-    else {
+    else
       setAutomaticFileIndexingEstimate(res)
-      indexType === IndexingType.QUALIFIED && setEstimateTokes({ tokens: res.tokens, total_price: res.total_price })
-    }
   }
 
   const confirmChangeCustomConfig = () => {
@@ -311,6 +309,19 @@ const StepTwo = ({
     defaultModel: rerankDefaultModel,
     currentModel: isRerankDefaultModelVaild,
   } = useModelListAndDefaultModelAndCurrentProviderAndModel(ModelTypeEnum.rerank)
+  const { data: embeddingModelList } = useModelList(ModelTypeEnum.textEmbedding)
+  const { data: defaultEmbeddingModel } = useDefaultModel(ModelTypeEnum.textEmbedding)
+  const [embeddingModel, setEmbeddingModel] = useState<DefaultModel>(
+    currentDataset?.embedding_model
+      ? {
+        provider: currentDataset.embedding_model_provider,
+        model: currentDataset.embedding_model,
+      }
+      : {
+        provider: defaultEmbeddingModel?.provider.provider || '',
+        model: defaultEmbeddingModel?.model || '',
+      },
+  )
   const getCreationParams = () => {
     let params
     if (segmentationType === SegmentType.CUSTOM && overlap > max) {
@@ -325,6 +336,8 @@ const StepTwo = ({
         process_rule: getProcessRule(),
         // eslint-disable-next-line @typescript-eslint/no-use-before-define
         retrieval_model: retrievalConfig, // Readonly. If want to changed, just go to settings page.
+        embedding_model: embeddingModel.model, // Readonly
+        embedding_model_provider: embeddingModel.provider, // Readonly
       } as CreateDocumentReq
     }
     else { // create
@@ -361,6 +374,8 @@ const StepTwo = ({
         doc_language: docLanguage,
 
         retrieval_model: postRetrievalConfig,
+        embedding_model: embeddingModel.model,
+        embedding_model_provider: embeddingModel.provider,
       } as CreateDocumentReq
       if (dataSourceType === DataSourceType.FILE) {
         params.data_source.info_list.file_info_list = {
@@ -614,14 +629,17 @@ const StepTwo = ({
                   <div className={s.formRow}>
                     <div className='w-full'>
                       <div className={s.label}>{t('datasetCreation.stepTwo.maxLength')}</div>
-                      <input
-                        type="number"
-                        className={s.input}
-                        placeholder={t('datasetCreation.stepTwo.maxLength') || ''}
-                        value={max}
-                        min={1}
-                        onChange={e => setMax(parseInt(e.target.value.replace(/^0+/, ''), 10))}
-                      />
+                      <div className='relative w-full'>
+                        <input
+                          type="number"
+                          className={s.input}
+                          placeholder={t('datasetCreation.stepTwo.maxLength') || ''}
+                          value={max}
+                          min={1}
+                          onChange={e => setMax(parseInt(e.target.value.replace(/^0+/, ''), 10))}
+                        />
+                        <div className='absolute top-2.5 right-2.5 text-text-tertiary system-sm-regular'>Tokens</div>
+                      </div>
                     </div>
                   </div>
                   <div className={s.formRow}>
@@ -636,14 +654,17 @@ const StepTwo = ({
                           <RiQuestionLine className='ml-1 w-3.5 h-3.5 text-gray-400' />
                         </TooltipPlus>
                       </div>
-                      <input
-                        type="number"
-                        className={s.input}
-                        placeholder={t('datasetCreation.stepTwo.overlap') || ''}
-                        value={overlap}
-                        min={1}
-                        onChange={e => setOverlap(parseInt(e.target.value.replace(/^0+/, ''), 10))}
-                      />
+                      <div className='relative w-full'>
+                        <input
+                          type="number"
+                          className={s.input}
+                          placeholder={t('datasetCreation.stepTwo.overlap') || ''}
+                          value={overlap}
+                          min={1}
+                          onChange={e => setOverlap(parseInt(e.target.value.replace(/^0+/, ''), 10))}
+                        />
+                        <div className='absolute top-2.5 right-2.5 text-text-tertiary system-sm-regular'>Tokens</div>
+                      </div>
                     </div>
                   </div>
                   <div className={s.formRow}>
@@ -676,7 +697,7 @@ const StepTwo = ({
                     !isAPIKeySet && s.disabled,
                     !hasSetIndexType && indexType === IndexingType.QUALIFIED && s.active,
                     hasSetIndexType && s.disabled,
-                    hasSetIndexType && '!w-full',
+                    hasSetIndexType && '!w-full !min-h-[96px]',
                   )}
                   onClick={() => {
                     if (isAPIKeySet)
@@ -691,16 +712,6 @@ const StepTwo = ({
                       {!hasSetIndexType && <span className={s.recommendTag}>{t('datasetCreation.stepTwo.recommend')}</span>}
                     </div>
                     <div className={s.tip}>{t('datasetCreation.stepTwo.qualifiedTip')}</div>
-                    <div className='pb-0.5 text-xs font-medium text-gray-500'>{t('datasetCreation.stepTwo.emstimateCost')}</div>
-                    {
-                      estimateTokes
-                        ? (
-                          <div className='text-xs font-medium text-gray-800'>{formatNumber(estimateTokes.tokens)} tokens(<span className='text-yellow-500'>${formatNumber(estimateTokes.total_price)}</span>)</div>
-                        )
-                        : (
-                          <div className={s.calculating}>{t('datasetCreation.stepTwo.calculating')}</div>
-                        )
-                    }
                   </div>
                   {!isAPIKeySet && (
                     <div className={s.warningTip}>
@@ -718,7 +729,7 @@ const StepTwo = ({
                     s.indexItem,
                     !hasSetIndexType && indexType === IndexingType.ECONOMICAL && s.active,
                     hasSetIndexType && s.disabled,
-                    hasSetIndexType && '!w-full',
+                    hasSetIndexType && '!w-full !min-h-[96px]',
                   )}
                   onClick={changeToEconomicalType}
                 >
@@ -727,13 +738,11 @@ const StepTwo = ({
                   <div className={s.typeHeader}>
                     <div className={s.title}>{t('datasetCreation.stepTwo.economical')}</div>
                     <div className={s.tip}>{t('datasetCreation.stepTwo.economicalTip')}</div>
-                    <div className='pb-0.5 text-xs font-medium text-gray-500'>{t('datasetCreation.stepTwo.emstimateCost')}</div>
-                    <div className='text-xs font-medium text-gray-800'>0 tokens</div>
                   </div>
                 </div>
               )}
             </div>
-            {hasSetIndexType && (
+            {hasSetIndexType && indexType === IndexingType.ECONOMICAL && (
               <div className='mt-2 text-xs text-gray-500 font-medium'>
                 {t('datasetCreation.stepTwo.indexSettedTip')}
                 <Link className='text-[#155EEF]' href={`/datasets/${datasetId}/settings`}>{t('datasetCreation.stepTwo.datasetSettingLink')}</Link>
@@ -768,12 +777,32 @@ const StepTwo = ({
                 )}
               </div>
             )}
+            {/* Embedding model */}
+            {indexType === IndexingType.QUALIFIED && (
+              <div className='mb-2'>
+                <div className={cn(s.label, datasetId && 'flex justify-between items-center')}>{t('datasetSettings.form.embeddingModel')}</div>
+                <ModelSelector
+                  readonly={!!datasetId}
+                  defaultModel={embeddingModel}
+                  modelList={embeddingModelList}
+                  onSelect={(model: DefaultModel) => {
+                    setEmbeddingModel(model)
+                  }}
+                />
+                {!!datasetId && (
+                  <div className='mt-2 text-xs text-gray-500 font-medium'>
+                    {t('datasetCreation.stepTwo.indexSettedTip')}
+                    <Link className='text-[#155EEF]' href={`/datasets/${datasetId}/settings`}>{t('datasetCreation.stepTwo.datasetSettingLink')}</Link>
+                  </div>
+                )}
+              </div>
+            )}
             {/* Retrieval Method Config */}
             <div>
               {!datasetId
                 ? (
                   <div className={s.label}>
-                    {t('datasetSettings.form.retrievalSetting.title')}
+                    <div className='shrink-0 mr-4'>{t('datasetSettings.form.retrievalSetting.title')}</div>
                     <div className='leading-[18px] text-xs font-normal text-gray-500'>
                       <a target='_blank' rel='noopener noreferrer' href='https://docs.dify.ai/guides/knowledge-base/create-knowledge-and-upload-documents#id-6-retrieval-settings' className='text-[#155eef]'>{t('datasetSettings.form.retrievalSetting.learnMore')}</a>
                       {t('datasetSettings.form.retrievalSetting.longDescription')}

web/app/components/datasets/create/steps-nav-bar/index.tsx

@@ -49,7 +49,7 @@ const StepsNavBar = ({
             key={item}
             className={cn(s.stepItem, s[`step${item}`], step === item && s.active, step > item && s.done, isMobile && 'px-0')}
           >
-            <div className={cn(s.stepNum)}>{item}</div>
+            <div className={cn(s.stepNum)}>{step > item ? '' : item}</div>
             <div className={cn(s.stepName)}>{isMobile ? '' : t(STEP_T_MAP[item])}</div>
           </div>
         ))}

web/app/components/datasets/documents/detail/embedding/index.tsx

@@ -18,9 +18,7 @@ import { ToastContext } from '@/app/components/base/toast'
 import type { FullDocumentDetail, ProcessRuleResponse } from '@/models/datasets'
 import type { CommonResponse } from '@/models/common'
 import { asyncRunSafe, sleep } from '@/utils'
-import { formatNumber } from '@/utils/format'
-import { fetchIndexingStatus as doFetchIndexingStatus, fetchIndexingEstimate, fetchProcessRule, pauseDocIndexing, resumeDocIndexing } from '@/service/datasets'
-import DatasetDetailContext from '@/context/dataset-detail'
+import { fetchIndexingStatus as doFetchIndexingStatus, fetchProcessRule, pauseDocIndexing, resumeDocIndexing } from '@/service/datasets'
 import StopEmbeddingModal from '@/app/components/datasets/create/stop-embedding-modal'
 
 type Props = {
@@ -108,16 +106,14 @@ const RuleDetail: FC<{ sourceData?: ProcessRuleResponse; docName?: string }> = (
   </div>
 }
 
-const EmbeddingDetail: FC<Props> = ({ detail, stopPosition = 'top', datasetId: dstId, documentId: docId, indexingType, detailUpdate }) => {
+const EmbeddingDetail: FC<Props> = ({ detail, stopPosition = 'top', datasetId: dstId, documentId: docId, detailUpdate }) => {
   const onTop = stopPosition === 'top'
   const { t } = useTranslation()
   const { notify } = useContext(ToastContext)
 
   const { datasetId = '', documentId = '' } = useContext(DocumentContext)
-  const { indexingTechnique } = useContext(DatasetDetailContext)
   const localDatasetId = dstId ?? datasetId
   const localDocumentId = docId ?? documentId
-  const localIndexingTechnique = indexingType ?? indexingTechnique
 
   const [indexingStatusDetail, setIndexingStatusDetail] = useState<any>(null)
   const fetchIndexingStatus = async () => {
@@ -160,14 +156,6 @@ const EmbeddingDetail: FC<Props> = ({ detail, stopPosition = 'top', datasetId: d
     }
   }, [startQueryStatus, stopQueryStatus])
 
-  const { data: indexingEstimateDetail, error: indexingEstimateErr } = useSWR({
-    action: 'fetchIndexingEstimate',
-    datasetId: localDatasetId,
-    documentId: localDocumentId,
-  }, apiParams => fetchIndexingEstimate(omit(apiParams, 'action')), {
-    revalidateOnFocus: false,
-  })
-
   const { data: ruleDetail, error: ruleError } = useSWR({
     action: 'fetchProcessRule',
     params: { documentId: localDocumentId },
@@ -250,21 +238,6 @@ const EmbeddingDetail: FC<Props> = ({ detail, stopPosition = 'top', datasetId: d
       </div>
       <div className={s.progressData}>
         <div>{t('datasetDocuments.embedding.segments')} {indexingStatusDetail?.completed_segments}/{indexingStatusDetail?.total_segments} · {percent}%</div>
-        {localIndexingTechnique === 'high_quaility' && (
-          <div className='flex items-center'>
-            <div className={cn(s.commonIcon, s.highIcon)} />
-            {t('datasetDocuments.embedding.highQuality')} · {t('datasetDocuments.embedding.estimate')}
-            <span className={s.tokens}>{formatNumber(indexingEstimateDetail?.tokens || 0)}</span>tokens
-            (<span className={s.price}>${formatNumber(indexingEstimateDetail?.total_price || 0)}</span>)
-          </div>
-        )}
-        {localIndexingTechnique === 'economy' && (
-          <div className='flex items-center'>
-            <div className={cn(s.commonIcon, s.economyIcon)} />
-            {t('datasetDocuments.embedding.economy')} · {t('datasetDocuments.embedding.estimate')}
-            <span className={s.tokens}>0</span>tokens
-          </div>
-        )}
       </div>
       <RuleDetail sourceData={ruleDetail} docName={detail?.name} />
       {!onTop && (

web/app/components/datasets/documents/detail/embedding/style.module.css

@@ -31,7 +31,7 @@
   @apply rounded-r-md;
 }
 .progressData {
-  @apply w-full flex justify-between items-center text-xs text-gray-700;
+  @apply w-full flex items-center text-xs text-gray-700;
 }
 .previewTip {
   @apply pb-1 pt-12 text-gray-900 text-sm font-medium;

web/i18n/en-US/dataset-creation.ts

@@ -86,7 +86,7 @@ const translation = {
     autoDescription: 'Automatically set chunk and preprocessing rules. Unfamiliar users are recommended to select this.',
     custom: 'Custom',
     customDescription: 'Customize chunks rules, chunks length, and preprocessing rules, etc.',
-    separator: 'Segment identifier',
+    separator: 'Delimiter',
     separatorPlaceholder: 'For example, newline (\\\\n) or special separator (such as "***")',
     maxLength: 'Maximum chunk length',
     overlap: 'Chunk overlap',
@@ -135,8 +135,8 @@ const translation = {
     previewSwitchTipStart: 'The current chunk preview is in text format, switching to a question-and-answer format preview will',
     previewSwitchTipEnd: ' consume additional tokens',
     characters: 'characters',
-    indexSettedTip: 'To change the index method, please go to the ',
-    retrivalSettedTip: 'To change the index method, please go to the ',
+    indexSettedTip: 'To change the index method & embedding model, please go to the ',
+    retrivalSettedTip: 'To change the retrieval setting, please go to the ',
     datasetSettingLink: 'Knowledge settings.',
   },
   stepThree: {

web/i18n/ja-JP/dataset-creation.ts

@@ -136,7 +136,7 @@ const translation = {
     previewSwitchTipEnd: ' 追加のトークンが消費されます',
     characters: '文字',
     indexSettedTip: 'インデックス方法を変更するには、',
-    retrivalSettedTip: 'インデックス方法を変更するには、',
+    retrivalSettedTip: '検索方法を変更するには、',
     datasetSettingLink: 'ナレッジ設定',
   },
   stepThree: {

web/i18n/zh-Hans/dataset-creation.ts

@@ -135,7 +135,7 @@ const translation = {
     previewSwitchTipStart: '当前分段预览是文本模式，切换到 Q&A 模式将会',
     previewSwitchTipEnd: '消耗额外的 token',
     characters: '字符',
-    indexSettedTip: '要更改索引方法，请转到',
+    indexSettedTip: '要更改索引方法和 embedding 模型，请转到',
     retrivalSettedTip: '要更改检索方法，请转到',
     datasetSettingLink: '知识库设置。',
   },

web/models/datasets.ts

@@ -226,6 +226,8 @@ export type DocumentReq = {
 export type CreateDocumentReq = DocumentReq & {
   data_source: DataSource
   retrieval_model: RetrievalConfig
+  embedding_model: string
+  embedding_model_provider: string
 }
 
 export type IndexingEstimateParams = DocumentReq & Partial<DataSource> & {