dify/api/services/dataset_service.py

import datetime
import json
import logging
import random
import time
import uuid
from typing import List, Optional, cast

from core.errors.error import LLMBadRequestError, ProviderTokenNotInitError
from core.index.index import IndexBuilder
from core.model_manager import ModelManager
from core.model_runtime.entities.model_entities import ModelType
from core.model_runtime.model_providers.__base.text_embedding_model import TextEmbeddingModel
from events.dataset_event import dataset_was_deleted
from events.document_event import document_was_deleted
from extensions.ext_database import db
from extensions.ext_redis import redis_client
from flask import current_app
from flask_login import current_user
from libs import helper
from models.account import Account
from models.dataset import (AppDatasetJoin, Dataset, DatasetCollectionBinding, DatasetProcessRule, DatasetQuery,
                            Document, DocumentSegment)
from models.model import UploadFile
from models.source import DataSourceBinding
from services.errors.account import NoPermissionError
from services.errors.dataset import DatasetNameDuplicateError
from services.errors.document import DocumentIndexingError
from services.errors.file import FileNotExistsError
from services.vector_service import VectorService
from sqlalchemy import func
from tasks.clean_notion_document_task import clean_notion_document_task
from tasks.deal_dataset_vector_index_task import deal_dataset_vector_index_task
from tasks.delete_segment_from_index_task import delete_segment_from_index_task
from tasks.document_indexing_task import document_indexing_task
from tasks.document_indexing_update_task import document_indexing_update_task
from tasks.recover_document_indexing_task import recover_document_indexing_task


class DatasetService:

    @staticmethod
    def get_datasets(page, per_page, provider="vendor", tenant_id=None, user=None):
        if user:
            permission_filter = db.or_(Dataset.created_by == user.id,
                                       Dataset.permission == 'all_team_members')
        else:
            permission_filter = Dataset.permission == 'all_team_members'
        datasets = Dataset.query.filter(
            db.and_(Dataset.provider == provider, Dataset.tenant_id == tenant_id, permission_filter)) \
            .order_by(Dataset.created_at.desc()) \
            .paginate(
            page=page,
            per_page=per_page,
            max_per_page=100,
            error_out=False
        )

        return datasets.items, datasets.total

    @staticmethod
    def get_process_rules(dataset_id):
        # get the latest process rule
        dataset_process_rule = db.session.query(DatasetProcessRule). \
            filter(DatasetProcessRule.dataset_id == dataset_id). \
            order_by(DatasetProcessRule.created_at.desc()). \
            limit(1). \
            one_or_none()
        if dataset_process_rule:
            mode = dataset_process_rule.mode
            rules = dataset_process_rule.rules_dict
        else:
            mode = DocumentService.DEFAULT_RULES['mode']
            rules = DocumentService.DEFAULT_RULES['rules']
        return {
            'mode': mode,
            'rules': rules
        }

    @staticmethod
    def get_datasets_by_ids(ids, tenant_id):
        datasets = Dataset.query.filter(Dataset.id.in_(ids),
                                        Dataset.tenant_id == tenant_id).paginate(
            page=1, per_page=len(ids), max_per_page=len(ids), error_out=False)
        return datasets.items, datasets.total

    @staticmethod
    def create_empty_dataset(tenant_id: str, name: str, indexing_technique: Optional[str], account: Account):
        # check if dataset name already exists
        if Dataset.query.filter_by(name=name, tenant_id=tenant_id).first():
            raise DatasetNameDuplicateError(
                f'Dataset with name {name} already exists.')
        embedding_model = None
        if indexing_technique == 'high_quality':
            model_manager = ModelManager()
            embedding_model = model_manager.get_default_model_instance(
                tenant_id=tenant_id,
                model_type=ModelType.TEXT_EMBEDDING
            )
        dataset = Dataset(name=name, indexing_technique=indexing_technique)
        # dataset = Dataset(name=name, provider=provider, config=config)
        dataset.created_by = account.id
        dataset.updated_by = account.id
        dataset.tenant_id = tenant_id
        dataset.embedding_model_provider = embedding_model.provider if embedding_model else None
        dataset.embedding_model = embedding_model.model if embedding_model else None
        db.session.add(dataset)
        db.session.commit()
        return dataset

    @staticmethod
    def get_dataset(dataset_id):
        dataset = Dataset.query.filter_by(
            id=dataset_id
        ).first()
        if dataset is None:
            return None
        else:
            return dataset

    @staticmethod
    def check_dataset_model_setting(dataset):
        if dataset.indexing_technique == 'high_quality':
            try:
                model_manager = ModelManager()
                model_manager.get_model_instance(
                    tenant_id=dataset.tenant_id,
                    provider=dataset.embedding_model_provider,
                    model_type=ModelType.TEXT_EMBEDDING,
                    model=dataset.embedding_model
                )
            except LLMBadRequestError:
                raise ValueError(
                    f"No Embedding Model available. Please configure a valid provider "
                    f"in the Settings -> Model Provider.")
            except ProviderTokenNotInitError as ex:
                raise ValueError(f"The dataset in unavailable, due to: "
                                 f"{ex.description}")

    @staticmethod
    def update_dataset(dataset_id, data, user):
        filtered_data = {k: v for k, v in data.items() if v is not None or k == 'description'}
        dataset = DatasetService.get_dataset(dataset_id)
        DatasetService.check_dataset_permission(dataset, user)
        action = None
        if dataset.indexing_technique != data['indexing_technique']:
            # if update indexing_technique
            if data['indexing_technique'] == 'economy':
                action = 'remove'
                filtered_data['embedding_model'] = None
                filtered_data['embedding_model_provider'] = None
                filtered_data['collection_binding_id'] = None
            elif data['indexing_technique'] == 'high_quality':
                action = 'add'
                # get embedding model setting
                try:
                    model_manager = ModelManager()
                    embedding_model = model_manager.get_default_model_instance(
                        tenant_id=current_user.current_tenant_id,
                        model_type=ModelType.TEXT_EMBEDDING
                    )
                    filtered_data['embedding_model'] = embedding_model.model
                    filtered_data['embedding_model_provider'] = embedding_model.provider
                    dataset_collection_binding = DatasetCollectionBindingService.get_dataset_collection_binding(
                        embedding_model.provider,
                        embedding_model.model
                    )
                    filtered_data['collection_binding_id'] = dataset_collection_binding.id
                except LLMBadRequestError:
                    raise ValueError(
                        f"No Embedding Model available. Please configure a valid provider "
                        f"in the Settings -> Model Provider.")
                except ProviderTokenNotInitError as ex:
                    raise ValueError(ex.description)

        filtered_data['updated_by'] = user.id
        filtered_data['updated_at'] = datetime.datetime.now()

        # update Retrieval model
        filtered_data['retrieval_model'] = data['retrieval_model']

        dataset.query.filter_by(id=dataset_id).update(filtered_data)

        db.session.commit()
        if action:
            deal_dataset_vector_index_task.delay(dataset_id, action)
        return dataset

    @staticmethod
    def delete_dataset(dataset_id, user):
        # todo: cannot delete dataset if it is being processed

        dataset = DatasetService.get_dataset(dataset_id)

        if dataset is None:
            return False

        DatasetService.check_dataset_permission(dataset, user)

        dataset_was_deleted.send(dataset)

        db.session.delete(dataset)
        db.session.commit()
        return True

    @staticmethod
    def check_dataset_permission(dataset, user):
        if dataset.tenant_id != user.current_tenant_id:
            logging.debug(
                f'User {user.id} does not have permission to access dataset {dataset.id}')
            raise NoPermissionError(
                'You do not have permission to access this dataset.')
        if dataset.permission == 'only_me' and dataset.created_by != user.id:
            logging.debug(
                f'User {user.id} does not have permission to access dataset {dataset.id}')
            raise NoPermissionError(
                'You do not have permission to access this dataset.')

    @staticmethod
    def get_dataset_queries(dataset_id: str, page: int, per_page: int):
        dataset_queries = DatasetQuery.query.filter_by(dataset_id=dataset_id) \
            .order_by(db.desc(DatasetQuery.created_at)) \
            .paginate(
            page=page, per_page=per_page, max_per_page=100, error_out=False
        )
        return dataset_queries.items, dataset_queries.total

    @staticmethod
    def get_related_apps(dataset_id: str):
        return AppDatasetJoin.query.filter(AppDatasetJoin.dataset_id == dataset_id) \
            .order_by(db.desc(AppDatasetJoin.created_at)).all()


class DocumentService:
    DEFAULT_RULES = {
        'mode': 'custom',
        'rules': {
            'pre_processing_rules': [
                {'id': 'remove_extra_spaces', 'enabled': True},
                {'id': 'remove_urls_emails', 'enabled': False}
            ],
            'segmentation': {
                'delimiter': '\n',
                'max_tokens': 500
            }
        }
    }

    DOCUMENT_METADATA_SCHEMA = {
        "book": {
            "title": str,
            "language": str,
            "author": str,
            "publisher": str,
            "publication_date": str,
            "isbn": str,
            "category": str,
        },
        "web_page": {
            "title": str,
            "url": str,
            "language": str,
            "publish_date": str,
            "author/publisher": str,
            "topic/keywords": str,
            "description": str,
        },
        "paper": {
            "title": str,
            "language": str,
            "author": str,
            "publish_date": str,
            "journal/conference_name": str,
            "volume/issue/page_numbers": str,
            "doi": str,
            "topic/keywords": str,
            "abstract": str,
        },
        "social_media_post": {
            "platform": str,
            "author/username": str,
            "publish_date": str,
            "post_url": str,
            "topic/tags": str,
        },
        "wikipedia_entry": {
            "title": str,
            "language": str,
            "web_page_url": str,
            "last_edit_date": str,
            "editor/contributor": str,
            "summary/introduction": str,
        },
        "personal_document": {
            "title": str,
            "author": str,
            "creation_date": str,
            "last_modified_date": str,
            "document_type": str,
            "tags/category": str,
        },
        "business_document": {
            "title": str,
            "author": str,
            "creation_date": str,
            "last_modified_date": str,
            "document_type": str,
            "department/team": str,
        },
        "im_chat_log": {
            "chat_platform": str,
            "chat_participants/group_name": str,
            "start_date": str,
            "end_date": str,
            "summary": str,
        },
        "synced_from_notion": {
            "title": str,
            "language": str,
            "author/creator": str,
            "creation_date": str,
            "last_modified_date": str,
            "notion_page_link": str,
            "category/tags": str,
            "description": str,
        },
        "synced_from_github": {
            "repository_name": str,
            "repository_description": str,
            "repository_owner/organization": str,
            "code_filename": str,
            "code_file_path": str,
            "programming_language": str,
            "github_link": str,
            "open_source_license": str,
            "commit_date": str,
            "commit_author": str,
        },
        "others": dict
    }

    @staticmethod
    def get_document(dataset_id: str, document_id: str) -> Optional[Document]:
        document = db.session.query(Document).filter(
            Document.id == document_id,
            Document.dataset_id == dataset_id
        ).first()

        return document

    @staticmethod
    def get_document_by_id(document_id: str) -> Optional[Document]:
        document = db.session.query(Document).filter(
            Document.id == document_id
        ).first()

        return document

    @staticmethod
    def get_document_by_dataset_id(dataset_id: str) -> List[Document]:
        documents = db.session.query(Document).filter(
            Document.dataset_id == dataset_id,
            Document.enabled == True
        ).all()

        return documents

    @staticmethod
    def get_batch_documents(dataset_id: str, batch: str) -> List[Document]:
        documents = db.session.query(Document).filter(
            Document.batch == batch,
            Document.dataset_id == dataset_id,
            Document.tenant_id == current_user.current_tenant_id
        ).all()

        return documents

    @staticmethod
    def get_document_file_detail(file_id: str):
        file_detail = db.session.query(UploadFile). \
            filter(UploadFile.id == file_id). \
            one_or_none()
        return file_detail

    @staticmethod
    def check_archived(document):
        if document.archived:
            return True
        else:
            return False

    @staticmethod
    def delete_document(document):
        # trigger document_was_deleted signal
        document_was_deleted.send(document.id, dataset_id=document.dataset_id)

        db.session.delete(document)
        db.session.commit()

    @staticmethod
    def pause_document(document):
        if document.indexing_status not in ["waiting", "parsing", "cleaning", "splitting", "indexing"]:
            raise DocumentIndexingError()
        # update document to be paused
        document.is_paused = True
        document.paused_by = current_user.id
        document.paused_at = datetime.datetime.utcnow()

        db.session.add(document)
        db.session.commit()
        # set document paused flag
        indexing_cache_key = 'document_{}_is_paused'.format(document.id)
        redis_client.setnx(indexing_cache_key, "True")

    @staticmethod
    def recover_document(document):
        if not document.is_paused:
            raise DocumentIndexingError()
        # update document to be recover
        document.is_paused = False
        document.paused_by = None
        document.paused_at = None

        db.session.add(document)
        db.session.commit()
        # delete paused flag
        indexing_cache_key = 'document_{}_is_paused'.format(document.id)
        redis_client.delete(indexing_cache_key)
        # trigger async task
        recover_document_indexing_task.delay(document.dataset_id, document.id)

    @staticmethod
    def get_documents_position(dataset_id):
        document = Document.query.filter_by(dataset_id=dataset_id).order_by(Document.position.desc()).first()
        if document:
            return document.position + 1
        else:
            return 1

    @staticmethod
    def save_document_with_dataset_id(dataset: Dataset, document_data: dict,
                                      account: Account, dataset_process_rule: Optional[DatasetProcessRule] = None,
                                      created_from: str = 'web'):

        # check document limit
        if current_app.config['EDITION'] == 'CLOUD':
            if 'original_document_id' not in document_data or not document_data['original_document_id']:
                count = 0
                if document_data["data_source"]["type"] == "upload_file":
                    upload_file_list = document_data["data_source"]["info_list"]['file_info_list']['file_ids']
                    count = len(upload_file_list)
                elif document_data["data_source"]["type"] == "notion_import":
                    notion_info_list = document_data["data_source"]['info_list']['notion_info_list']
                    for notion_info in notion_info_list:
                        count = count + len(notion_info['pages'])
        # if dataset is empty, update dataset data_source_type
        if not dataset.data_source_type:
            dataset.data_source_type = document_data["data_source"]["type"]

        if not dataset.indexing_technique:
            if 'indexing_technique' not in document_data \
                    or document_data['indexing_technique'] not in Dataset.INDEXING_TECHNIQUE_LIST:
                raise ValueError("Indexing technique is required")

            dataset.indexing_technique = document_data["indexing_technique"]
            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.embedding_model = embedding_model.model
                dataset.embedding_model_provider = embedding_model.provider
                dataset_collection_binding = DatasetCollectionBindingService.get_dataset_collection_binding(
                    embedding_model.provider,
                    embedding_model.model
                )
                dataset.collection_binding_id = dataset_collection_binding.id
                if not dataset.retrieval_model:
                    default_retrieval_model = {
                        'search_method': 'semantic_search',
                        'reranking_enable': False,
                        'reranking_model': {
                            'reranking_provider_name': '',
                            'reranking_model_name': ''
                        },
                        'top_k': 2,
                        'score_threshold_enabled': False
                    }

                    dataset.retrieval_model = document_data.get('retrieval_model') if document_data.get(
                        'retrieval_model') else default_retrieval_model

        documents = []
        batch = time.strftime('%Y%m%d%H%M%S') + str(random.randint(100000, 999999))
        if 'original_document_id' in document_data and document_data["original_document_id"]:
            document = DocumentService.update_document_with_dataset_id(dataset, document_data, account)
            documents.append(document)
        else:
            # save process rule
            if not dataset_process_rule:
                process_rule = document_data["process_rule"]
                if process_rule["mode"] == "custom":
                    dataset_process_rule = DatasetProcessRule(
                        dataset_id=dataset.id,
                        mode=process_rule["mode"],
                        rules=json.dumps(process_rule["rules"]),
                        created_by=account.id
                    )
                elif process_rule["mode"] == "automatic":
                    dataset_process_rule = DatasetProcessRule(
                        dataset_id=dataset.id,
                        mode=process_rule["mode"],
                        rules=json.dumps(DatasetProcessRule.AUTOMATIC_RULES),
                        created_by=account.id
                    )
                db.session.add(dataset_process_rule)
                db.session.commit()
            position = DocumentService.get_documents_position(dataset.id)
            document_ids = []
            if document_data["data_source"]["type"] == "upload_file":
                upload_file_list = document_data["data_source"]["info_list"]['file_info_list']['file_ids']
                for file_id in upload_file_list:
                    file = db.session.query(UploadFile).filter(
                        UploadFile.tenant_id == dataset.tenant_id,
                        UploadFile.id == file_id
                    ).first()

                    # raise error if file not found
                    if not file:
                        raise FileNotExistsError()

                    file_name = file.name
                    data_source_info = {
                        "upload_file_id": file_id,
                    }
                    document = DocumentService.build_document(dataset, dataset_process_rule.id,
                                                              document_data["data_source"]["type"],
                                                              document_data["doc_form"],
                                                              document_data["doc_language"],
                                                              data_source_info, created_from, position,
                                                              account, file_name, batch)
                    db.session.add(document)
                    db.session.flush()
                    document_ids.append(document.id)
                    documents.append(document)
                    position += 1
            elif document_data["data_source"]["type"] == "notion_import":
                notion_info_list = document_data["data_source"]['info_list']['notion_info_list']
                exist_page_ids = []
                exist_document = dict()
                documents = Document.query.filter_by(
                    dataset_id=dataset.id,
                    tenant_id=current_user.current_tenant_id,
                    data_source_type='notion_import',
                    enabled=True
                ).all()
                if documents:
                    for document in documents:
                        data_source_info = json.loads(document.data_source_info)
                        exist_page_ids.append(data_source_info['notion_page_id'])
                        exist_document[data_source_info['notion_page_id']] = document.id
                for notion_info in notion_info_list:
                    workspace_id = notion_info['workspace_id']
                    data_source_binding = DataSourceBinding.query.filter(
                        db.and_(
                            DataSourceBinding.tenant_id == current_user.current_tenant_id,
                            DataSourceBinding.provider == 'notion',
                            DataSourceBinding.disabled == False,
                            DataSourceBinding.source_info['workspace_id'] == f'"{workspace_id}"'
                        )
                    ).first()
                    if not data_source_binding:
                        raise ValueError('Data source binding not found.')
                    for page in notion_info['pages']:
                        if page['page_id'] not in exist_page_ids:
                            data_source_info = {
                                "notion_workspace_id": workspace_id,
                                "notion_page_id": page['page_id'],
                                "notion_page_icon": page['page_icon'],
                                "type": page['type']
                            }
                            document = DocumentService.build_document(dataset, dataset_process_rule.id,
                                                                      document_data["data_source"]["type"],
                                                                      document_data["doc_form"],
                                                                      document_data["doc_language"],
                                                                      data_source_info, created_from, position,
                                                                      account, page['page_name'], batch)
                            db.session.add(document)
                            db.session.flush()
                            document_ids.append(document.id)
                            documents.append(document)
                            position += 1
                        else:
                            exist_document.pop(page['page_id'])
                # delete not selected documents
                if len(exist_document) > 0:
                    clean_notion_document_task.delay(list(exist_document.values()), dataset.id)
            db.session.commit()

            # trigger async task
            document_indexing_task.delay(dataset.id, document_ids)

        return documents, batch

    @staticmethod
    def build_document(dataset: Dataset, process_rule_id: str, data_source_type: str, document_form: str,
                       document_language: str, data_source_info: dict, created_from: str, position: int,
                       account: Account,
                       name: str, batch: str):
        document = Document(
            tenant_id=dataset.tenant_id,
            dataset_id=dataset.id,
            position=position,
            data_source_type=data_source_type,
            data_source_info=json.dumps(data_source_info),
            dataset_process_rule_id=process_rule_id,
            batch=batch,
            name=name,
            created_from=created_from,
            created_by=account.id,
            doc_form=document_form,
            doc_language=document_language
        )
        return document

    @staticmethod
    def get_tenant_documents_count():
        documents_count = Document.query.filter(Document.completed_at.isnot(None),
                                                Document.enabled == True,
                                                Document.archived == False,
                                                Document.tenant_id == current_user.current_tenant_id).count()
        return documents_count

    @staticmethod
    def update_document_with_dataset_id(dataset: Dataset, document_data: dict,
                                        account: Account, dataset_process_rule: Optional[DatasetProcessRule] = None,
                                        created_from: str = 'web'):
        DatasetService.check_dataset_model_setting(dataset)
        document = DocumentService.get_document(dataset.id, document_data["original_document_id"])
        if document.display_status != 'available':
            raise ValueError("Document is not available")
        # update document name
        if 'name' in document_data and document_data['name']:
            document.name = document_data['name']
        # save process rule
        if 'process_rule' in document_data and document_data['process_rule']:
            process_rule = document_data["process_rule"]
            if process_rule["mode"] == "custom":
                dataset_process_rule = DatasetProcessRule(
                    dataset_id=dataset.id,
                    mode=process_rule["mode"],
                    rules=json.dumps(process_rule["rules"]),
                    created_by=account.id
                )
            elif process_rule["mode"] == "automatic":
                dataset_process_rule = DatasetProcessRule(
                    dataset_id=dataset.id,
                    mode=process_rule["mode"],
                    rules=json.dumps(DatasetProcessRule.AUTOMATIC_RULES),
                    created_by=account.id
                )
            db.session.add(dataset_process_rule)
            db.session.commit()
            document.dataset_process_rule_id = dataset_process_rule.id
        # update document data source
        if 'data_source' in document_data and document_data['data_source']:
            file_name = ''
            data_source_info = {}
            if document_data["data_source"]["type"] == "upload_file":
                upload_file_list = document_data["data_source"]["info_list"]['file_info_list']['file_ids']
                for file_id in upload_file_list:
                    file = db.session.query(UploadFile).filter(
                        UploadFile.tenant_id == dataset.tenant_id,
                        UploadFile.id == file_id
                    ).first()

                    # raise error if file not found
                    if not file:
                        raise FileNotExistsError()

                    file_name = file.name
                    data_source_info = {
                        "upload_file_id": file_id,
                    }
            elif document_data["data_source"]["type"] == "notion_import":
                notion_info_list = document_data["data_source"]['info_list']['notion_info_list']
                for notion_info in notion_info_list:
                    workspace_id = notion_info['workspace_id']
                    data_source_binding = DataSourceBinding.query.filter(
                        db.and_(
                            DataSourceBinding.tenant_id == current_user.current_tenant_id,
                            DataSourceBinding.provider == 'notion',
                            DataSourceBinding.disabled == False,
                            DataSourceBinding.source_info['workspace_id'] == f'"{workspace_id}"'
                        )
                    ).first()
                    if not data_source_binding:
                        raise ValueError('Data source binding not found.')
                    for page in notion_info['pages']:
                        data_source_info = {
                            "notion_workspace_id": workspace_id,
                            "notion_page_id": page['page_id'],
                            "notion_page_icon": page['page_icon'],
                            "type": page['type']
                        }
            document.data_source_type = document_data["data_source"]["type"]
            document.data_source_info = json.dumps(data_source_info)
            document.name = file_name
        # update document to be waiting
        document.indexing_status = 'waiting'
        document.completed_at = None
        document.processing_started_at = None
        document.parsing_completed_at = None
        document.cleaning_completed_at = None
        document.splitting_completed_at = None
        document.updated_at = datetime.datetime.utcnow()
        document.created_from = created_from
        document.doc_form = document_data['doc_form']
        db.session.add(document)
        db.session.commit()
        # update document segment
        update_params = {
            DocumentSegment.status: 're_segment'
        }
        DocumentSegment.query.filter_by(document_id=document.id).update(update_params)
        db.session.commit()
        # trigger async task
        document_indexing_update_task.delay(document.dataset_id, document.id)

        return document

    @staticmethod
    def save_document_without_dataset_id(tenant_id: str, document_data: dict, account: Account):
        count = 0
        if document_data["data_source"]["type"] == "upload_file":
            upload_file_list = document_data["data_source"]["info_list"]['file_info_list']['file_ids']
            count = len(upload_file_list)
        elif document_data["data_source"]["type"] == "notion_import":
            notion_info_list = document_data["data_source"]['info_list']['notion_info_list']
            for notion_info in notion_info_list:
                count = count + len(notion_info['pages'])

        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
            )
            dataset_collection_binding_id = dataset_collection_binding.id
            if 'retrieval_model' in document_data and document_data['retrieval_model']:
                retrieval_model = document_data['retrieval_model']
            else:
                default_retrieval_model = {
                    'search_method': 'semantic_search',
                    'reranking_enable': False,
                    'reranking_model': {
                        'reranking_provider_name': '',
                        'reranking_model_name': ''
                    },
                    'top_k': 2,
                    'score_threshold_enabled': False
                }
                retrieval_model = default_retrieval_model
        # save dataset
        dataset = Dataset(
            tenant_id=tenant_id,
            name='',
            data_source_type=document_data["data_source"]["type"],
            indexing_technique=document_data["indexing_technique"],
            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,
            collection_binding_id=dataset_collection_binding_id,
            retrieval_model=retrieval_model
        )

        db.session.add(dataset)
        db.session.flush()

        documents, batch = DocumentService.save_document_with_dataset_id(dataset, document_data, account)

        cut_length = 18
        cut_name = documents[0].name[:cut_length]
        dataset.name = cut_name + '...'
        dataset.description = 'useful for when you want to answer queries about the ' + documents[0].name
        db.session.commit()

        return dataset, documents, batch

    @classmethod
    def document_create_args_validate(cls, args: dict):
        if 'original_document_id' not in args or not args['original_document_id']:
            DocumentService.data_source_args_validate(args)
            DocumentService.process_rule_args_validate(args)
        else:
            if ('data_source' not in args and not args['data_source']) \
                    and ('process_rule' not in args and not args['process_rule']):
                raise ValueError("Data source or Process rule is required")
            else:
                if 'data_source' in args and args['data_source']:
                    DocumentService.data_source_args_validate(args)
                if 'process_rule' in args and args['process_rule']:
                    DocumentService.process_rule_args_validate(args)

    @classmethod
    def data_source_args_validate(cls, args: dict):
        if 'data_source' not in args or not args['data_source']:
            raise ValueError("Data source is required")

        if not isinstance(args['data_source'], dict):
            raise ValueError("Data source is invalid")

        if 'type' not in args['data_source'] or not args['data_source']['type']:
            raise ValueError("Data source type is required")

        if args['data_source']['type'] not in Document.DATA_SOURCES:
            raise ValueError("Data source type is invalid")

        if 'info_list' not in args['data_source'] or not args['data_source']['info_list']:
            raise ValueError("Data source info is required")

        if args['data_source']['type'] == 'upload_file':
            if 'file_info_list' not in args['data_source']['info_list'] or not args['data_source']['info_list'][
                'file_info_list']:
                raise ValueError("File source info is required")
        if args['data_source']['type'] == 'notion_import':
            if 'notion_info_list' not in args['data_source']['info_list'] or not args['data_source']['info_list'][
                'notion_info_list']:
                raise ValueError("Notion source info is required")

    @classmethod
    def process_rule_args_validate(cls, args: dict):
        if 'process_rule' not in args or not args['process_rule']:
            raise ValueError("Process rule is required")

        if not isinstance(args['process_rule'], dict):
            raise ValueError("Process rule is invalid")

        if 'mode' not in args['process_rule'] or not args['process_rule']['mode']:
            raise ValueError("Process rule mode is required")

        if args['process_rule']['mode'] not in DatasetProcessRule.MODES:
            raise ValueError("Process rule mode is invalid")

        if args['process_rule']['mode'] == 'automatic':
            args['process_rule']['rules'] = {}
        else:
            if 'rules' not in args['process_rule'] or not args['process_rule']['rules']:
                raise ValueError("Process rule rules is required")

            if not isinstance(args['process_rule']['rules'], dict):
                raise ValueError("Process rule rules is invalid")

            if 'pre_processing_rules' not in args['process_rule']['rules'] \
                    or args['process_rule']['rules']['pre_processing_rules'] is None:
                raise ValueError("Process rule pre_processing_rules is required")

            if not isinstance(args['process_rule']['rules']['pre_processing_rules'], list):
                raise ValueError("Process rule pre_processing_rules is invalid")

            unique_pre_processing_rule_dicts = {}
            for pre_processing_rule in args['process_rule']['rules']['pre_processing_rules']:
                if 'id' not in pre_processing_rule or not pre_processing_rule['id']:
                    raise ValueError("Process rule pre_processing_rules id is required")

                if pre_processing_rule['id'] not in DatasetProcessRule.PRE_PROCESSING_RULES:
                    raise ValueError("Process rule pre_processing_rules id is invalid")

                if 'enabled' not in pre_processing_rule or pre_processing_rule['enabled'] is None:
                    raise ValueError("Process rule pre_processing_rules enabled is required")

                if not isinstance(pre_processing_rule['enabled'], bool):
                    raise ValueError("Process rule pre_processing_rules enabled is invalid")

                unique_pre_processing_rule_dicts[pre_processing_rule['id']] = pre_processing_rule

            args['process_rule']['rules']['pre_processing_rules'] = list(unique_pre_processing_rule_dicts.values())

            if 'segmentation' not in args['process_rule']['rules'] \
                    or args['process_rule']['rules']['segmentation'] is None:
                raise ValueError("Process rule segmentation is required")

            if not isinstance(args['process_rule']['rules']['segmentation'], dict):
                raise ValueError("Process rule segmentation is invalid")

            if 'separator' not in args['process_rule']['rules']['segmentation'] \
                    or not args['process_rule']['rules']['segmentation']['separator']:
                raise ValueError("Process rule segmentation separator is required")

            if not isinstance(args['process_rule']['rules']['segmentation']['separator'], str):
                raise ValueError("Process rule segmentation separator is invalid")

            if 'max_tokens' not in args['process_rule']['rules']['segmentation'] \
                    or not args['process_rule']['rules']['segmentation']['max_tokens']:
                raise ValueError("Process rule segmentation max_tokens is required")

            if not isinstance(args['process_rule']['rules']['segmentation']['max_tokens'], int):
                raise ValueError("Process rule segmentation max_tokens is invalid")

    @classmethod
    def estimate_args_validate(cls, args: dict):
        if 'info_list' not in args or not args['info_list']:
            raise ValueError("Data source info is required")

        if not isinstance(args['info_list'], dict):
            raise ValueError("Data info is invalid")

        if 'process_rule' not in args or not args['process_rule']:
            raise ValueError("Process rule is required")

        if not isinstance(args['process_rule'], dict):
            raise ValueError("Process rule is invalid")

        if 'mode' not in args['process_rule'] or not args['process_rule']['mode']:
            raise ValueError("Process rule mode is required")

        if args['process_rule']['mode'] not in DatasetProcessRule.MODES:
            raise ValueError("Process rule mode is invalid")

        if args['process_rule']['mode'] == 'automatic':
            args['process_rule']['rules'] = {}
        else:
            if 'rules' not in args['process_rule'] or not args['process_rule']['rules']:
                raise ValueError("Process rule rules is required")

            if not isinstance(args['process_rule']['rules'], dict):
                raise ValueError("Process rule rules is invalid")

            if 'pre_processing_rules' not in args['process_rule']['rules'] \
                    or args['process_rule']['rules']['pre_processing_rules'] is None:
                raise ValueError("Process rule pre_processing_rules is required")

            if not isinstance(args['process_rule']['rules']['pre_processing_rules'], list):
                raise ValueError("Process rule pre_processing_rules is invalid")

            unique_pre_processing_rule_dicts = {}
            for pre_processing_rule in args['process_rule']['rules']['pre_processing_rules']:
                if 'id' not in pre_processing_rule or not pre_processing_rule['id']:
                    raise ValueError("Process rule pre_processing_rules id is required")

                if pre_processing_rule['id'] not in DatasetProcessRule.PRE_PROCESSING_RULES:
                    raise ValueError("Process rule pre_processing_rules id is invalid")

                if 'enabled' not in pre_processing_rule or pre_processing_rule['enabled'] is None:
                    raise ValueError("Process rule pre_processing_rules enabled is required")

                if not isinstance(pre_processing_rule['enabled'], bool):
                    raise ValueError("Process rule pre_processing_rules enabled is invalid")

                unique_pre_processing_rule_dicts[pre_processing_rule['id']] = pre_processing_rule

            args['process_rule']['rules']['pre_processing_rules'] = list(unique_pre_processing_rule_dicts.values())

            if 'segmentation' not in args['process_rule']['rules'] \
                    or args['process_rule']['rules']['segmentation'] is None:
                raise ValueError("Process rule segmentation is required")

            if not isinstance(args['process_rule']['rules']['segmentation'], dict):
                raise ValueError("Process rule segmentation is invalid")

            if 'separator' not in args['process_rule']['rules']['segmentation'] \
                    or not args['process_rule']['rules']['segmentation']['separator']:
                raise ValueError("Process rule segmentation separator is required")

            if not isinstance(args['process_rule']['rules']['segmentation']['separator'], str):
                raise ValueError("Process rule segmentation separator is invalid")

            if 'max_tokens' not in args['process_rule']['rules']['segmentation'] \
                    or not args['process_rule']['rules']['segmentation']['max_tokens']:
                raise ValueError("Process rule segmentation max_tokens is required")

            if not isinstance(args['process_rule']['rules']['segmentation']['max_tokens'], int):
                raise ValueError("Process rule segmentation max_tokens is invalid")


class SegmentService:
    @classmethod
    def segment_create_args_validate(cls, args: dict, document: Document):
        if document.doc_form == 'qa_model':
            if 'answer' not in args or not args['answer']:
                raise ValueError("Answer is required")
            if not args['answer'].strip():
                raise ValueError("Answer is empty")
        if 'content' not in args or not args['content'] or not args['content'].strip():
            raise ValueError("Content is empty")

    @classmethod
    def create_segment(cls, args: dict, document: Document, dataset: Dataset):
        content = args['content']
        doc_id = str(uuid.uuid4())
        segment_hash = helper.generate_text_hash(content)
        tokens = 0
        if dataset.indexing_technique == 'high_quality':
            model_manager = ModelManager()
            embedding_model = model_manager.get_model_instance(
                tenant_id=current_user.current_tenant_id,
                provider=dataset.embedding_model_provider,
                model_type=ModelType.TEXT_EMBEDDING,
                model=dataset.embedding_model
            )
            # calc embedding use tokens
            model_type_instance = cast(TextEmbeddingModel, embedding_model.model_type_instance)
            tokens = model_type_instance.get_num_tokens(
                model=embedding_model.model,
                credentials=embedding_model.credentials,
                texts=[content]
            )
        max_position = db.session.query(func.max(DocumentSegment.position)).filter(
            DocumentSegment.document_id == document.id
        ).scalar()
        segment_document = DocumentSegment(
            tenant_id=current_user.current_tenant_id,
            dataset_id=document.dataset_id,
            document_id=document.id,
            index_node_id=doc_id,
            index_node_hash=segment_hash,
            position=max_position + 1 if max_position else 1,
            content=content,
            word_count=len(content),
            tokens=tokens,
            status='completed',
            indexing_at=datetime.datetime.utcnow(),
            completed_at=datetime.datetime.utcnow(),
            created_by=current_user.id
        )
        if document.doc_form == 'qa_model':
            segment_document.answer = args['answer']

        db.session.add(segment_document)
        db.session.commit()

        # save vector index
        try:
            VectorService.create_segment_vector(args['keywords'], segment_document, dataset)
        except Exception as e:
            logging.exception("create segment index failed")
            segment_document.enabled = False
            segment_document.disabled_at = datetime.datetime.utcnow()
            segment_document.status = 'error'
            segment_document.error = str(e)
            db.session.commit()
        segment = db.session.query(DocumentSegment).filter(DocumentSegment.id == segment_document.id).first()
        return segment

    @classmethod
    def multi_create_segment(cls, segments: list, document: Document, dataset: Dataset):
        embedding_model = None
        if dataset.indexing_technique == 'high_quality':
            model_manager = ModelManager()
            embedding_model = model_manager.get_model_instance(
                tenant_id=current_user.current_tenant_id,
                provider=dataset.embedding_model_provider,
                model_type=ModelType.TEXT_EMBEDDING,
                model=dataset.embedding_model
            )
        max_position = db.session.query(func.max(DocumentSegment.position)).filter(
            DocumentSegment.document_id == document.id
        ).scalar()
        pre_segment_data_list = []
        segment_data_list = []
        for segment_item in segments:
            content = segment_item['content']
            doc_id = str(uuid.uuid4())
            segment_hash = helper.generate_text_hash(content)
            tokens = 0
            if dataset.indexing_technique == 'high_quality' and embedding_model:
                # calc embedding use tokens
                model_type_instance = cast(TextEmbeddingModel, embedding_model.model_type_instance)
                tokens = model_type_instance.get_num_tokens(
                    model=embedding_model.model,
                    credentials=embedding_model.credentials,
                    texts=[content]
                )
            segment_document = DocumentSegment(
                tenant_id=current_user.current_tenant_id,
                dataset_id=document.dataset_id,
                document_id=document.id,
                index_node_id=doc_id,
                index_node_hash=segment_hash,
                position=max_position + 1 if max_position else 1,
                content=content,
                word_count=len(content),
                tokens=tokens,
                status='completed',
                indexing_at=datetime.datetime.utcnow(),
                completed_at=datetime.datetime.utcnow(),
                created_by=current_user.id
            )
            if document.doc_form == 'qa_model':
                segment_document.answer = segment_item['answer']
            db.session.add(segment_document)
            segment_data_list.append(segment_document)
            pre_segment_data = {
                'segment': segment_document,
                'keywords': segment_item['keywords']
            }
            pre_segment_data_list.append(pre_segment_data)

        try:
            # save vector index
            VectorService.multi_create_segment_vector(pre_segment_data_list, dataset)
        except Exception as e:
            logging.exception("create segment index failed")
            for segment_document in segment_data_list:
                segment_document.enabled = False
                segment_document.disabled_at = datetime.datetime.utcnow()
                segment_document.status = 'error'
                segment_document.error = str(e)
        db.session.commit()
        return segment_data_list

    @classmethod
    def update_segment(cls, args: dict, segment: DocumentSegment, document: Document, dataset: Dataset):
        indexing_cache_key = 'segment_{}_indexing'.format(segment.id)
        cache_result = redis_client.get(indexing_cache_key)
        if cache_result is not None:
            raise ValueError("Segment is indexing, please try again later")
        try:
            content = args['content']
            if segment.content == content:
                if document.doc_form == 'qa_model':
                    segment.answer = args['answer']
                if 'keywords' in args and args['keywords']:
                    segment.keywords = args['keywords']
                if'enabled' in args and args['enabled'] is not None:
                    segment.enabled = args['enabled']
                db.session.add(segment)
                db.session.commit()
                # update segment index task
                if args['keywords']:
                    kw_index = IndexBuilder.get_index(dataset, 'economy')
                    # delete from keyword index
                    kw_index.delete_by_ids([segment.index_node_id])
                    # save keyword index
                    kw_index.update_segment_keywords_index(segment.index_node_id, segment.keywords)
            else:
                segment_hash = helper.generate_text_hash(content)
                tokens = 0
                if dataset.indexing_technique == 'high_quality':
                    model_manager = ModelManager()
                    embedding_model = model_manager.get_model_instance(
                        tenant_id=current_user.current_tenant_id,
                        provider=dataset.embedding_model_provider,
                        model_type=ModelType.TEXT_EMBEDDING,
                        model=dataset.embedding_model
                    )

                    # calc embedding use tokens
                    model_type_instance = cast(TextEmbeddingModel, embedding_model.model_type_instance)
                    tokens = model_type_instance.get_num_tokens(
                        model=embedding_model.model,
                        credentials=embedding_model.credentials,
                        texts=[content]
                    )
                segment.content = content
                segment.index_node_hash = segment_hash
                segment.word_count = len(content)
                segment.tokens = tokens
                segment.status = 'completed'
                segment.indexing_at = datetime.datetime.utcnow()
                segment.completed_at = datetime.datetime.utcnow()
                segment.updated_by = current_user.id
                segment.updated_at = datetime.datetime.utcnow()
                if document.doc_form == 'qa_model':
                    segment.answer = args['answer']
                db.session.add(segment)
                db.session.commit()
                # update segment vector index
                VectorService.update_segment_vector(args['keywords'], segment, dataset)
        except Exception as e:
            logging.exception("update segment index failed")
            segment.enabled = False
            segment.disabled_at = datetime.datetime.utcnow()
            segment.status = 'error'
            segment.error = str(e)
            db.session.commit()
        segment = db.session.query(DocumentSegment).filter(DocumentSegment.id == segment.id).first()
        return segment

    @classmethod
    def delete_segment(cls, segment: DocumentSegment, document: Document, dataset: Dataset):
        indexing_cache_key = 'segment_{}_delete_indexing'.format(segment.id)
        cache_result = redis_client.get(indexing_cache_key)
        if cache_result is not None:
            raise ValueError("Segment is deleting.")

        # enabled segment need to delete index
        if segment.enabled:
            # send delete segment index task
            redis_client.setex(indexing_cache_key, 600, 1)
            delete_segment_from_index_task.delay(segment.id, segment.index_node_id, dataset.id, document.id)
        db.session.delete(segment)
        db.session.commit()


class DatasetCollectionBindingService:
    @classmethod
    def get_dataset_collection_binding(cls, provider_name: str, model_name: str,
                                       collection_type: str = 'dataset') -> DatasetCollectionBinding:
        dataset_collection_binding = db.session.query(DatasetCollectionBinding). \
            filter(DatasetCollectionBinding.provider_name == provider_name,
                   DatasetCollectionBinding.model_name == model_name,
                   DatasetCollectionBinding.type == collection_type). \
            order_by(DatasetCollectionBinding.created_at). \
            first()

        if not dataset_collection_binding:
            dataset_collection_binding = DatasetCollectionBinding(
                provider_name=provider_name,
                model_name=model_name,
                collection_name="Vector_index_" + str(uuid.uuid4()).replace("-", "_") + '_Node',
                type=collection_type
            )
            db.session.add(dataset_collection_binding)
            db.session.commit()
        return dataset_collection_binding

    @classmethod
    def get_dataset_collection_binding_by_id_and_type(cls, collection_binding_id: str,
                                                      collection_type: str = 'dataset') -> DatasetCollectionBinding:
        dataset_collection_binding = db.session.query(DatasetCollectionBinding). \
            filter(DatasetCollectionBinding.id == collection_binding_id,
                   DatasetCollectionBinding.type == collection_type). \
            order_by(DatasetCollectionBinding.created_at). \
            first()

        return dataset_collection_binding