Copiar versões do processador e conjuntos de dados entre projetos

Esta página contém etapas para copiar versões de processadores treinados da Document AI de um projeto para outro, junto com o esquema do conjunto de dados e exemplos da origem para o processador de destino. Essas etapas automatizam o processo de importação da versão do processador, implantação e definição como padrão no projeto de destino.

Antes de começar

  • Receba um ID do projeto Google Cloud .
  • Tenha o ID do processador da Document AI.
  • Ter o Cloud Storage.
  • Usar Python: notebook Jupyter (Vertex AI).
  • Você precisa de permissões para dar acesso à conta de serviço nos projetos de origem e destino.

Procedimento detalhado

O procedimento é descrito nas etapas a seguir.

Etapa 1: identificar a conta de serviço associada ao Vertex AI Notebook

!gcloud config list account

Saída:

[core]
account = example@automl-project.iam.gserviceaccount.com

Your active configuration is: [default]

Etapa 2: conceder as permissões necessárias à conta de serviço

No projeto Google Cloud que é o destino pretendido da migração, adicione a conta de serviço adquirida na etapa anterior como um principal e atribua os dois papéis a seguir:

  • Administrador da Document AI
  • Administrador de armazenamento

Consulte Como conceder papéis a contas de serviço e Chaves de criptografia gerenciadas pelo cliente (CMEK) para mais informações.

processor-version-migrate-1

Para que a migração funcione, a conta de serviço usada para executar este notebook precisa ter:

  • Funções nos projetos de origem e destino para criar o bucket do conjunto de dados ou criá-lo se ele não existir, além de permissões de leitura e gravação para todos os objetos.
  • Função de editor da Document AI no projeto de origem, conforme descrito em Importar uma versão do processador.

Faça o download de uma chave JSON para a conta de serviço, assim você poderá autenticar e autorizar como conta de serviço. Para mais informações, consulte Chaves de conta de serviço.

Próximo:

  1. Acesse a conta de serviço.
  2. Selecione a conta de serviço destinada a realizar essa tarefa.
  3. Acesse a guia Chaves, clique em Add Key e escolha Criar nova chave.
  4. Selecione o tipo de chave (de preferência JSON).

  5. Clique em Create e faça o download para um caminho específico. processor-version-migrate-2

  6. Atualize o caminho na variável service_account_key no snippet a seguir.

service_account_key='path_to_sa_key.json'

from google.oauth2 import service_account
from google.cloud import storage

# Authenticate the service account
credentials = service_account.Credentials.from_service_account_file(
    service_account_key
)

# pass this credentials variable to all client initializations
# storage_client = storage.Client(credentials=credentials)
# docai_client = documentai.DocumentProcessorServiceClient(credentials=credentials)

Etapa 3: importar as bibliotecas

import time
from pathlib import Path
from typing import Optional, Tuple
from google.cloud.documentai_v1beta3.services.document_service import pagers
from google.api_core.client_options import ClientOptions
from google.api_core.operation import Operation
from google.cloud import documentai_v1beta3 as documentai
from google.cloud import storage
from tqdm import tqdm

Etapa 4: inserir detalhes

  • source_project_id: forneça o ID do projeto de origem.
  • source_location: informe o local do processador de origem (us ou eu).
  • source_processor_id: forneça o ID do processador da Document AI. Google Cloud
  • source_processor_version_to_import: forneça o ID da versão do processador da Google Cloud Document AI para a versão treinada.
  • migrate_dataset::forneça esse valor como True ou False. Se você quiser migrar o conjunto de dados do processador de origem para o de destino, forneça True. Caso contrário, forneça False. O valor padrão é False.
  • source_exported_gcs_path: informe o caminho do Cloud Storage para armazenar arquivos JSON.
  • destination_project_id: forneça o ID do projeto de destino.
  • destination_processor_id: forneça o ID do processador da Document AI, "" ou processor_id do projeto de destino. Google Cloud 
source_project_id = "source-project-id"
source_location = "processor-location"
source_processor_id = "source-processor-id"
source_processor_version_to_import = "source-processor-version-id"
migrate_dataset = False  # Either True or False
source_exported_gcs_path = (
    "gs://bucket/path/to/export_dataset/"
)
destination_project_id = "< destination-project-id >"
# Give an empty string if you wish to create a new processor
destination_processor_id = ""

Etapa 5: executar o código

import time
from pathlib import Path
from typing import Optional, Tuple
from google.cloud.documentai_v1beta3.services.document_service import pagers
from google.api_core.client_options import ClientOptions
from google.api_core.operation import Operation
from google.cloud import documentai_v1beta3 as documentai
from google.cloud import storage
from tqdm import tqdm

source_project_id = "source-project-id"
source_location = "processor-location"
source_processor_id = "source-processor-id"
source_processor_version_to_import = "source-processor-version-id"
migrate_dataset = False # Either True or False
source_exported_gcs_path = (
    "gs://bucket/path/to/export_dataset/"
)
destination_project_id = "< destination-project-id >"
# Give empty string if you wish to create a new processor
destination_processor_id = ""

exported_bucket_name = source_exported_gcs_path.split("/")[2]
exported_bucket_path_prefix = "/".join(source_exported_gcs_path.split("/")[3:])
destination_location = source_location

def sample_get_processor(project_id: str, processor_id: str, location: str)->Tuple[str, str]:
    """
    This function returns Processor Display Name and Type of Processor from source project

    Args:
        project_id (str): Project ID
        processor_id (str): Document AI Processor ID
        location (str): Processor Location

    Returns:
        Tuple[str, str]: Returns Processor Display name and type
    """
    client = documentai.DocumentProcessorServiceClient()
    print(
        f"Fetching processor({processor_id}) details from source project ({project_id})"
    )
    name = f"projects/{project_id}/locations/{location}/processors/{processor_id}"
    request = documentai.GetProcessorRequest(
        name=name,
    )
    response = client.get_processor(request=request)
    print(f"Processor Name: {response.name}")
    print(f"Processor Display Name: {response.display_name}")
    print(f"Processor Type: {response.type_}")
    return response.display_name, response.type_

def sample_create_processor(project_id: str, location: str, display_name: str, processor_type: str)->documentai.Processor:
    """It will create Processor in Destination project

    Args:
        project_id (str): Project ID
        location (str): Location fo processor
        display_name (str): Processor Display Name
        processor_type (str): Google Cloud Document AI Processor type

    Returns:
        documentai.Processor: Returns details abouts newly created processor
    """

    client = documentai.DocumentProcessorServiceClient()
    request = documentai.CreateProcessorRequest(
        parent=f"projects/{project_id}/locations/{location}",
        processor={
            "type_": processor_type,
            "display_name": display_name,
        },
    )
    print(f"Creating Processor in project: {project_id} in location: {location}")
    print(f"Display Name: {display_name} & Processor Type: {processor_type}")
    res = client.create_processor(request=request)
    return res

def initialize_dataset(project_id: str, processor_id: str, location: str)-> Operation:
    """It will configure dataset for target processor in destination project

    Args:
        project_id (str): Project ID
        processor_id (str): DocuemntAI Processor ID
        location (str): Processor Location

    Returns:
        Operation: An object representing a long-running operation
    """

    # opts = ClientOptions(api_endpoint=f"{location}-documentai.googleapis.com")
    client = documentai.DocumentServiceClient()  # client_options=opts
    dataset = documentai.types.Dataset(
        name=f"projects/{project_id}/locations/{location}/processors/{processor_id}/dataset",
        state=3,
        unmanaged_dataset_config={},
        spanner_indexing_config={},
    )
    request = documentai.types.UpdateDatasetRequest(dataset=dataset)
    print(
        f"Configuring Dataset in  project: {project_id} for processor: {processor_id}"
    )
    response = client.update_dataset(request=request)
    return response

def get_dataset_schema(project_id: str, processor_id: str, location: str)->documentai.DatasetSchema:
    """It helps to fetch processor schema

    Args:
        project_id (str): Project ID
        processor_id (str): DocumentAI Processor ID
        location (str): Processor Location

    Returns:
        documentai.DatasetSchema: Return deails about Processor Dataset Schema
    """

    # Create a client
    processor_name = (
        f"projects/{project_id}/locations/{location}/processors/{processor_id}"
    )
    client = documentai.DocumentServiceClient()
    request = documentai.GetDatasetSchemaRequest(
        name=processor_name + "/dataset/datasetSchema"
    )
    # Make the request
    print(f"Fetching schema from source processor: {processor_id}")
    response = client.get_dataset_schema(request=request)
    return response

def upload_dataset_schema(schema: documentai.DatasetSchema)->documentai.DatasetSchema:
    """It helps to update the schema in destination processor

    Args:
        schema (documentai.DatasetSchema): Document AI Processor Schema details & Metadata

    Returns:
        documentai.DatasetSchema: Returns Dataset Schema object
    """

    client = documentai.DocumentServiceClient()
    request = documentai.UpdateDatasetSchemaRequest(dataset_schema=schema)
    print("Updating Schema in destination processor")
    res = client.update_dataset_schema(request=request)
    return res

def store_document_as_json(document: str, bucket_name: str, file_name: str)->None:
    """It helps to upload data to Cloud Storage and stores as a blob

    Args:
        document (str): Processor response in json string format
        bucket_name (str): Cloud Storage bucket name
        file_name (str): Cloud Storage blob uri
    """

    print(f"\tUploading file to Cloud Storage gs://{bucket_name}/{file_name}")
    storage_client = storage.Client()
    process_result_bucket = storage_client.get_bucket(bucket_name)
    document_blob = storage.Blob(
        name=str(Path(file_name)), bucket=process_result_bucket
    )
    document_blob.upload_from_string(document, content_type="application/json")

def list_documents(project_id: str, location: str, processor: str, page_size: Optional[int]=100, page_token: Optional[str]="")->pagers.ListDocumentsPager:
    """This function helps to list the samples present in processor dataset

    Args:
        project_id (str): Project ID
        location (str): Processor Location 
        processor (str): DocumentAI Processor ID
        page_size (Optional[int], optional): The maximum number of documents to return. Defaults to 100.
        page_token (Optional[str], optional): A page token, received from a previous ListDocuments call. Defaults to "".

    Returns:
        pagers.ListDocumentsPager: Returns all details about documents present in Processor Dataset
    """
    client = documentai.DocumentServiceClient()
    dataset = (
        f"projects/{project_id}/locations/{location}/processors/{processor}/dataset"
    )
    request = documentai.types.ListDocumentsRequest(
        dataset=dataset,
        page_token=page_token,
        page_size=page_size,
        return_total_size=True,
    )
    print(f"Listingll  documents/Samples present in processor: {processor}")
    operation = client.list_documents(request)
    return operation

def get_document(project_id: str, location: str, processor: str, doc_id: documentai.DocumentId)->documentai.GetDocumentResponse:
    """It will fetch data for individual sample/document present in dataset

    Args:
        project_id (str): Project ID
        location (str): Processor Location
        processor (str): Document AI Processor ID
        doc_id (documentai.DocumentId): Document identifier

    Returns:
        documentai.GetDocumentResponse: Returns data related to doc_id
    """

    client = documentai.DocumentServiceClient()
    dataset = (
        f"projects/{project_id}/locations/{location}/processors/{processor}/dataset"
    )
    request = documentai.GetDocumentRequest(dataset=dataset, document_id=doc_id)
    operation = client.get_document(request)
    return operation

def import_documents(project_id: str, processor_id: str, location: str, gcs_path: str)->Operation:
    """It helps to import samples/docuemnts from Cloud Storage path to processor via API call

    Args:
        project_id (str): Project ID
        processor_id (str): Document AI Processor ID
        location (str): Processor Location
        gcs_path (str): Cloud Storage path uri prefix 

    Returns:
        Operation: An object representing a long-running operation
    """

    client = documentai.DocumentServiceClient()
    dataset = (
        f"projects/{project_id}/locations/{location}/processors/{processor_id}/dataset"
    )
    request = documentai.ImportDocumentsRequest(
        dataset=dataset,
        batch_documents_import_configs=[
            {
                "dataset_split": "DATASET_SPLIT_TRAIN",
                "batch_input_config": {
                    "gcs_prefix": {"gcs_uri_prefix": gcs_path + "train/"}
                },
            },
            {
                "dataset_split": "DATASET_SPLIT_TEST",
                "batch_input_config": {
                    "gcs_prefix": {"gcs_uri_prefix": gcs_path + "test/"}
                },
            },
            {
                "dataset_split": "DATASET_SPLIT_UNASSIGNED",
                "batch_input_config": {
                    "gcs_prefix": {"gcs_uri_prefix": gcs_path + "unassigned/"}
                },
            },
        ],
    )
    print(
        f"Importing Documents/samples from {gcs_path} to corresponding tran_test_unassigned sections"
    )
    response = client.import_documents(request=request)
    return response

def import_processor_version(source_processor_version_name: str, destination_processor_name: str)->Operation:
    """It helps to import processor version from source processor to destanation processor

    Args:
        source_processor_version_name (str): source processor name in this format projects/{project}/locations/{location}/processors/{processor}
        destination_processor_name (str): destination processor name in this format projects/{project}/locations/{location}/processors/{processor}

    Returns:
        Operation: An object representing a long-running operation
    """

    from google.cloud import documentai_v1beta3

    # provide the source version(to copy) processor details in the following format
    client = documentai_v1beta3.DocumentProcessorServiceClient()
    # provide the new processor name in the parent variable in format 'projects/{project_number}/locations/{location}/processors/{new_processor_id}'
    import google.cloud.documentai_v1beta3 as documentai

    op_import_version_req = (
        documentai.types.document_processor_service.