Textract-Caller

amazon-textract-caller 提供了一系列可用的函数和示例实现，以加快任何使用 Amazon Textract 的项目的评估和开发。

无论文件类型和位置如何，都易于调用 Amazon Textract。

安装

> python -m pip install amazon-textract-caller

函数

from textractcaller import call_textract
def call_textract(input_document: Union[str, bytes],
                  features: Optional[List[Textract_Features]] = None,
                  queries_config: Optional[QueriesConfig] = None,
                  output_config: Optional[OutputConfig] = None,
                  adapters_config: Optional[AdaptersConfig] = None,
                  kms_key_id: str = "",
                  job_tag: str = "",
                  notification_channel: Optional[NotificationChannel] = None,
                  client_request_token: str = "",
                  return_job_id: bool = False,
                  force_async_api: bool = False,
                  call_mode: Textract_Call_Mode = Textract_Call_Mode.DEFAULT,
                  boto3_textract_client=None,
                  job_done_polling_interval=1) -> dict:

当接收到异步作业（start_document_text_detection 或 start_document_analysis）的 JSON 响应时也很有用。

from textractcaller import get_full_json
def get_full_json(job_id: str = None,
                  textract_api: Textract_API = Textract_API.DETECT,
                  boto3_textract_client=None)->dict:

当从 OutputConfig 位置接收到 JSON 时，此方法也很有用。

from textractcaller import get_full_json_from_output_config
def get_full_json_from_output_config(output_config: OutputConfig = None,
                                     job_id: str = None,
                                     s3_client = None)->dict:

示例

仅使用 detect_text 功能从本地文件系统中调用文件

textract_json = call_textract(input_document="/folder/local-filesystem-file.png")

仅使用 detect_text 从本地文件系统中调用文件并使用在 Textract 响应解析器中

（需要 trp 依赖项，通过 python -m pip install amazon-textract-response-parser 安装）

import json
from trp import Document
from textractcaller import call_textract

textract_json = call_textract(input_document="/folder/local-filesystem-file.png")
d = Document(textract_json)

使用查询调用多页文档并提取答案

该示例也使用了 amazon-textract-response-parser

python -m pip install amazon-textract-caller amazon-textract-response-parser

import textractcaller as tc
import trp.trp2 as t2
import boto3

textract = boto3.client('textract', region_name="us-east-2")
q1 = tc.Query(text="What is the employee SSN?", alias="SSN", pages=["1"])
q2 = tc.Query(text="What is YTD gross pay?", alias="GROSS_PAY", pages=["2"])
textract_json = tc.call_textract(
    input_document="s3://amazon-textract-public-content/blogs/2-pager.pdf",
    queries_config=tc.QueriesConfig(queries=[q1, q2]),
    features=[tc.Textract_Features.QUERIES],
    force_async_api=True,
    boto3_textract_client=textract)
t_doc: t2.TDocument = t2.TDocumentSchema().load(textract_json)  # type: ignore
for page in t_doc.pages:
    query_answers = t_doc.get_query_answers(page=page)
    for x in query_answers:
        print(f"{x[1]},{x[2]}")

使用适配器调用多页文档的定制查询

该示例也使用了 amazon-textract-response-parser

python -m pip install amazon-textract-caller amazon-textract-response-parser

import textractcaller as tc
import trp.trp2 as t2
import boto3

textract = boto3.client('textract', region_name="us-east-2")
q1 = tc.Query(text="What is the employee SSN?", alias="SSN", pages=["1"])
q2 = tc.Query(text="What is YTD gross pay?", alias="GROSS_PAY", pages=["2"])
adapter1 = tc.Adapter(adapter_id="2e9bf1c4aa31", version="1", pages=["1"])
textract_json = tc.call_textract(
    input_document="s3://amazon-textract-public-content/blogs/2-pager.pdf",
    queries_config=tc.QueriesConfig(queries=[q1, q2]),
    adapters_config=tc.AdaptersConfig(adapters=[adapter1])
    features=[tc.Textract_Features.QUERIES],
    force_async_api=True,
    boto3_textract_client=textract)
t_doc: t2.TDocument = t2.TDocumentSchema().load(textract_json)  # type: ignore
for page in t_doc.pages:
    query_answers = t_doc.get_query_answers(page=page)
    for x in query_answers:
        print(f"{x[1]},{x[2]}")

使用具有 TABLES 功能的本地文件系统中的文件调用

from textractcaller import call_textract, Textract_Features
features = [Textract_Features.TABLES]
response = call_textract(
    input_document="/folder/local-filesystem-file.png", features=features)

使用位于 S3 上的图像但强制使用异步 API 调用

from textractcaller import call_textract
response = call_textract(input_document="s3://some-bucket/w2-example.png", force_async_api=True)

使用 OutputConfig 和 Customer-Managed-Key 调用

from textractcaller import call_textract
from textractcaller import OutputConfig, Textract_Features
output_config = OutputConfig(s3_bucket="somebucket-encrypted", s3_prefix="output/")
response = call_textract(input_document="s3://someprefix/somefile.png",
                          force_async_api=True,
                          output_config=output_config,
                          kms_key_id="arn:aws:kms:us-east-1:12345678901:key/some-key-id-ref-erence",
                          return_job_id=False,
                          job_tag="sometag",
                          client_request_token="sometoken")

使用位于 S3 上的 PDF 并强制返回 JobId 而不是 JSON 响应

from textractcaller import call_textract
response = call_textract(input_document="s3://some-bucket/some-document.pdf", return_job_id=True)
job_id = response['JobId']