Amazon Textract boto3 response changed when using in lambda

Hi, thank you for using Textract. Textract did update the table model to support merged_cells and table_headers. https://aws.amazon.com/about-aws/whats-new/2022/03/amazon-textract-updates-tables-check-detection/

The update adds a new BlockType called "MERGED_CELLS" and Relationships Type "MERGED_CELL" and an EntityType "COLUMN_HEADER". If you don't need those, you can ignore them.

Outside of those additions the response is the same as the "older" one with all CELLs of a TABLE being the CHILD Relationship. See: https://docs.aws.amazon.com/textract/latest/dg/how-it-works-tables.html

I recommend using https://pypi.org/project/amazon-textract-response-parser/ for parsing the response in Python.

There is no difference in response format when calling Textract from Lambda vs. locally or from an EC2 instance. BUT Lambda does not always include the latest SDK version.

Check https://docs.aws.amazon.com/lambda/latest/dg/lambda-runtimes.html for more information. You may have to install the same boto3 version (here is an example: https://aws.amazon.com/premiumsupport/knowledge-center/lambda-python-runtime-errors/).

I usually use containers with Lambda now (generated with SAM or CDK) where I have full control.