How to avoid full table scan in Glue's create_dynamic_frame.from_options for dynamodb

My Dynamodb table has both PK and SK. it has huge data set(500 GB).

I'm using below syntax for querying data based on PK in Glue, But it does a full table scan leading to the glue timeout. Have checked the RCU, that is also very high.

dyf = glueContext.create_dynamic_frame.from_options(
    connection_type="dynamodb",
    connection_options={
        "dynamodb.region": region,
        "dynamodb.input.tableName": tablename,
        "dynamodb.query.filterExpression": f"prt_key = :pv",
        "dynamodb.query.expressionAttributeValues": '{":pv": {"S": "'+ id + '"}}'
    }
)
df= dyf.toDF()

Can someone please suggest a solution to avoid full table scan using create_dynamic_frame.from_options?

I already have boto3 based solution for this, looking for a solution with create_dynamic_frame.from_options.

AWS Glue will always do a full table scan. If you only want data for a given partition key, use boto3 query to get the data you need.