Schema mismatch not detected locally when serializing Number/String attributes

[Abhi904485]

When using PynamoDB with DynamoDB Local (amazon/dynamodb-local:latest), schema validation errors are not surfaced.
This leads to code that works against local DynamoDB but fails in production AWS DynamoDB with SerializationException.

Actual behavior

In production AWS DynamoDB:
{
  "Error": {
    "Code": "SerializationException",
    "Message": "NUMBER_VALUE cannot be converted to String"
  }
}

But when running locally with:

dynamodb-local:
  image: "amazon/dynamodb-local:latest"
  command: "-jar DynamoDBLocal.jar -sharedDb -dbPath ./data"
  ports:
    - "8030:8000"

the same write succeeds silently, even when attribute types are mismatched.

Example Code:

from pynamodb.models import Model
from pynamodb.attributes import UnicodeAttribute, NumberAttribute

class MyModel(Model):
    class Meta:
        table_name = "TestTable"
        host = "http://localhost:8030"

    id = UnicodeAttribute(hash_key=True)
    count = NumberAttribute()

# Works locally but fails on real DynamoDB
item = MyModel(id="123", count="not_a_number")
item.save()

Local DynamoDB → succeeds, record written with wrong type
AWS DynamoDB → raises SerializationException: NUMBER_VALUE cannot be converted to String

Proposal

Could PynamoDB add stricter client-side validation to catch these mismatches before sending the request?

Options:

• Enforce type checking in Attribute.serialize (e.g., NumberAttribute should only accept int/float, not str).
• Raise a ValueError or PynamoDB-specific ValidationError when serialization is impossible.
• Provide a setting like STRICT_VALIDATION=True in Meta to toggle this behavior, so existing code isn’t broken.

This would make local DynamoDB testing more consistent with real AWS behavior.

Environment

• Python: 3.10 (prod), 3.11 (local)
• PynamoDB version: 6.0.2
• boto3: 1.28.26
• botocore: ==1.31.26
• DynamoDB Local: amazon/dynamodb-local:latest
• AWS DynamoDB: ap-south-1