Type Mismatch Error for Start_TS in Foundry Builds

Hello Palantir Community,

I am encountering a build failure in Foundry related to the Start_TS property. Below is a summary of the issue and the troubleshooting steps I have taken.

Summary of Error:
The build failed due to a type mismatch in the property Start_TS. The system expected an offsetdatetimeudt type, but the field is being treated as a struct type during transformation/output generation. Because of this mismatch, the ontology build fails.

What I Observed:
When I import the dataset, Start_TS has values like 2026-03-26T02:45:52.604Z and Foundry identifies the type as datetime. However, after I transform the dataset or generate the output dataset, Foundry changes the datatype of this field to struct. Later, when I try to build the ontology type and ontology objects, Foundry expects this field to be offsetdatetimeudt, which causes the build to fail.

Why the Failure May Be Happening:
It looks like the source dataset initially has the column in a datetime-compatible format, but during transformation or output dataset generation, the schema for Start_TS is being converted into a struct instead of remaining as a timestamp/datetime field. As a result, the ontology expects offsetdatetimeudt, but the transformed dataset provides struct, creating a schema mismatch between the dataset and the ontology definition.

Troubleshooting Steps Taken:

I checked the source dataset and confirmed that Start_TS is originally shown as datetime.
I noticed that after transformation, the field changes to struct.
I attempted to correct the mismatch by transforming Start_TS into the expected type.
I used code to cast/convert the column, but the ontology build still fails due to the datatype mismatch.

Question:
How can I make sure that Start_TS stays as offsetdatetimeudt (or a compatible datetime/timestamp type) in the transformed/output dataset instead of being converted to struct? Has anyone faced a similar issue while building ontology types and ontology objects in Foundry?

Hey @TransformEngineerAM, thanks for reaching out!

I would try and make sure you are using the cast expression at some point in your integration pipeline either in pipeline builder or Python transforms to ensure the column time is Timestamp before it hits the ontology.

If this is not working for you, please attach relevant screenshots or code snippets and we can help!