DuckDBPolarsIOManager with partitioned assets

[chrwnz-eng]

Dear community,

I'm struggling with I/O Manager (DuckDBPolarsIOManager) and partitioned assets. I defined the following asset:

@asset(description="Extracts raw change requests from jira server",
       io_manager_key="io_manager",
       partitions_def=DailyPartitionsDefinition(start_date="2025-12-05"),
       metadata={"partition_expr": "extract_date"})
def extract_change_requests(context: AssetExecutionContext) -> polars.DataFrame:
    jira_base_url_str: str = "http://..."

    extract_date = context.partition_key
    jr = JiraRequests("CRTR", jira_base_url_str)
    js = jr.get_jira_stories()

    org_crs: dict = {"org_change_requests": js.text, "extract_date": extract_date}

    return polars.DataFrame(org_crs)

The materialization and the persistence in the DuckDB works fine if I run this asset. With the help of another asset I try to load these data from the first asset and transform them.

@asset(description="Transforms raw change requests",
       io_manager_key="io_manager",
       partitions_def=DailyPartitionsDefinition(start_date="2025-12-05"),
       metadata={"partition_expr": "extract_date"})
def transform_change_requests(context: AssetExecutionContext, extract_change_requests: polars.DataFrame) -> polars.DataFrame:

    extract_date = context.partition_key

    js_list = <do some transformation>

    trans_crs: dict = {js_list, "extract_date": extract_date, "project": "CRTR"}

    return polars.DataFrame(trans_crs)

Unfortunately is the DataFrame extract_change_requests in the asset transform_change_requests empty. If I remove the partition_def and the metadata in the annotation of transform_change_requests the mentioned DataFrame is populated with data. Please, could somebody explain this to me and give me a hint how to use partions also in the asset transform_change_requests.

BR