Prefect Artifacts and Result Payloads¶

Prefect users often want a result that is both programmatic and visible in the Prefect UI. Truthound supports that by keeping a shared result contract and offering artifact-oriented helpers such as to_prefect_artifact.

Who This Is For¶

flow authors publishing quality summaries to Prefect
operators debugging why a flow succeeded with warnings
teams standardizing reporting across tasks and flows

When To Use It¶

Use this page when:

a quality task result should appear as a Prefect artifact
a downstream step needs structured status and failure counts
flow-level reporting should look consistent across environments

Prerequisites¶

truthound-orchestration[prefect] installed
a Prefect flow using tasks or a DataQualityBlock
access to the serialized Truthound result

Minimal Quickstart¶

Convert a result into a Prefect-friendly artifact payload:

from truthound_prefect import data_quality_check_task, to_prefect_artifact

result = await data_quality_check_task(
    data,
    rules=[{"column": "id", "check": "not_null"}],
)
artifact = to_prefect_artifact(result)

Block-based execution can follow the same output path:

from truthound_prefect import DataQualityBlock

block = DataQualityBlock(engine_name="truthound")
result = await block.check(data, rules=[{"column": "email", "check": "email_format"}])

Production Pattern¶

Use a layered output model:

Layer	Purpose
shared Truthound result	canonical status, counts, and failure detail
Prefect artifact payload	UI-friendly presentation and operator context

Recommended practice:

keep the shared result as the machine-readable source of truth
use artifacts for dashboards and operator visibility
standardize naming for artifacts so repeated flows are comparable

Failure Modes and Troubleshooting¶

Symptom	Likely Cause	What To Do
artifact exists but is hard to automate	only display text was persisted	preserve the structured result dictionary too
a warning run looks successful but operators miss it	only Prefect state is inspected	include warning counts in the artifact summary
outputs differ between flows	each flow formats artifacts differently	route through `to_prefect_artifact` or a shared wrapper