Truthound for Kestra¶

Truthound's Kestra integration is YAML-first. It gives Kestra users Python helpers for script tasks, generated flow templates, structured outputs, and SLA-aware execution without asking them to abandon native Kestra flow design.

Who This Is For¶

teams orchestrating data quality with Kestra flow YAML
operators who prefer script-task execution over heavy embedded framework code
platform engineers standardizing reusable flow templates

When To Use It¶

Choose Kestra when:

your team wants flow YAML to remain the primary control plane
Python script tasks are already a normal part of the platform
generated flow templates are preferable to hand-written repetitive validation jobs

Prerequisites¶

truthound-orchestration[kestra] installed in the script execution environment
a Kestra deployment with namespaces and task runners
a supported URI, file, or relation-oriented validation source

What The Package Provides¶

script entry points such as check_quality_script
reusable executor classes for check, stream, profile, and learn workflows
flow generation helpers such as generate_check_flow
output handlers for Kestra-friendly result emission
SLA monitoring hooks and presets

Minimal Quickstart¶

from truthound_kestra import check_quality_script

result = check_quality_script(
    data_uri="s3://warehouse/curated/users.parquet",
    rules=[
        {"column": "id", "check": "not_null"},
        {"column": "email", "check": "email_format"},
    ],
)

Generate a reusable flow template when the pattern should be standardized:

from truthound_kestra import generate_check_flow, RetryConfig

yaml_content = generate_check_flow(
    flow_id="users_quality",
    namespace="production",
    retry=RetryConfig(max_attempts=3),
)

Decision Table¶

Need	Recommended Kestra Surface	Why
execute a one-off quality script	`check_quality_script`	easiest task-level entry point
standardize repeated flow structure	`generate_check_flow` or related template helpers	keeps YAML generation consistent
publish structured outputs	`KestraOutputHandler` or `send_check_result`	aligns with Kestra output conventions
separate environments or teams	namespace-aware flow config	keeps ownership explicit

Execution Lifecycle¶

flowchart LR
    A["Kestra task or generated flow starts"] --> B["Script config and namespace resolved"]
    B --> C["Shared runtime resolves the source and preflight"]
    C --> D["Engine executes check/profile/learn/stream"]
    D --> E["Shared result serialized"]
    E --> F["Kestra outputs, metrics, retries, and notifications consume the result"]

Result Surface¶

script helpers return shared Truthound results
Kestra output handlers translate them into task outputs and metrics
downstream flows should consume the structured output contract instead of parsing logs

Config Surface¶

Config Area	Kestra Boundary
validation input	URI, file, or task output passed into the script
runtime behavior	script config and retry config
environment ownership	namespace and Kestra variables/secrets
output contract	output handlers and task outputs