Skip to content

Upstream Source

This page is part of Truthound Orchestration 3.x.

Source repository: seadonggyun4/truthound-orchestration Upstream docs path: docs/mage/production-rollout.md Edit upstream page: Edit in orchestration

Mage Production Rollout

Mage is often adopted first in a single pipeline and only later standardized across projects. This page documents the rollout path that keeps Truthound quality blocks maintainable as the surface grows.

Who This Is For

  • teams promoting Mage quality checks from pilot to production
  • platform engineers defining a shared rollout pattern
  • operators creating incident and rollback procedures for Mage pipelines

When To Use It

Use this page when:

  • more than one Mage pipeline needs the same quality pattern
  • SLA-aware rollout and alerting are becoming operational requirements
  • a local io_config.yaml path needs to become environment-safe

Prerequisites

  • at least one working Truthound Mage pipeline
  • project configuration under version control
  • an environment strategy for Mage execution

Minimal Quickstart

Start with a dedicated validation transformer block:

from truthound_mage import CheckTransformer, CheckBlockConfig

check_block = CheckTransformer(
    config=CheckBlockConfig(
        rules=[{"column": "id", "check": "not_null"}],
        fail_on_error=True,
    )
)

Then add a sensor or condition for the rollout boundary:

from truthound_mage import DataQualityCondition

Production Pattern

Recommended rollout ladder:

  1. validate in one transformer block
  2. externalize source config through io_config.yaml and profiles
  3. add sensors or conditions for branching and gating
  4. enable SLA hooks and notification routing
  5. standardize shared block config across pipelines

Failure Modes and Troubleshooting

Symptom Likely Cause What To Do
every pipeline rolls its own quality behavior no shared rollout template exists publish one blessed transformer + gating pattern
incident response is slow blocks emit results but no operational hooks are wired add SLA monitoring before wider rollout
local assumptions leak into production io_config.yaml and env policy are not formalized promote profile and secret conventions early