Production profiles

Profiles provide safe defaults without hiding configuration. The first profile is production_safe.

production_safe

profile: production_safe

Runtime defaults:

Area	Default
Schema contract	enforcement: strict
Type inference	enabled: true, conflict_policy: fail
Schema evolution	enabled: true
Physical design	enabled: true, apply: online
Runtime evidence	enabled: true, output_dir: .dpone/evidence/latest
Lineage	enabled: true, preset: standard

User-specified values win over profile defaults:

profile: production_safe

sink:
  options:
    type_inference:
      sample_rows: 50000

sample_rows remains 50000; profile defaults only fill missing values.

Runtime integration

LoadConfigRuntimeService.prepare() applies profile defaults before ETLProcessor runs lifecycle gates. This means CLI, Python API, and orchestrated runs all see the same profile behavior.

Runbook

Need	Action
Certified production pipeline	Start with profile: production_safe and add explicit column contracts.
Dirty API source	Override schema_contract.enforcement: quarantine and monitor quarantine evidence.
Planned DDL window	Override physical_design.apply: safe_window only with approval evidence.
Experimental connector	Keep profile on, then relax one policy at a time with documented risk.

Related docs:

• Runtime lifecycle
• Runtime data contracts
• Physical DDL apply
• Certification suite