Skip to main content
goldencheck-types is the suite’s shared canonical field-type registry — a single source of truth for “what does the field type email mean?”: its name hints, its value signals, and its confidence thresholds. It is deliberately tiny and dependency-light, because its whole value is being depended on by everything without dragging anything along.

​
Producer and consumers

GoldenCheck profiles a dataset and emits inferred field types. Downstream packages consume them:
  • Producer — goldencheck (data-quality profiling).
  • Consumers — GoldenPipe (stage I/O contracts), InferMap (target-schema inference), and the TypeScript mirror goldencheck-types over a JSON wire.

​
Why it is its own package

Two reasons, both about keeping a vocabulary honest:
  1. Cross-language byte-identity. The Python and TypeScript sides produce the same types because they read the same registry, not because two implementations were written to agree and then drifted.
  2. Dependency-light consumption. A consumer that needs the vocabulary should not have to install the full profiler to get it.
This package is a vocabulary, not an engine — it has no config surface, no CLI, and no MCP server. Field-type detection lives in GoldenCheck; this package only defines what the resulting types mean.