Crate zer_schema 

Schema registry and model persistence for zer.

This crate provides three cooperating components:

1. SchemaInferrer, automatic FieldKind detection from column names and value patterns; produces a Schema without requiring the caller to know the dataset structure upfront.

2. SchemaFingerprint, a compact identity for a schema plus its data distribution (SHA-256 hash of field names/kinds, per-field null rates, cardinalities).

3. SchemaRegistry, a sled-backed persistent store for ModelArtifacts (trained Fellegi-Sunter parameters). On startup the pipeline calls SchemaRegistry::lookup_startup_mode to decide whether to load params directly (exact match), warm-start EM (similar schema), or run full EM from priors (new/incompatible schema).

Re-exports

pub use artifact::ModelArtifact;

pub use config::NameHeuristics;

pub use config::ValuePatterns;

pub use fingerprint::FieldStats;

pub use fingerprint::SchemaFingerprint;

pub use infer::SchemaInferrer;

pub use registry::SchemaRegistry;

pub use registry::StartupMode;

pub use similarity::fingerprint_distance;

pub use similarity::EXACT_MATCH_THRESHOLD;

pub use similarity::WARM_START_THRESHOLD;

Modules

artifact

config

fingerprint

infer

registry

similarity