Changelog

Generated by releaseledger. Manual edits may be overwritten. Update releaseledger entries and rebuild this file instead.

[v0.1.9] - Unreleased

Added

  • Added direct truth-value checks for case-alias mappings without enumerating aliases

  • Added lazy, position-aware prefix lookup to lexicons

[v0.1.8] - Unreleased

Added

  • Added an IPA TSV adapter with slash-delimited pronunciation parsing and CLI support

[v0.1.7] - 2026-08-30

Added

  • Added deterministic lexicon and cascade benchmarks with JSON/TSV reports for variants, Unicode, collisions, and coverage

Fixed

  • Fixed packed asset metadata to ignore checkout paths while retaining original parsed source paths

[v0.1.6] - 2026-08-30

Added

  • Added layer-prioritized lookup for ordered candidate spellings with provenance-preserving hit results

[v0.1.5] - 2026-08-29

Added

  • Added ordered pronunciation lookup helpers while preserving exact typed values

Changed

  • Changed CLI handlers and expanded the documentation site for stable and experimental workflows

Fixed

  • Fixed binary asset validation and resource cleanup for malformed or failed loads

[v0.1.4] - 2026-08-29

Changed

  • Changed layered lexicons to expose lookup provenance and explicit ownership and closure

Fixed

  • Fixed collisions between user metadata and reserved binary manifest fields by rejecting them

Deprecated

  • Changed the Kokoro compatibility helper to stop retaining live lexicons

Documentation

  • Added PyPI package version, Python-version, and download badges to the README

Quality

  • Improved runtime type safety and deterministic selector behavior across adapters and serialized assets

[0.1.3] - 2026-08-28

Quality

  • Improved runtime type safety and concrete value handling across adapters, assets, and lexicon operations

[0.1.2] - 2026-08-28

Quality

  • Improved package publishing compatibility across release versions

[0.1.1] - 2026-08-28

Documentation

  • Added PyPI version, supported Python version, and download-count badges to the README

[v0.1.0] - 2026-08-28

Added

  • Added the Python package and CLI for lossless lexicon reduction with German profiles and exact reconstruction checks

  • Added indexed V4 assets with memory-mapped loading, bounded reconstruction, configurable selectors, and training support

  • Added typed lexicon storage with deterministic assets, lazy lookup, source adapters, layering, and exact verification

Changed

  • Changed the public package and CLI name to g2lex, updating asset identities, metadata, docs, scripts, and tests

Quality

  • Added reproducible German de-gold benchmarks with locked sources, ablation runs, and result aggregation

  • Added Crane Wiktionary source configuration and benchmark verification coverage