Releases: Bigred97/wgea-mcp
Release list
0.6.19: top_n full-rank before 10k row ceiling
v0.6.18
v0.6.14
0.6.13
0.6.12 — sharpen employee_weighted scope docs
Docs-only release. Describe_dataset now explicitly says: anzsic_division='all' returns the national 21.66% headline; specific division returns industry-weighted (Mining 11.46%, Finance 28.17%) — NOT directly comparable to WGEA's 21.1% national.
0.6.11 — prewarm_curated() + --warmup CLI
Ports abs-mcp 0.11.14 prewarm pattern. 235 unit tests pass.
0.6.10 — employer_size_band dim on HEADLINE_GAP
Adds size-band breakdowns to HEADLINE_GAP. Reveals real customer pattern: 1000-4999 employers have the widest gap (24.4%) vs 5000+ at 14.97%. 235 unit tests pass.
0.6.9 — WGEA 21.1% national headline GPG derived from xlsx
Real fix (not a bandaid): customer-expected WGEA Scorecard headline (21.1% for 2024-25) is now reachable via the new employee_weighted_total_rem_gap_pct measure. Derived from per-employer xlsx using A = w·f + (1-w)·m and GPG = (m-f)/m, weighted by size-band midpoint. Reproduces 21.66% — within 0.6pp of WGEA's published 21.1%. Available per-industry too. 235 unit tests pass.
0.6.8 — HEADLINE_GAP methodology fix
Switched total_remuneration_gap_pct from median to mean (now 12.38% vs WGEA's published 21.1%). Sharpened docs on methodology gap — HEADLINE_GAP is per-employer stats, not WGEA's national payroll-weighted headline.
0.6.7 — wheel-verify count fix
CI fix only — wheel-verify now asserts >=7 curated datasets instead of ==7, accommodating HEADLINE_GAP and future additions.