v0.6.1 -- PPRL Auto-Config, Vectorized Similarity, Benchmarks

PPRL Auto-Configuration

GoldenMatch now automatically picks optimal PPRL parameters from your data -- zero manual tuning needed.

goldenmatch pprl auto-config data.csv

Profiles every column, scores usefulness for privacy-preserving linkage, recommends fields, bloom filter parameters, and threshold.

Results: auto-config beats manual tuning on both benchmark datasets:

Dataset	Auto-Config F1	Manual F1
FEBRL4 (synthetic, 5K vs 5K)	92.4%	89.8%
NCVR (real voter data, 5K+2.5K)	76.1%	65.8%

Performance

Vectorized PPRL similarity computation: 13x speedup (183s -> 14s on 5Kx5K).

MCP Tools

Two new tools for Claude Desktop:

• pprl_auto_config -- analyze data, recommend PPRL config
• pprl_link -- run cross-party linkage

Stats

• 903 tests passing
• CI green on Python 3.11/3.12/3.13

pip install --upgrade goldenmatch