Python API: import goldenmatch as gm -- 95 exports

[Ben Severn (benzsevern)]

Python API: All Features from a Single Import

All GoldenMatch features are now accessible from a single import:

import goldenmatch as gm

This exposes 95 public exports covering the full pipeline -- from zero-config deduplication to advanced probabilistic matching, LLM-assisted scoring, and privacy-preserving record linkage.

Quick Start: Zero-Config Dedupe

import goldenmatch as gm

results = gm.dedupe("customers.csv")
print(f"{results.stats.clusters_found} duplicate groups found")
results.save("deduped.csv")

Two-Source Matching

import goldenmatch as gm

results = gm.match("crm.csv", "erp.csv", on=["name", "email", "phone"])
for cluster in results.clusters.values():
    print(cluster["members"], f"confidence={cluster["confidence"]:.2f}")

Advanced: Custom Matchkeys + LLM Scoring

import goldenmatch as gm

config = gm.GoldenMatchConfig(
    matchkeys=[
        gm.MatchKeyConfig(fields=["name", "zip"], type="fuzzy", threshold=0.85),
        gm.MatchKeyConfig(fields=["email"], type="exact"),
    ],
    llm_scorer=gm.LLMScorerConfig(model="gpt-4o-mini", budget=gm.BudgetConfig(max_cost_usd=0.05)),
)
results = gm.run_pipeline(["data.csv"], config=config)

Privacy-Preserving Linkage (PPRL)

import goldenmatch as gm

config = gm.GoldenMatchConfig(
    matchkeys=[
        gm.MatchKeyConfig(
            fields=["name", "dob"],
            type="fuzzy",
            threshold=0.80,
            transforms=["bloom_filter"],
        )
    ]
)
results = gm.run_pipeline([("hospital_a.csv", "A"), ("hospital_b.csv", "B")], config=config)

Key Modules Available

Area	Exports
Pipeline	run_pipeline, dedupe, match, match_one
Config	GoldenMatchConfig, MatchKeyConfig, LLMScorerConfig, BudgetConfig
Scoring	find_fuzzy_matches, find_exact_matches, score_blocks_parallel
Clustering	build_clusters, unmerge_record, unmerge_cluster
Probabilistic	train_em, score_probabilistic
Blocking	generate_blocking_key, LearnedBlocker
Streaming	StreamProcessor, match_one, add_to_cluster
Explainability	explain_pair_nl, explain_cluster_nl, build_lineage
PPRL	bloom_filter transform, dice/jaccard scorers
LLM	llm_score_pairs, BudgetTracker

Full list: print(gm.__all__)

Feedback and questions welcome!