TypeScript / JavaScript

GoldenCheck has a full TypeScript port with complete Python feature parity.

Installation 

npm install goldencheck

Quick Start 

import { scanData, TabularData } from "goldencheck";

const data = new TabularData([
  { id: 1, email: "alice@example.com", age: 30 },
  { id: 2, email: "not-an-email", age: -5 },
]);

const { findings } = scanData(data);
findings.forEach(f => console.log(`[${f.check}] ${f.column}: ${f.message}`));

Architecture

The TypeScript package is split into two layers:

Layer Import Runtimes
Core goldencheck/core Browsers, Cloudflare Workers, Vercel Edge Runtime, Deno, Bun
Node goldencheck/node Node.js (CSV/Parquet reading, CLI, MCP server, A2A server)

What’s included

  • 14 profilers — all 10 column + 4 relation profilers ported from Python
  • Semantic types — classifier, suppression engine, 3 domain packs (healthcare, finance, ecommerce)
  • Baseline/drift — statistical profiling + 13 drift checks
  • LLM integration — Anthropic + OpenAI via raw fetch (edge-safe, no SDK deps)
  • MCP server — 7 tools for Claude Desktop integration
  • CLI — npx goldencheck-js scan data.csv

CLI Usage 

npx goldencheck-js scan data.csv
npx goldencheck-js baseline data.csv
npx goldencheck-js scan data.csv --baseline goldencheck_baseline.yaml
npx goldencheck-js health-score data.csv

Edge Runtime

The core module works in any JavaScript runtime without Node.js-specific APIs:

import { scanData, TabularData } from "goldencheck/core";

export default {
  async fetch(request: Request) {
    const data = new TabularData(await request.json());
    const { findings } = scanData(data);
    return Response.json({ findings });
  },
};

Domain Packs 

import { scanData, TabularData } from "goldencheck";

const { findings } = scanData(data, { domain: "healthcare" });

Available domains: healthcare, finance, ecommerce.

LLM Boost 

import { scanDataWithLLM, TabularData } from "goldencheck";

const { findings } = await scanDataWithLLM(data, {
  provider: "anthropic",
  apiKey: process.env.ANTHROPIC_API_KEY,
});

API Reference

scanData(data, options?)

Scan tabular data for quality issues.

Parameter Type Description
data TabularData The dataset to scan
options.domain string Domain pack: "healthcare", "finance", "ecommerce"
options.baseline string \| Baseline Path to baseline YAML or Baseline object

Returns { findings: Finding[], profile: Profile }.

scanDataWithLLM(data, options)

Scan with LLM boost for additional rule discovery.

Parameter Type Description
data TabularData The dataset to scan
options.provider string "anthropic" or "openai"
options.apiKey string API key for the provider

Returns Promise<{ findings: Finding[], profile: Profile }>.

createBaseline(data)

Create a statistical baseline from a dataset.

import { createBaseline } from "goldencheck";

const baseline = createBaseline(data);
baseline.save("goldencheck_baseline.yaml");

healthScore(data)

Get a letter grade and numeric score.

import { healthScore } from "goldencheck";

const score = healthScore(data);
console.log(score); // e.g. "B (78/100)"

TabularData

Wrapper for tabular data, accepts arrays of objects or column-oriented data.

// Row-oriented
const data = new TabularData([
  { name: "Alice", age: 30 },
  { name: "Bob", age: 25 },
]);

// Column-oriented
const data = new TabularData({
  name: ["Alice", "Bob"],
  age: [30, 25],
});

Finding

Each finding has these fields:

Field Type Description
column string Column name
check string Check identifier
message string Human-readable description
severity "ERROR" \| "WARNING" \| "INFO" Issue severity
confidence number 0.0 to 1.0
source string \| null null = profiler, "llm" = LLM-generated