v1.2.0 — TypeScript Port

What's New

GoldenFlow is now available in TypeScript with full feature parity to the Python library.

TypeScript Port (packages/goldenflow-js)

npm install goldenflow
npx goldenflow-js transform data.csv

import { TransformEngine } from "goldenflow";

const result = new TransformEngine().transformDf([
  { name: "  JOHN  ", phone: "(555) 123-4567", email: "John@Example.COM" },
]);

Highlights:

• 83 transforms across 11 categories (text, phone, names, address, dates, email, numeric, categorical, identifiers, URL, auto-correct)
• Edge-safe core — runs in browsers, Cloudflare Workers, Vercel Edge Runtime (import from "goldenflow/core")
• Node layer — CLI (12 commands), CSV/JSON connectors, MCP server (10 tools), REST API, directory watcher, scheduler
• 5 domain packs — people_hr, healthcare, finance, ecommerce, real_estate
• LLM-enhanced corrections — edge-safe via raw fetch() to Anthropic/OpenAI APIs
• 71 tests — smoke, parity, CSV parser, engine error handling, transform correctness
• Dual ESM + CJS build with TypeScript declarations
• Zero type errors with strict TypeScript (noUncheckedIndexedAccess, exactOptionalPropertyTypes)

Security Fixes (TS)

• Path traversal prevention in REST API and MCP server
• LLM corrector validates JSON responses and logs errors
• Config loader throws on invalid YAML (no silent defaults)

Architecture

• goldenflow/core — edge-safe (no Node APIs)
• goldenflow/node — Node 20+ (file I/O, MCP, CLI)
• Same YAML config format as Python
• Same transform names, same behavior

New CI

• TypeScript typecheck + test + build in GitHub Actions
• npm publish workflow on goldenflow-js-v* tags

Examples

• 6 TypeScript examples: zero-config, configured, schema-mapping, streaming, profiling, edge-safe

Full Changelog: benzsevern/goldenflow@v1.1.2...v1.2.0

v1.1.2 — CVE fixes

Bump python-multipart to >=0.0.20 (fixes CVE-2024-53981, CVE-2026-24486) and pydantic to >=2.7 (fixes CVE-2024-3772).

v1.1.1 — Silence Smithery -32601

Register empty list_resources/list_prompts MCP handlers.

v1.1.0

What's New

33 new transforms (43 to 76 total)

Three new modules and expansions across all existing categories:

Category	New Transforms
Email (new)	email_lowercase, email_normalize, email_extract_domain, email_validate
Identifiers (new)	ssn_format, ssn_mask, ein_format
URL (new)	url_normalize, url_extract_domain
Text	remove_html_tags, remove_urls, remove_digits, remove_emojis, fix_mojibake, normalize_line_endings, extract_numbers, pad_left, pad_right
Phone	phone_country_code
Name	nickname_standardize, merge_name
Address	country_standardize, unit_normalize
Date	datetime_iso8601, extract_year, extract_month, extract_day, extract_quarter, extract_day_of_week, date_shift, date_validate
Numeric	to_integer, abs_value, fill_zero, comma_decimal, scientific_to_decimal

Fixed GoldenCheck integration

• --from-findings now works (was using invented check names that matched nothing)
• Profiler bridge uses GoldenCheck semantic types (person_name to name, currency to numeric)
• Safer LLM fallback with logged warnings

Bug fixes

• comma_decimal no longer corrupts US-format decimals
• email_normalize preserves invalid emails instead of silently dropping
• merge_name handles missing last_name column gracefully

Stats

• 234 tests (up from 158)
• All tests pass on Python 3.11, 3.12, 3.13

v1.0.0

First stable release. GoldenFlow is production-ready.

• 43+ built-in transforms across 8 categories
• 5 domain packs (people/HR, healthcare, finance, e-commerce, real estate)
• Zero-config and config-driven modes
• MCP server with 10 tools
• DQBench Transform Score: 100/100
• 170 tests passing

See CHANGELOG.md for full details.