v0.1.0 — first preview

Changelog

All notable changes to this project are documented here. Format loosely follows
Keep a Changelog; this project uses SemVer.

0.1.0 — unreleased

First public preview. A schema-driven, category-agnostic engine: one product
dataset → many asset formats × platforms, with localization.

Added

• Core — neutral product schema (schema/product.schema.json) with
  variable attribute_groups + dynamic-column spec_table + i18n; loader,
  validation, and a Jinja rendering engine; listingforge CLI
  (build / localize / adapt / kit).
• Templates (3 asset formats, same data): sku-card (spec sheet),
  social-square (1:1 creative), detail-page (A+/detail long-page).
• Platform adapters: Amazon listing fields (title/bullets/description, with
  limit validation; bullets synthesized from attributes when absent) and a
  Google Merchant Center TSV feed.
• Localization (M3): offline glossary, human/LLM translation-memory file,
  and a real Claude AnthropicTranslator — layered (MemoryTranslator >
  glossary > LLM).
• kit: one command → localize + all templates + all platforms + a manifest.
• Data tools: scripts/ingest_sa.py (vendor pages → schema),
  scripts/anonymize.py (de-identify a dataset into a shippable example).
• Examples: apparel-demo (synthetic) and fly-demo (anonymized,
  dense-parameter fishing case — fictional "Northbank").
• Project: Apache-2.0, pyproject.toml + console entry point, stdlib
  unittest suite, GitHub Actions CI (py3.9/3.11/3.12 + build/kit smoke),
  CONTRIBUTING.md.

Known limitations

• One rendering path (HTML → open in a browser); PNG export needs Playwright.
• Adapters cover Amazon + Google Merchant only.
• No upstream messy-data ingestion, online playground, or hosted service yet.