v1.1.0 — Extraction types + bug fixes

First minor release. Adds 8 specialised extraction prompts (article, product, contact, event, job, recipe, review, plus auto), fixes a handful of real bugs in the HTML→Markdown converter and request handling, and adds safer defaults (timeout + size cap).

✨ Added

• extractionType request field — one of auto, article, product, contact, event, job, recipe, review. Auto-detected from your schema if not passed.
• 15 s fetch timeout on the target URL, surfaced as TIMEOUT_ERROR (504).
• 100 KB markdown cap before content reaches Gemini — prevents silent token-limit failures.

🐛 Fixed

• Ordered lists were broken. <ol><li>foo</li></ol> was rendering as 1. $1 because $1 inside a String.replace callback return value is a literal string, not a backreference. Now correctly renders 1. foo.
• Code block indentation was being destroyed by an over-eager whitespace collapse that ran after fenced blocks were created.
• Invalid JSON in the request body now returns 400 VALIDATION_ERROR instead of 500 INTERNAL_ERROR.
• Removed a misconfigured Accept-Encoding header that could cause some servers to return content the worker couldn't decode.

🔧 Changed

• Bumped Cloudflare Workers compatibility_date 2024-04-05 → 2025-05-01.
• Dropped the unused REST-based extractWithGeminiRest extractor.
• README rewritten to document the 8 extraction types, the full error code taxonomy, safe defaults, and an honest "this is a Worker, not an npm library yet" framing — a proper client SDK is on the v2 roadmap.

Full changelog: CHANGELOG.md · Diff: v1.0.0...v1.1.0