v1.1.0

Optional Harper pre-pass

Turn it on in Options to fix mechanical mistakes with Harper — a fast, rule-based, fully-local grammar engine — before the AI runs.

• Harper instantly fixes the deterministic stuff (capitalization, punctuation, spacing, subject-verb agreement, repeated words), offline.
• The local AI then only handles fluency and word choice.
• Lexical guesses (spelling, typos, word choice) are deliberately left to the AI, which has the full context.
• Off by default. Stays 100% on-device: Harper is WebAssembly run in the offscreen document, loaded only when enabled — the default built-in-AI path never pays for it.

Full Changelog: v1.0.0...v1.1.0

v1.0.0

Full Changelog: v0.2.0...v1.0.0

v0.2.0

Full Changelog: v0.1.0...v0.2.0

v0.1.0 — Track-Changes for any textarea, locally

First release.

What it does: press Alt+G in any textarea / text input / contenteditable and your text comes back as a reviewable Track-Changes diff — deletions struck through, insertions in green, each accepted (✓) or rejected (✕) individually. Corrections come from your own OpenAI-compatible endpoint (default: local Ollama + llama3.2). Nothing leaves your machine.

Design guarantees

• The LLM never produces the diff — it returns corrected prose; hunks are computed by a deterministic word-level diff.
• Your text is never modified until you accept.
• Works with stock Ollama: the extension strips the Origin header (declarativeNetRequest), so no OLLAMA_ORIGINS setup.

Install (unpacked): clone → npm install && npm run build → chrome://extensions → Load unpacked → dist/.

18 unit tests; smoke-tested in Chromium against live llama3.2.