AIterval v0.2.1

AIterval v0.2.1

AIterval v0.2.1 improves the real listening experience and accessibility.

Changes

• Fixed unreadable Shadow DOM overlay colors.
• Improved light and dark theme contrast.
• Added a visible and accessible AI-ready status.
• Listening audio now continues to the end after the host AI finishes.
• Added an explicit Return to AI now action.
• Prevented duplicate sessions and recovered-time updates.
• Expanded accessibility, theme, state-machine, and E2E regression coverage.
• Added no browser permissions, analytics, tracking, or runtime AI-generation
  dependency. Learning data remains local-first.

Chrome Web Store publication is not complete. Download the ZIP, extract it, and
load the extracted folder containing manifest.json through Developer mode.

Full Changelog: v0.2.0...v0.2.1

AIterval 0.2.0 — OpenAI Build Week Edition

AIterval 0.2.0 — OpenAI Build Week Edition

Highlights

• Lecture-to-Sprints: turn an explicitly submitted upcoming lecture abstract into 1–5 schema-validated listening exercises with GPT-5.6.
• Current OpenAI Responses API and Zod Structured Outputs with server revalidation, trusted provenance, and one bounded schema retry.
• Public no-login judge journey plus clearly labeled curated sample mode.
• Safe JSON export/import keeps the extension’s host permissions and CSP unchanged.
• Generated-pack progress, pause/resume, rename, export, regenerate, delete, source badge, and scheduling.
• Aggregate-only opt-in personalization without raw session history.
• Storage schema v2 migration and 38 unit/component/API tests plus Chromium web/extension E2E.
• Accurate external-call privacy copy, server-only credentials, access-code enforcement, timeout, rate limits, and daily quota.

Install

Download the Chrome ZIP and SHA256SUMS.txt, verify the checksum, unzip, and load the extracted manifest.json directory through chrome://extensions Developer mode. See docs/installation.md for the full no-build path.

Compatibility and limitations

Current desktop Chrome is supported. Automatic wait detection depends on third-party UI; manual start remains available. Browser speech quality depends on installed voices. Live GPT-5.6 generation requires a configured server key and private demo code, while built-in exercises and the public sample require neither. This release is not a Chrome Web Store publication.