Skip to content

[6.1.0] Enable Miteiru background server, updated UX

Choose a tag to compare

@hockyy hockyy released this 16 Jun 01:35

What's New in Miteiru 6.1.0

Changes since v6.0.0.

Features

  • Local analyzer server — While Miteiru is running, a localhost HTTP server exposes the same tokenization used by subtitles. External tools can call GET /health, POST /analyze, and POST /analyze-batch. Registration details are written to analyzer-server.json in the app user-data directory. See Analyzer Server.
  • Official analyzer CLIscripts/miteiru_cli.py is a thin client for the running desktop app. Examples:
    • python scripts/miteiru_cli.py health
    • python scripts/miteiru_cli.py analyze "見たよ"
    • python scripts/miteiru_cli.py analyze --stdin < lines.txt
  • Unified video page layout — The video player, subtitles, live captions, sidebars, and dropzone now share one render tree. The player sits behind the subtitle overlay instead of switching to a separate screen when media loads.
  • Linux 26.04 builds — Added npm run build:linux26 for Ubuntu 26.04 .deb and AppImage artifacts.

Improvements

  • In-house Japanese analysis — Replaced the shunou dependency with a local analyzer module. Kuromoji loading is now lazy and shared across requests.
  • Centralized tokenization — All languages route through a single analyzeText path in the main process, used by both the UI and the analyzer server.
  • Language asset tooling — Copy and unpack scripts now validate asset-manifest.json contents, warn only for missing required paths, and unpack archives with PowerShell on Windows or unzip elsewhere.
  • Dependency updatesto-jyutping moved to the published npm package (^3.1.1); added wanakana. New npm run update:safe updates packages only if they were published at least 7 days ago.
  • Windows packaging — NSIS, portable, and default Windows builds now bundle the full live-captions folder instead of only the bridge executable.

Fixes

  • Vocab sidebar — Japanese learned words now display as readable text instead of raw token objects.
  • Vocab sidebar — Ignores invalid/non-array tokenization results instead of crashing.
  • Word of the Day — Removed unnecessary z-index stacking that could interfere with overlays.

Developer notes

  • Added npm test (tsx --test tests/**/*.test.ts) with coverage for the analyzer server and Japanese analysis.
  • README updated for the local Japanese analyzer and Linux 26.04 build target.

Full Changelog: v6.0.0...v6.1.0