- Add standalone src/plugin-api.ts types and manifest apiVersion check
- Add ctx.media playback snapshots, ctx.signal, ctx.onDispose, and awaitable ctx.mpv.request
- Bound async activation; actions wait only for their own plugin
- card-mined now carries per-note provenance (media + subtitle captured at mining start) through merges and YouTube queues
- Port episode-tags and auto-jimaku to provenance and media snapshots; drop mining-goal docs
- episode-tags: match each note id to the file history instead of the current media, so late card-mined events (YouTube media, field grouping) get the right episode
- episode-tags: group notes per file into one addTags call; skip cards added before the plugin started
- auto-jimaku: track a generation counter so a pending check goes stale when the same file reloads
- docs: describe the per-card episode matching in plugins.md
- episode-tags: tag mined cards with show/episode parsed from the file name or stream title
- auto-jimaku: open Jimaku search when a file starts without a Japanese subtitle track
- Hold card-mined for notes merged into a note still waiting on queued YouTube media
- Route YouTube media settle reports through reportCardsReady
- Document both plugins in docs-site and changelog fragment
- Add ctx.anki.on('card-mined') fed by polling, proxy, and card creation once enrichment finishes
- Add ctx.anki.invoke to send one AnkiConnect action to the effective url with no retries
- Ship example anki-sync plugin that syncs to AnkiWeb after mining goes quiet or playback ends
- Unload plugins that finish activating after unloadAll gave up on them
- Document the Anki API and anki-sync plugin in docs-site and architecture docs
- Escalate timed-out or aborted plugin exec children to SIGKILL after a grace period
- Move ai-explain example to assets/plugins and convert it to CommonJS
- Update plugin docs for the new path and the .js/.mjs module formats
sudacode