v1.4.1 — Recordings the watcher finished are no longer stranded

Fixes a bug in v1.4.0. If you used the watcher, upgrade.

The bug

Stopping a watch-agent recording from the menu bar left /meetey stop reporting "No active recording" — with the audio sitting on disk and no obvious way to reach it. The post-stop notification tells you to run /meetey stop, so it pointed straight at the dead end.

A recording started by the watcher ends in a different process from the MCP server. That process clears the "something is recording" state, which is correct, but nothing was left behind for /meetey stop to collect. Recordings started from /meetey start were unaffected.

The fix

Finished-but-uncollected recordings are now recorded on disk, so they survive the process that made them — and a Claude Code restart between stopping and writing up.

When there is genuinely nothing to collect, /meetey stop no longer dead-ends: it names the recordings that have audio but no notes, and offers to write the most recent one up.

Nothing was lost

Recordings affected by this are still in ~/.meetey/recordings/, and the watcher had already transcribed them. After upgrading, /meetey stop will point you at them.

Upgrade

npx jej2k5/meetey update

Then restart Claude Code so the MCP server reconnects.