fix(timeline): a post could be liked and disliked at once, and the Cat's replies were unlabelled

[catomean]

Two things found sweeping the timeline in production.

1. Liking and disliking were not actually exclusive

Like a post, then dislike it: both buttons render as active, and the like count keeps a number with no row behind it.

The RPCs do enforce exclusivity — liking DELETEs any dislike row, and the reverse. What neither did was recompute the count belonging to the reaction it had just removed:

timeline_likes         no row for (event, user)   ← correctly deleted
timeline_event_stats   like_count = 1             ← never recomputed
                       dislike_count = 1

That number is wrong for every reader, not just the person who clicked, and it stays wrong — nothing recomputes these except the next reaction on the same post.

All four functions now recount both totals whenever either changes and return both. Toggle-off cannot move the opposite count but returns it anyway, so all four responses have one shape. Existing skewed rows are recounted from the membership tables, which were right all along.

DROP + CREATE was forced by the changed return type, so grants are restored explicitly — a reaction that silently stopped being callable by authenticated would look exactly like the bug being fixed.

The client half was the same mistake from the other side: handleLike only wrote userLiked/likesCount, leaving the opposite button lit. It now clears the opposite optimistically and takes both totals from the response instead of inferring them; a failed switch restores both sides.

Rehearsed against production in a transaction and rolled back: like → (1,0), dislike → (0,1), rows and cached stats agree, one already-skewed row repaired.

2. A Cat reply looked like a person's

is_cat_reply: true is stamped on every Cat reply under a comment saying it exists "so the UI can render a Cat reply distinctly rather than leaving a reader to work out from the avatar that this one was written by an agent". Nothing read it — two writers, zero readers.

That is the actual gap versus how @grok works: the threading was already right (the Cat replies as a child of the post that tagged it), but the model's replies were not visibly the model's. Attributing machine-written text to a human reader is not a styling detail.

Cat-authored posts now carry the same AI badge the mention menu already uses — one definition of what the Cat looks like. Author is the primary signal; the metadata flag is honoured too. @catalogue is not badged.

Both proven by mutation.

🤖 Generated with Claude Code

https://claude.ai/code/session_012dpTLxh5GJWeWTF1UEvcD5