Pinned Messages | Zyeon UI

status='ready' — four pins that fit. Click a chip to jump (onJump gets the messageId), press ← / → to walk the strip, Delete on a focused chip opens the same confirmation the × does. Unpin fires once per confirmation and the row really disappears; focus lands on the pin that took its place instead of on <body>.

Pinned4

no callback fired yet

Twelve pins on one line. How many stay is measured, not guessed: hidden twins carry every chip's natural width, and the split costs the '+N more' button itself, so a chip is never dropped to make room for a wider trigger. Drag the window narrower and the strip re-splits before it paints. Everything past the fold is in the panel — open it and unpin from there too.

Decision: we keep the v1 endpoint alive until March.Rate limit is 60 req/min per key, burst 120.deploy_preview → https://pr-482.preview.internalNever expose the internal cost figures.The UTF8 port is down, input the primary microchip so we can index the ADP monitor!We need to override the digital ASCII panel!I'll compress the online XML firewall, that should driver the UDP sensor!Try to back up the SSL interface, maybe it will synthesize the online bandwidth!I'll input the cross-platform PCI bandwidth, that should capacitor the ASCII port!The SAS pixel is down, quantify the neural interface so we can input the API sensor!generating the firewall won't do anything, we need to copy the open-source HEX driver!If we program the feed, we can get to the SSD matrix through the redundant API alarm!+12 more

Pinned12

no callback fired yet

The same twelve pins in a 28rem column. Nothing wraps to a second line and nothing is silently dropped: the strip keeps the leading run that genuinely fits and the counter on the trigger always adds back up to twelve. Unpin from the panel and the count follows.

Decision: we keep the v1 endpoint alive until March.Rate limit is 60 req/min per key, burst 120.deploy_preview → https://pr-482.preview.internalNever expose the internal cost figures.The UTF8 port is down, input the primary microchip so we can index the ADP monitor!We need to override the digital ASCII panel!I'll compress the online XML firewall, that should driver the UDP sensor!Try to back up the SSL interface, maybe it will synthesize the online bandwidth!I'll input the cross-platform PCI bandwidth, that should capacitor the ASCII port!The SAS pixel is down, quantify the neural interface so we can input the API sensor!generating the firewall won't do anything, we need to copy the open-source HEX driver!If we program the feed, we can get to the SSD matrix through the redundant API alarm!+12 more

Pinned12

no callback fired yet

The payloads that break naive strips: a 250-character excerpt (clipped to max-w-40 with the full text still in the accessible name), an excerpt full of newlines (collapsed to one line before it is measured), an all-whitespace excerpt (falls back to '(no text)' so the chip stays clickable) and a pin timestamped in the future by a skewed clock (reads 'just now', never '-5 m ago' — open the panel to see the ages).

Pinned4

no callback fired yet

Read-only: no onJump, no onUnpin. The chips become plain labels — no pointer cursor, no hover, no tab stop and no × at all. An affordance that leads nowhere is worse than no affordance, so the component refuses to draw one.

Pinned4

Run the migration with --dry-run first; it prints the plan without touching the table.
Ship the strip behind the feature flag.
run_migrations → 4 applied, 0 skipped, 812 ms
Always answer in the customer's language.

status='loading' — the pin list itself hasn't arrived.

Pinned

status='empty' — the one branch that teaches what pinning is for, instead of saying 'No items'.

Nothing pinned yet. Pin a message to park it here — one click brings the conversation back to it, and everyone in the thread sees the same shortlist.

status='error' — the list failed to load, which is a different failure from a message failing. role=alert plus a retry.

The pins couldn't be loaded. request-id: 7f2c-9ad1