herdr web ui v0.3.19

Added

• Quick replies: one-tap messages above the chat's message box (continue, yes, no,
  commit and push, retry to start with). Each is sent as if typed: queued while the agent works,
  taken as the answer when a question is open, and a draft in the box stays. Settings → Quick
  replies edits the list on each device; the row is hidden until the ⚡ button beside the paperclip
  shows it.
• The chat's status line says how much context is left: 74% left from the window Codex records,
  or for Claude once a request has run past 200k (the 1M window); 68k used where the transcript
  names no window (omp, omo, gjc, and Claude below 200k). It turns red at 20% left. Hovering shows
  the token counts.
• A tool call that failed says so: its row reads "failed" in red, its output is headed Error, and
  the folded "Worked for…" header counts the failures. Claude and omp record the failure; for Codex
  it is read from the output (a command that exited non-zero, a script or patch that failed).
• A file a tool call names (an edit, a read, a patch's files) opens in the viewer on a tap.
• A Codex patch reads as a diff: each file as a header, then its lines in red and green, and the row
  is summed up by the files it touches, also when an exec script applies it.
• While reading further up the chat, a ↓ button takes you back to the end, new messages or not.
• Images you sent show above your message: ones pasted into Claude's prompt, fetched only when
  shown, and ones attached here (an @…png mention). A tap opens them.
• An edit reads as one diff: unchanged lines once, removed and added lines in place between them,
  instead of the whole old block and then the whole new one. Several edits to one file each get
  theirs.
• Where a compaction folded a Claude conversation, a divider says so, and opens to its summary.
• Headings down to ###### render as headings.
• A tool output cut for the chat (past 4,000 characters) has a "Show the whole output" button that
  fetches the rest, up to 2 MB, into a box of its own.
• The / menu lists Claude's skills (yours and the project's) and the skills and commands of the
  plugins turned on in its settings, as /<plugin>:<name>. In a Codex pane it lists your saved
  prompts as /prompts:<name>, and $ opens Codex's skills.
• On a phone, the terminal lens has an input line above the key bar. A phone keyboard rewrites
  what it typed (dictation revising a phrase, a Korean syllable being composed, autocorrect), and
  the terminal could not take back keys it had sent, so every revision arrived as more text. The
  line is written with the keyboard's own editing and goes to the pane whole, then Enter (typed
  like the keyboard, into an agent's open menu too); an empty line's button presses Enter alone.
  Tapping the terminal no longer raises the keyboard; the ⌨ key on the key bar switches to typing
  straight into it, remembered per device. Desktops are unchanged.

Changed

• Remote PCs use the remote-v5 runtime, which carries this release's server side to them: the
  context left and failed calls in their chats, images and whole tool outputs, skills in the menu,
  questions read in a narrow pane, and the terminal's input line. A connected PC's bridge updates
  itself as for remote-v3.
• Alerts wait before they go out, and a change of the pane meanwhile calls them off: a question
  answered at the PC within 10 seconds, or a finish followed by the next prompt within a minute,
  never buzzes the phone. A finish is told only after a turn that worked a minute or more, since a
  quick answer is read where it was asked.
• Settings → Alerts chooses, per device, whether a waiting agent alerts, and whether a finish
  does never, after a long turn (the default) or every time (then after 10 seconds). An ended
  terminal follows the finish choice.

Fixed

• An agent's chat said "No conversation yet" until its first answer arrived, seconds on a remote PC.
  It now says it is loading.
• A Claude question never showed as a card in the chat when its pane was narrow (a phone, a split):
  the hint under the menu wraps (… Esc to / cancel), and it was looked for on one line. Hints
  are now read across the wrap, in every agent's menus, and so is the check that a menu is still
  open before an answer is typed into it. A single question's card is titled by its header chip.