Search is the one capability that fits the profile holding none: every other capability either
reads this installation or changes it, and chat's whole contract is that it can do neither. The
assistant deliberately does not get it — its contract is that answers come from this install's
own documents, and a web result there is an answer that looks sourced and is not. A chat question
about this machine hands off to the assistant before the search would run, so it never reaches
the internet even with the box ticked.
Off by default, and not because it is dangerous: searching sends the operator's question outside
the house, which is theirs to decide. Asked for per turn as well as enabled in conf.
Provider-agnostic, as asked — searxng, brave, tavily. Only whichever is configured here can be
verified; all three read every field by name, so a shape that changes underneath yields no
results rather than wrong ones.
The explain fixtures asserted chat holds no capabilities at all, which is exactly the guarantee
worth keeping. caps=only: now states the set rather than its emptiness, so the check still fails
the day something else is granted there.
The 55 alert() calls carried the same suppression as the confirms, and go wrong in the worse
direction: a silenced confirm makes a button do nothing, while a silenced alert lets the action
run and says nothing about it failing. vvAlert returns a promise nobody has to await, so these
converted by rename with no caller becoming async. The icon is inferred from the message rather
than asked of fifty call sites, and an explicit type still wins.
Every confirm() and prompt() in the plugin could be switched off from inside itself — one tick of
"prevent this page from creating additional dialogs" and all 32 of them returned false while
drawing nothing, across every tab, until a full reload. swal is already global on every webGUI
page and core uses it 370 times without a single confirm(), so this costs no new dependency.
vvConfirmRun() is the one that mattered: it returned a boolean to three callers testing !it, and
an unawaited promise is always truthy, so leaving those alone would have run every job without
asking. The wrapper's callback is a classic function expression on purpose — SweetAlert only
calls back on cancel when the callback's own source declares a parameter, and an arrow would
have hung the promise forever.
Both browsers offer "prevent this page from creating additional dialogs" inside the dialog
itself, and once it is ticked every later confirm() returns false without drawing anything — so
Fix and Never a problem silently declined and read as dead buttons. Unraid swaps tabs by AJAX
without reloading, so the suppression outlived leaving the tab and only clearing history got it
back. Both now arm on the first press and act on the second, and say what the press will do
while they wait for it.
Seven findings had been sitting in needs_operator with no way to reach the operator: the only
way to read one was the sweep tool over SSH. The card states the gate as well as the list,
because an empty list means nothing found when repair is on and nothing is looking when it is
off. Buttons come from the actions the endpoint returned for that row, so a tab left open
overnight cannot offer a choice the store has moved past.
The readmes still described a single expand glyph and a chevron for saved chats, and neither knew
about the banner, the two expanded sizes or the shortcuts list.
Counting every step reported a fully scheduled system as 10 of 82, and reopening the last thread
brought its profile with it, so the tab came up as Troubleshoot pointed at nothing.
Flex items default to min-height:auto, so a view holding a log ignored the height it was given
and pushed the composer out of a panel that clips instead of scrolling.
The Scheduler sizes the views above the panel from what the chat leaves, so a chat that resizes
silently grew off the bottom and took its own composer with it.
Several headers argued the Scheduler dock was deliberately separate, and the Scheduler's help
never mentioned the assistant at all — including the fix flow that just changed shape.
It kept its own bar, send loop and poll, which is why a thread there died on reload while the
other two were saved and why its controls had drifted from the same controls everywhere else.
The picker sat above the transcript and the buttons below it, so the same component read
differently on each surface; size is the only thing a placement should get to choose.
A stretched grid item is measured before it is stretched, so a long card grew the row it was
meant to fit inside and then filled it; the ceiling is taken from the Assistant instead.
The AI cards sat in the monitor grid, which caps every row at a quarter of the viewport and
clips the card, so expanding grew the transcript inside a box the row had already sized.
Run was the only mutating control here that did not confirm, and it is the one that executes
shell as root — the library holds cleaners that delete files, and six Run buttons sit a row apart
in a dense tree.
Every poll after the first asked for ?live=1, so the page paid a full collection — partner SSH
timeouts included — every two seconds while the tmpfs cache it was built around went unused.
Polls are now guarded against overlap and stop while the tab is hidden, repeated failures say so
instead of leaving the last good reading on screen, and two functions nothing called are gone.
The response was discarded, so a refused stop looked like a completed one, and the payload the
card redraws from is cached for 300s with no invalidation — the container carried on showing as
running until the once-a-minute writer caught up. Stop now confirms; start still does not.
The page interpolated media titles, partner hostnames read over the mesh, and docker folder
names straight into innerHTML — its own header claimed otherwise, and the helpers that would
have fixed it were defined in a page it never loads. Container WebUI values now get a scheme
check before they reach window.open().
The Tools card adopted the AI scripts on the host check alone, and api/ai.php only
tested AI_ENABLED in front of ask, so a disabled subsystem still had rows to run and
an endpoint that answered.
Gmer4Lfe