* fix(mobile-relay): resolve on persistent 4404 and make deferral logs readable
* fix(mobile-relay): bound 4404 director resolves for a desktop that stays offline
Review follow-up:
- Resolve on the 2nd consecutive 4404 of an episode, then the 4th, 8th, 16th,
and never twice within 5 minutes per host. A sleeping laptop with a
foregrounded phone now costs 6 resolves in 30 minutes, not ~1 per 2 dials.
- If the director names the same cell and epoch, keep the normal 4404 wait:
no re-route and no immediate redial.
- The streak ends on any other outcome (including aborts), when the phone has
connected since the last 4404 (a direct-path win), or after a 2-minute gap.
- Clear the gate reprobe due time with its timer so the deferral log never
reports a stale remainingMs.
* fix(mobile-relay): double the gap between 4404 resolves while the desktop stays offline
The 5-minute floor alone set the pace (12 resolves/hour, forever). After each
resolve the minimum gap now doubles, 5 -> 10 -> 20 -> 40 -> 60 min, and
returns to 5 min after a successful relay dial or any connection. A laptop
asleep for two hours with the phone in the foreground costs 5 resolves.
* refactor(mobile-relay): namespace the supervisor's diagnostic log import to stay within max-lines
Desktop: drop IPv4 169.254.0.0/16 (and, as before, IPv6 fe80::/10)
addresses from the pairing interface list, so an APIPA or idle adapter is
never the QR's direct endpoint. When nothing else is usable the offer
already omits the direct address (Relay carries the invite; LAN-only
refuses). Among non-Tailscale candidates, an adapter known to hold the
default route now wins ties.
Mobile: a relay-paired host whose stored endpoint is link-local seeds its
logical client with a never-dialing 'reconnecting' stand-in and disables
the direct return probe, so already-paired phones go straight to Relay
instead of redialing the dead address. Relay-less hosts still dial.
Reconciles the reviewed automatic-refresh parent with the setting child:
the child's refresh/lookup owners already carry the parent's completion
contract, so they are kept and the parent's read-admission adapter is
dropped; main's moved modules take the child's imports at their new paths.
Co-authored-by: Jinjing <6427696+AmethystLiang@users.noreply.github.com>
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Use the accepted provider completion for publication and retry planning. Recheck live demand before automatic starts, preserve user actions under Off, and keep the current failure schedule after interval changes.
Co-authored-by: Jinjing <6427696+AmethystLiang@users.noreply.github.com>
* feat(native-chat): show each subagent's model, current tool and result in the roster
Subagent roster rows now carry the child's role, model, reasoning effort,
the tool it is running and the last thing it said, so a row says what the
subagent is doing and, once done, what it found. Codex children take model
and effort from the spawn call; Claude children take the model from their
own messages. A hover card shows the longer text.
* feat(native-chat): name subagent models and roles the way people say them
Show Opus 5.5 · High instead of claude-opus-5-5 · high, reusing the
model-id label and effort picker copy, and General purpose instead of
general-purpose.
* fix(native-chat): read Codex subagent model and effort from its started activity
Current Codex announces a spawn only through subAgentActivity, which carries
the model and reasoning effort the child resolved to.
* fix(native-chat): use the hover card's own spacing variant for the subagent preview
* fix(native-chat): show a subagent's reply as plain text in its roster row and preview
Its last message is markdown; the row now drops backticks, emphasis, list
and heading marks and joins lines into one readable line.
- Checks: overlapping reads of one PR head share one record kept until the last
of them settles, so a read replaced by a user refresh that has already ended
still answers with a newer manual answer after it was evicted, rechecked after
waiting on its successor.
- A row shown again after its settled age reads instead of being answered by a
cached settled result; one foreground age decides both whether the return
earns a read and whether the current cached answer satisfies it.
- Off back to Default restores a failed row's own retry from its failure, still
an error retry the next Off retires, without a new report or manual action.
- Test fixes: advance the replaced job's start cooldown, run the checks reads
without command-slot limits, typed fixtures and clock-rule notes.
Co-authored-by: Jinjing <6427696+AmethystLiang@users.noreply.github.com>
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
- A window return or re-shown row that is due under the Default interval now
reads past the stretched timed cache instead of being answered from it.
- Turning auto-refresh Off also retires a host revalidation that was waiting
for quota, so it never starts or requeues afterwards.
- A newer manual checks answer stays authoritative while an older read of the
same checks is still out, even when 256 other checks fill the cache.
- A paired host keeps no auto-refresh state after it is forgotten; a read from
before unpairing never answers for it, and a re-paired host is read afresh.
- A host's cadence change re-plans only the visible rows that host runs, and
this app's own setting re-plans its rows in main instead of re-reporting.
- Settled rows wait at least a minute on return under a stretched cadence;
Default keeps its cooldown.
- The setting's copy no longer promises a read on selecting a workspace.
Co-authored-by: Jinjing <6427696+AmethystLiang@users.noreply.github.com>
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Brings in the automatic host-side PR read sharing (#42033) and finishes the
auto-refresh setting on top of it, so one owner per resource decides when PRs
and checks are read again:
- The setting now stretches the host's own interval, including the five-minute
quiet interval for settled open PRs (Default 5, Reduced 25, Minimal 50 min,
Off none), both for paired clients' timer reads and this host's own queue.
- The earlier duplicate PR answer reuse for paired clients is removed; the
shared host result owner now serves those reads.
- Switching cadence re-times errors, their shared retry time, host gates and
queued host reads from when the read failed; Off drops them without pinning
anything forever, and Default restores the original retry time.
- Window return reads data older than the Default interval in every mode and
never sooner, so Off no longer re-reads no-PR or closed rows on each return.
- Checks keep their own per-head result, now recorded by the actual checks read
in start order, so a manual refresh cannot be overwritten by an older read
and a read after a PR change never joins one from before it.
- Phones send timer reads as revalidations, keep what is shown on a receipt and
read again soon; returning to the app reads fresh. No foreground wire flag.
Co-authored-by: Jinjing <6427696+AmethystLiang@users.noreply.github.com>
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
* test: remove low-value push and relay coverage
* test: delete orphaned fixture and closure support
* test: trim runner and mobile recording matrices
---------
Co-authored-by: OrcaWin <293788423+OrcaWin@users.noreply.github.com>
The tab bar built every tab's sheet row, including the agent preview over
each terminal's last message, on every tab snapshot and 60 s tick, and the
always-mounted closed sheet ran its own clock and row elements. The bar now
builds only the title/dot/group summary it draws; the sheet's clock, frozen
groups and detail rows live in content mounted only while the drawer is
shown.
The queued list cached each row's height by message id for the life of the
chat, so a long conversation kept one entry per message ever queued and a
requeued id could place a failed row by the height it left with. Keep heights
only for queued ids, and drop a layout a departed row reports late.