On the game PC agent-desktop's press --window-id answers APP_NOT_FOUND for
the new Calculator ("not found with exact process identity", nothing sent),
though focus-window brings it in front. The keys now go to the window in
front while it is still that window. The Windows test checks Notepad's typed
text by element values only (a label such as Ctrl+Shift+X contains "hi").
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
User decision 2026-11-19 (mixed way). agent-desktop only presses keys in the
window in front, so the app now picks one way before any key is sent:
- behind: one key without modifiers that types or moves is posted to the
window's one shown text control (Edit, RichEdit, RichEditD2DPT, Scintilla);
the person's screen and focus stay. Windows does not tell a background
thread's focused control (win-ci VM: GetFocus is null even attached), so
only windows with exactly one text control get keys behind.
- in front for a moment: combinations, other keys and every other window;
the window that was in front is brought back after (also after a failed
press). A window already in front gets the keys there.
- never in front while game mode is on or the window in front is full screen.
- never both ways, so a key is never pressed twice.
Specs: admin.spec.md "PC control" (Keys), pc-features.md, pc_control.spec.md.
Tests: six new fake-desktop tests in pc.rs; tests/pc_keys_windows.rs on a real
desktop (win-ci VM: letters behind Notepad with Character Map kept in front,
ctrl+a in front and back, the classic Calculator in front and back).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
From the live conversation on 0.1.40 (2026-11-18):
- The bot browser told pages a program drives it (navigator.webdriver), and
YouTube stopped every video about 45 s in with "문제가 발생했습니다"
(ump.spsrejectfailure, reproduced twice on the game PC). Chrome now starts
with --disable-blink-features=AutomationControlled, as the M0 video check
already did; the live Chrome test checks webdriver is false.
- "네." to "계산기를 실행해도 될까요?" got webSearch and botBrowser from the
router, so the consent call was refused and the calculator never started.
Waiting tools: pcControl stays on offer while a request asked in the
previous reply waits, botBrowser while it holds the asker's page.
- "소리 반으로 줄여 줘" for the bot's video worked on the person's own Edge
window. botBrowser gets a volume action (media elements and YouTube's player),
the page shows its volume, its description names its actions first (the part
the planner sees), pcControl's says the bot's browser is not among its
windows, and a turn whose plan has no step while a tool holds something for
the asker gets the one-step nudge for it.
Evals (qwen3.6 on the game PC): volume 0/3 before, 8/8 after; "네." consent
10/10; pcControl 10/10; botBrowser 16/18 (the NewJeans play case fails 0/6
before this change too). Unit: Python 2542 passed, 13 failed (the same 13 as
before); app 348 passed.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Checked against agent-desktop 1.2.0 on the Windows test VM (normal and
elevated sessions); each fault has a test:
- the state folder is made private to the current user: a data folder
outside the user profile inherits Authenticated Users: Modify and
agent-desktop then refused every reading (INTERNAL), which broke all
of PC control on such PCs
- run with the program already open uses the window agent-desktop names
instead of waiting for a new one (failed after 5.3 s, now 0.6 s);
several copies get their own reason
- keys to a window Windows will not bring to the front says so
(NOT_FRONT) and sends nothing
- close that a save dialog holds up, or several copies, get their own
reasons instead of a generic failure
- an action that closes its own window answers done with no elements
- fields a reading showed as secure are refused as password fields
before anything is sent
Verified on the VM: calculator 1 2 + 3 = read back 15, Notepad type,
ctrl+a and a letter, elevated windows refused from a normal session and
normal ones not, the emergency stop killing a running command and
blocking acting while reading works.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
app/src/pc.rs runs every pc_action from the worker (admin.spec.md, "PC
control"): one agent-desktop 1.2.0 run per command with a 15 s limit and
its state in the data folder, never --force or raw input, typed text
only through stdin. Each action is decided by permissions::decide for
its asker (not registered, never, ask without consent), then refused
during the emergency stop, while an anti-cheat game runs (Vanguard,
EAC, BattlEye, FACEIT, GameGuard, XIGNCODE3 and the person's list), on
an administrator window and in a password field. Results carry the
window and its elements read again (200 at most, cut to 120 chars),
agent-desktop's error codes become Korean reasons, and each action is
recorded as pc.<action> with the program's file name only.
The emergency stop is a global hotkey (pcControl.stopHotkey, default
Ctrl+Shift+F12) on its own message loop; it kills the running command
and blocks everything but reading until pressed again or lifted on the
PC control screen. features.pcControl (on by default, nothing happens
until someone is registered) and the pinned exe hash decide whether
the brain gets the tool. The installer ships agent-desktop.exe from the
pinned release archive, both hashes checked.
Tested with a fake agent-desktop (342 lib tests pass) and cargo check
for x86_64-pc-windows-gnu; not yet run against real agent-desktop.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
getWeather takes days_ahead (days from today, up to 10) and days (a span,
2 for a weekend). Inside Korea with the app's KMA key: 단기예보 for today
to 3 days ahead (morning and afternoon weather, lowest and highest, highest
chance of rain), then 중기기온 and 중기육상 for the nearest of 32 KMA cities
and its region. Days KMA does not give, and every KMA failure, come from
open-meteo's daily forecast. days_ahead 0 with more days gives the weather
now, then each day. Eval: evals/test_weather_forecast_days.py.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The 기능 screen's 기상청 날씨 card stores the KMA API hub key encrypted
(kmaApiKey, masked in logs). On start the app passes it to the worker in
JARVIS_KMA_API_KEY; the worker takes it out of its environment before any
child process starts. Inside Korea getWeather reads 초단기실황 and the
coming hour's sky from 초단기예보, and falls back to open-meteo, then
met.no, on any KMA failure. The key is in KMA's URLs, so failures are
logged by their kind only.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Free memory is nvidia-smi's memory.free in the app and the voice worker, so the
driver's reserved share no longer counts as free. A card reading still running
when an answer ends finishes on its own and then logs, so an answer shorter
than one reading is logged too, and the answer still never waits for it.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The app logs how much of the card is free once the model answers, and the
voice worker reads the card once a second while an answer or a scheduled task
is written and logs the least free memory seen (no text, llm tab), so a slow
answer can be matched to a full card.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
pcControl (v2 M7): windows, read, click, type, keys, run, close through the
app's agent-desktop. Ask-level actions wait for the asker's own next message
and then cover that window's asked kinds for the reply; payments, passwords,
deleting files and admin rights are filtered before asking and never done.
The app refuses during an anti-cheat game and the emergency stop.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
On the game PC, Wallpaper Engine (steamapps\common\wallpaper_engine) and the
Riot client's crash handler (Riot Games\Riot Client) were taken for games
running behind other windows as the services started, so no reading of other
programs was kept and the game stage placed 0 GiB of experts instead of 2.
- A launcher's own folders and non-game apps under its game folder are not
game folders (Steamworks Shared, wallpaper_engine; Launcher, Epic Online
Services; Riot Client).
- The game log tab names the program found running at the start.
- The placement keeps the TTS's share (MeloTTS, 1005 MiB, M0 tts_melo) beside
the STT's, and takes the worker's loaded models off what the card holds, so
a start (TTS not yet loaded) and a switch (TTS loaded) place the same.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
A chat channel answering every message no longer starts Strata with
--parallel 2: on the 16 GB card that read prompts about three times
slower (2026-11-16, answers took 10-20 s). Chat and voice answers now
take turns at Strata. response.policy defaults to latest.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
- botBrowser `play`: searches YouTube for the words as said, opens the first video (not an ad,
without the playlist) and starts it, in one call (an address given opens that page); `type`
also takes `query`. Live 0.1.19: the model stopped on YouTube's home page or the results, a
retry was refused as a third call, and the answer still said the video played.
- `brain.location` (기본 위치, default 대한민국) goes to the brain as `location_name`: the time line
says it without GeoIP (no database on the game PC), and getWeather without a place uses it.
- Planner and webSearch keep search words in the user's language (never translated into
English), so recommendations fit where the user is.
Evals (qwen3:8b, 3 runs): play ending on a playing video found with the words as said 0/6 -> 5/6;
Korean searches for Korean requests 6/9 -> 8/9.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
- 로그 screen: "보이는 글 txt로 저장" (logs_save_text), a 대화 view that saves the days'
conversation with every trace in full (history_export), and a 목소리 view with the
voice.record switch and the days' recordings as a zip with an index (voice_export).
- Voice worker: with JARVIS_RECORDINGS_DIR (voice.record, off by default) each final and each
spoken answer is kept as a WAV with an index line; day folders older than 7 days removed.
- response.interrupt (default on): off, latest/important finish the running answer first.
- Bot browser: Chrome lets pages start video and sound; botBrowser reports each page's media
state, waits for a player to start after open/click, and presses play on a video page a click
led to when its player stayed paused (live: a clicked YouTube video stayed paused).
- webSearch reads `query` when `search_query` is missing (live: both searches of an answer failed).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The screen re-read its window every 3 seconds unless a trace was already
shown as being made, so an answer that ended between two reads came with
its trace. It now asks twice a second for history_stamp (the newest
record and trace ids and the steps of the traces being made, cheap and not
logged) and reads the window again as soon as it changes.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The 한 일 screen is gone: 로그 has 로그 and 한 일 views switched at the top.
An update before 0.1.12 could install into Program Files while the desktop
shortcut kept pointing at the first install's folder, an old copy that no
update reaches: its jarvis.exe shows the old icon and starts a second, old
Jarvis next to the running one. The installer now points every desktop and
start menu shortcut to another Jarvis install's jarvis.exe at this install.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The separate 업데이트 screen is gone: its card is first on the 앱 screen
and a newer release renames the 앱 button. On 대화 a trace appears when
the question is recognised, opens and fills in step by step (re-read
every second while one is being made) and stays open with its answer.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
- The icon pass also walks the start menus' folders, so a shortcut kept
in one (and a later icon change) is not missed.
- The install folder and icon path reach PowerShell through the
environment: a folder name with an apostrophe broke the whole command.
A failure is now a line in the install details.
- InstallDirRegKey read the 32-bit registry view before .onInit switched
to the 64-bit one where the folder is written, so a silent update of
Jarvis installed elsewhere put a second copy in Program Files. .onInit
now reads the 64-bit key; a folder given with /D wins.
- Version 0.1.12.
Checked on the Windows VM with Jarvis 0.1.6 in C:\Tools\Kim's Jarvis and
shortcuts on both desktops and in start menu subfolders: 0.1.11 left the
desktop and subfolder shortcuts old; 0.1.12 (no /D) updated that folder
and every shortcut, and a fresh 0.1.12 still goes to Program Files.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
- The side menu has 대화 right under 대시보드.
- 대화 screen: a "현재 세션" chip (off by default) shows only what came
since the bot was last started. The controller keeps when the latest
start began; history_query takes `session` and answers it back, so the
count line names that start and a new start starts the list over.
- Installer: the shortcuts' icon is a file named by its content
(jarvis-<sha256 prefix>.ico). Windows keeps an icon it has shown by its
path, so after an update the desktop shortcuts kept the old icon. Every
.lnk to jarvis.exe on the desktops and in the start menus (an earlier
version's and the person's own too) now points at the new file, set by
the 64-bit PowerShell: a 32-bit one reads Program Files links as
Program Files (x86). Checked on the Windows VM: 0.1.6 to 0.1.9 left both
desktop shortcuts old; to this build they and search show the new icon.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
- New icon (voice bars in a ring on the accent gradient; three thicker
bars at 32 px and below) for the program file, window, taskbar,
installer and a tray icon of its own.
- 앱 settings screen with app.closeButton: "quit" (default, as before)
or "tray", read each time X is pressed (no restart). Tray mode only
hides the window; the bot and services keep running. The tray icon is
always there: a left click shows the window, its menu has 열기 and
완전 종료 (the same stop-then-exit as closing).
- tauri-plugin-single-instance: starting Jarvis again while it runs
shows the running one's window instead of a second app driving the
same services.
- Version 0.1.8.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
- paging moves to app/ui/talk.js (pure, node --test app/tests/ui/talk.test.mjs)
- items the full live window no longer reaches join the older pages instead of being dropped
- "이전 기록 더 보기" runs one request at a time and never adds a loaded item twice
- the screen and spec state that only the newest 2,000 items are re-read live
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
- history_query takes item cursors (before / from) ordered by time, source and id,
so paging never skips or repeats an item and has no 2,000 item ceiling
- live refresh re-reads the window from the oldest item shown and compares whole
items, so an edited message is redrawn
- answers to requests made before a filter change, and live answers older than
the newest one asked, are dropped
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The management app reads the worker's history store (read only) and shows
people's words, Jarvis's answers and the answer traces in time order, with
filters for kind, place, person, channel, day and text, live follow, older
items on demand and expandable traces (records given, tools offered, plan,
model thinking, tool calls and results).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
MeloTTS's own speed shortens every sound and slurred fast speech (Whisper
large-v3 on five Korean sentences, three renders each: 5.5% character
errors at 1.5 and 20% at 2.0, against 1.4% at 1.0). The MeloTTS child now
speaks at most at its natural pace and makes faster speech with a WSOLA
time stretch that keeps the pitch (1.9% at 1.5, 2.2% at 2.0, about 15 ms
per 4 s of speech).
A voice.tts.speed saved while the services run goes to the worker at once
(new tts_speed message) and applies from the next sentence; the client
sends it again whenever the worker connects again. Version 0.1.4.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The first-run setup writes voiceWorker.cwd; when it ran from another
install, an updated app kept starting that install's older sources (found
on the game PC: 0.1.1 ran without the screenView tool). Shipped sources
next to the executable now always win. Version 0.1.2.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
- Excluded windows are read before and after each monitor or whole-screen capture; when they
differ, the capture is thrown away and taken again (at most 3 times, then refused).
- A window whose program cannot be told counts as excluded: not listed, not picked, painted black.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
- screenView tool (screen_view.spec.md): whole screen, monitor by number, window by part of its
title or program, or the list of monitors and windows; fast, normal or detailed pictures.
- The app captures with Windows Graphics Capture (app/src/screen.rs, admin.spec.md "Screen
viewing"), decides for the asker (only people registered for PC control), never picks, lists
or shows excluded programs (pcControl.screenExclude, password managers by default), refuses
minimised windows and black pictures, and records screen.view in the activity log.
- Pictures ride on the tool message (ToolExecutionResult.images); llm.py sends them to
OpenAI-compatible servers as image parts after the tool messages. They stay in that reply only:
not in the tool carryover, logs or prompt dumps.
- The LLM reads pictures only when screen viewing is available (--vision gpu normally, cpu in a
game), measured apart for the placement; otherwise screenView is a disabled tool.
- Worker: screen_action / screen_action_result, 4 MiB messages.
- Eval evals/test_screen_view_tool.py; Windows check tests/screen_windows.rs.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
On the shared screen a tab brought to the front makes Chrome bring its window forward and take the
keyboard focus from the game or program the user is in (seen on the game PC). The botBrowser tool
now opens each person's page in a window of its own, behind and at the place and size of the bot
browser's window (Target.createTarget with newWindow and background), and closes it again if the
connection to it fails. While the browser runs on the virtual monitor, the app looks for new Chrome
windows every 200 ms and settles each like the first (tool window, fitted to the work area, without
activating it), the newest in front of the older ones, so the shared screen shows the page the bot
opened last.
Checked on the game PC: the bot's window loading a public page comes in front of the first, both
out of the taskbar and Alt+Tab, the focus staying in Discord also with the user idle for hours.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Chrome 157 no longer reads --disable-features=Translate, so on the virtual monitor every page in
another language got a translate bubble over it on the shared screen (seen on the game PC). The
profile's preferences now turn translation off before each start, keeping the other preferences.
The live screen test also checks that opening the window leaves the focus where it was.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The app writes the reason in UTF-8; FileRead took it for the system's ANSI
code page, so a declined driver install showed mojibake in the message box
and in the details. Read the file's bytes and convert them from UTF-8, for the
install and the uninstall message alike.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The driver component is on by default, so installing Jarvis installs the
driver too (Windows asks the SignPath question once). A silent install, which
is how updates run, never touches the driver: no question during an update and
no driver brought back after the person removed it.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Turning a daily target on only marks its current items as seen, so that
day's digest still goes out at its hour. Each round saves its state before
posting and posts nothing when the save fails, so the same items are not
posted again every minute.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Each collection target can announce new items at once or in a daily digest
(off by default) in collect.notifyChannel, as titles and links only, with no
LLM; each post is recorded in the activity log. The brain gets a read-only
collectedNews tool (every target's newest items by default, fenced as
untrusted web content), offered with the app's collection file and the
collection feature, and available to scheduled tasks.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
A personalNotes tool (add, list, delete the asker's own notes) offered only
with a notes backend. The app keeps the notes in config/notes.json and sends
the asker's with every voice, chat and task reply; the reply engine puts them
in a <person_notes> fence as that person's preferences, which never permit an
action or a tool. The worker's schedule bridge becomes a shared app bridge
for schedule_action and note_action. A 사람별 지침 screen edits and deletes
any note; features.notes turns the tool and the notes off.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
A chat action's channel, message and asker are recorded only when they are
ids (else '(잘못된 ID)'), an unknown action as chat.other, and a failure as a
fixed note, so text the model put in those fields never reaches the log. Old
day files are also deleted on the first record of each new day.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Each chat channel action the brain's tool asked for and each schedule
delivery is recorded in activity/<day>.jsonl by kind, target, who asked and
outcome, never its text, kept 30 days. The worker's chat_action carries the
asker. A 한 일 screen lists the newest entries.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
permissions::decide gives allow, ask or refuse for an action on the PC: only
people in pcControl.allowedUsers may ask (nobody by default), viewing is
automatic and clicking, typing and running ask first by default, and
payments, passwords, deleting files and administrator rights are never done.
A PC 제어 screen edits the levels and shows the chat channels' send and
delete switches beside them. The PC control tool (M7) will use it.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
- A repeating schedule the app was off for is judged by its latest time up to
now, so today's reminder is not skipped after a night off.
- A spoken reminder counts as delivered only when its speech ends played to
the end; speech that failed or was stopped goes to the chat channel.
- A schedule being delivered is marked firing; one still marked at the next
start gets a failed outcome instead of vanishing without a reason.
- Deliveries run in the loop's JoinSet, so stopping the services stops them.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The brain gets a schedule tool (add, list, delete the asker's own reminders,
timers and tasks) offered only with a schedule backend. The voice worker asks
the app through schedule_action and runs fired tasks as reply via task with
read-only tools, no MCP and a fresh short-term memory. The app keeps the
schedules in config/schedules.json, fires them every 5 s, speaks them when the
person is in Jarvis's voice channel and otherwise posts in the chat channel
they asked in (or schedule.chatChannel for voice). A feature switch and a
예약 screen list and delete everyone's schedules.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
- update.rs finds the newest app-v<version> release above the running
version (v1 tags, drafts, prereleases, incomplete releases and files on
another server are skipped), downloads Jarvis-Setup-<version>.exe into
cache/updates checked against its .sha256, and hands it over.
- update_install hashes the file again, stops what runs, starts the
installer with /S /RUN as administrator and the window quits; a declined
administrator prompt keeps the app with the services stopped.
- New 업데이트 screen; the app checks once when the main screens open.
- tauri.conf.json no longer sets its own version, so the program file
carries the Cargo.toml version the update compares.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
- The lock is closed in a last hidden section once every change is done,
so /S /RUN and the finish page's "Jarvis 실행" start the app instead of
its "installing" message.
- An installer or uninstaller that finds the lock taken (or cannot make
it) stops at once, silently with exit code 5, so two never run together.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
- installer.sh lists every packaged file for the uninstaller, and folders
are removed only when empty, so a chosen folder keeps other files.
- A virtual monitor driver removal that fails stops the uninstall with
everything kept (silent exit code 4); the elevated half exits 2 when
there is no driver of ours to remove, which the uninstaller accepts.
- WebView2 is installed before any file is copied; when it is still
missing the installer asks to retry, or silently stops with code 3
without installing or starting anything.
- The installer holds Global\JarvisBotSetup while it runs; jarvis.exe
started meanwhile shows a message and exits before any window.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
- app/shell/installer.nsi with installer.sh: package.sh's folder into Jarvis-Setup-<version>.exe
(NSIS, about 10 MB) and its .sha256; models and runtimes still come with the first-run wizard
- Program Files with one administrator prompt, Korean pages, components (start menu and desktop
shortcuts, the virtual monitor's driver off by default, installed through the app's elevated
half), the WebView2 runtime only when missing, a wait for a running Jarvis, /S and /S /RUN for
updates, the app started through Explorer so it never runs as administrator
- uninstall removes this app's driver, only the files it installed, the shortcuts and its
registry keys; the data folder stays
- checked on the win-ci VM: silent install, start, update while running and start again,
uninstall; and an install with the driver from the pages, whose uninstall leaves no device,
driver package or registry key
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
- a normal placement keeps its reading only when no game runs: the perf loop passes whether a
game runs to every switch, so the way back from a failed game stage while the game runs keeps
the reading from before the game
- game detection also looks at every running program once as the services start: a program in
a launcher's game folder or the person's games list counts, also behind other windows
- checked on the win-ci VM: a minimised program in a Steam game folder is found at the start
while nothing is in front
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
- the game stage reads the card again: other programs as at the last normal placement made with
no game running, and a reserve of at least what grew since (the game and programs opened after
that reading); with a game already running at the start (game detection now looks once before
the LLM starts) it counts what the card holds now and leaves only the room for other programs,
so it no longer pauses for counting the game twice
- the retry count starts over whenever another stage is wanted, also when the game ends with the
normal stage already running, so the next game tries again
- the wait before the next try counts from when the failed switch and its way back are over
- the worker client keeps, sends and readies a connection under one lock (llm, then state), so an
LLM state set while a worker connects is not lost
- a game watch dropped by a start that fails part way stops its thread
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
- the normal stage reads other programs' memory afresh (the card after the LLM stopped, less the
STT worker's model) and keeps it; the game stage counts from the latest normal reading, so the
game is not counted against its own reserve
- a restart that fails brings back the stage that ran before and tells the brain ready; the
failed stage is tried again after 60 s, three times at most; when nothing comes back the brain
is told paused so requests fail at once
- the worker client keeps the latest llm state and sends it first on every connection
- switches are numbered: llm switching carries an id, llm_idle names it, and a late idle of an
earlier switch no longer lets a new one through
- a request that waited 180 s for a switch fails without calling the brain
- the dashboard shows the stage that runs also on a fallback placement without experts
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>