db21abdef00de03dcee54fd7222b2ab3205c300e
6
Commits
| Author | SHA1 | Message | Date | |
|---|---|---|---|---|
 David Heinemeier HanssonandClaude Opus 5.5
|
c05d90196f |
Switch between several Claude and Codex subscriptions, and build apps the Omarchy way (#13770)
* Plan multiple Claude and Codex accounts with manual or automatic switching Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * Keep several Claude and Codex accounts and start new sessions as the active one Each added account gets its own home holding only its login, with history, settings and skills linked back to ~/.claude or ~/.codex so --continue works across a switch. Accounts are added through the CLI's own login, and cx, cy, plain claude/codex and omarchy-agent all start as the active account unless CLAUDE_CONFIG_DIR or CODEX_HOME is already set. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * Report limits for every Claude and Codex account in the usage records Each registered account is probed with its own sign-in and cached on its own, and a parked account whose sign-in lapsed keeps its last-known numbers marked stale. The record's top-level limits keep describing the active account. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * Switch accounts automatically near a limit, or notify with a one-click switch After each usage update, an active account at or over its provider's threshold (95% by default) moves new sessions to the account with the most headroom in auto mode, or offers that switch as a notification in manual mode. It never flaps back to an account that just reset, and says once when every account is over. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * Show every subscription account's limits in the agents panel With several Claude or Codex accounts, the limits become one card per account with the active one badged. Pick a card with its number and press Enter (or click Use) to move new sessions to it, press a to add an account, and m to toggle automatic switching. Limits refresh every minute while an active account is above 80%. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * Add Setup > Agent Accounts to list, switch and add Claude and Codex accounts Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * Document switching between several Claude and Codex subscriptions Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * Count a rested parked account as available, and pin Main to its own home A parked Claude account whose windows all reset now reads as 0% instead of unknown, so switching can pick it. Main's Codex limits come from ~/.codex even when CODEX_HOME is set, and duplicate logins are checked against who each home is signed in as now. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * Key limits caches by subscription and report when an account truly frees up An added account's limits cache follows its account id, so a new account reusing a removed one's label never inherits its allowance. The all-accounts notice now names when an account's blocking windows have all reset, not the earliest reset of any window. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * Keep account ids clear of routing keywords and refresh the panel after removal Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * Share Codex plugins and hooks across accounts, and key Claude caches by current sign-in Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * Drop the key hint from the Add account button Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * Sign new agent accounts in through a private browser window The main browser is almost certainly signed in to the account you already have, and the login would silently reuse it. Both CLIs open their login page through $BROWSER, so it now points at omarchy-launch-browser --private. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * Let agent account commands default to your default agent's provider With Claude or Codex as the default agent, the provider can be left out: omarchy agent account use work, and primary names the primary account. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * Drop the primary alias, which shadowed an account named Primary Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * Replace the auto switch button with a small Notify / Autoswitch toggle Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * Add accounts from a small + beside the switch toggle Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * Fix the punctuation of the switch toggle's README line Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * Drop the border around the add account + Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * Rename Claude and Codex accounts without moving their homes Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * Lay out agent accounts on accent rails instead of boxed cards The active account gets an accent rail and the others a quiet one, each window is a single compact line, and the switch toggle, add, and Use are text. Clicking an account's name renames it in place. A window without a reset time no longer leaves an empty line, and last-known numbers are only red when the sign-in needs attention. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * Read Codex limits without waiting on account/read Codex 0.158's app-server can leave account/read unanswered, and asking it first lost the limits whenever it did, leaving the agents panel showing "Codex limits unavailable". The limits name the plan themselves, so they're asked first and account/read is only a short fallback when they don't. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * Switch agent providers from their marks in the panel header The row of provider buttons gives way to a small mark per provider in the hero's corner, the selected one at full strength, with the add account + beside them in place of the + by the switch toggle and the full-width Add account button. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * Keep the agents panel's scrollbar off its contents The panel's content narrows to leave the scrollbar its own strip whenever it scrolls, and the add account + leads the provider marks instead of sitting at the very edge. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * Add Claude, Codex, or Grok subscriptions from one Add Account menu The first account for a provider installs its CLI if needed and signs in to the CLI's own home through the normal browser; only additional accounts get a home of their own and a private window. Grok signs in its first account. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * Reveal Autoswitch beside Use instead of a Notify / Autoswitch row The panel's + now opens the Add Account menu for any provider. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * Split agent accounts with plain separators instead of rails ACTIVE already marks the account new sessions use. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * Only call Claude limits stale once they're old, and poll near a limit less Anthropic rate-limits its usage endpoint, and polling two accounts every minute near a limit got every re-check refused, so both accounts read "Last known" with numbers a minute old. A refused check of numbers under 15 minutes old now counts as current, older ones say how old they are, and near-limit polling is every three minutes. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * Show every agent's limits on one page under an Agents hero The panel stops switching between one provider at a time and lists every agent and account with its limits, dropping the tokens-by-day and by-model charts. The hero carries the agents robot and rotates through what the token counts add up to: tokens this week and today, the most used model, the busiest day, and today's prompts and sessions. Middle-clicking the bar icon refreshes, since there's no provider left to advance to. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * Report Codex's free rate-limit resets in its usage record Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * Give the agents panel more room, and show Codex's free resets Limit lines are a notch larger with more space between them, each account's limits sit a clear step below its name, and sections breathe. Codex has one limit on Pro, so its section now also says how many free full resets are waiting and when the next one lapses. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * Mark Claude's Fable limit on the Weekly meter instead of its own row A model-scoped allowance on the same clock as a base window is drawn as a tick on that window's meter and named in the row's tooltip. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * Let omarchy-default-agent set the default without launching it Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * Let the agents panel drive adding an account without a terminal --check says whether adding one now would be each provider's first sign-in or an additional account, and --events reports progress as tagged lines alongside the CLI's own output, notifies with the result, and cleans up the login and its scratch home when cancelled. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * Add subscriptions and pick the default agent right in the agents panel The + is a proper accent button, and it swaps the list for a picker of Claude, Codex, and Grok that signs in without a terminal: name a further account, then follow the sign-in with its status, Grok's confirmation code, a field for Claude's pasted code, and a link to reopen the page. A dropdown at the bottom sets the default agent without launching it. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * Draw the Fable tick in its meter's own color Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * Revert "Let omarchy-default-agent set the default without launching it" This reverts commit da96261f7d26bb76774cbdc1cb9e0abde92bb841. * Make the first agent signed in the default when none is picked Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * Keep the agents panel in the bar, with setup, starters, and adding in it A machine with no agent opens the panel on setting one up. Once set up, the list ends with starter prompts for a new theme, plugin, or app, and a quiet Add a subscription. The panel no longer sets the default agent, and the hero drops its add button. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * Offer the starter prompts as tiles, and adding as a row beneath them Theme, Plugin, and App each get a tile with its glyph, and Add a subscription a matching row with the + in a tinted square. The panel is allowed to grow tall enough to show it all without scrolling. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * Put the tinted add button in the agents panel's hero corner Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * Show when a limit resets in small type under its meter The percentage keeps the right edge to itself, so the meter runs longer. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * Revert "Show when a limit resets in small type under its meter" This reverts commit 108d9de59da1af12cddbf232b0b5333ad8994c64. * Put each account's plan beside its name, and its email in a tooltip Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * Stack each limit's percentage over its time left, so the meter runs longer Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * Show only each limit's time left, with the exact percentage on hover The meter already says how full a window is. Dropping the percentage beside it leaves one calm figure per row, and the row's tooltip carries the number. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * Sign an existing account in again with omarchy-agent-account-add --reauth It signs in where the account already lives: the CLI's own home (:primary) through the normal browser, an added one through a private window. The usage records now say which account is the primary. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * Offer Sign-in required in the agents panel instead of how old the numbers are A lapsed sign-in shows as a link that signs that account in again right in the panel. The "as of" and "last known" notes are gone; trouble that isn't about signing in still shows as text. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * Keep the separator dot out of the Sign-in required link Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * Drop the Setup > Agent Accounts menu and the stale plan The agents panel now lists, switches, and adds accounts, so the menu's duplicate of it goes, and the plan written before the design settled no longer describes it. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * Install the compiler and Qt pieces for building Omarchy-style apps base-devel plus qt6-base, qt6-declarative, qt6-multimedia, and qt6-wayland, which is what Hype, Monologue, and Omacut build with through qmake6 and make. Qt was only there as a dependency of those apps, and the compiler not at all. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * Add an omarchy-app agent skill for building apps the Omarchy way It teaches how Hype, Monologue, and Omacut are built: C++ and Qt Quick in one flat project, qmake6 and make into a single binary, a theme that follows Omarchy's accent live, portal dialogs, keyboard-first conventions, Qt Test offscreen, and a PKGBUILD that puts the app in the launcher, with starter files that build and pass their tests as written. The agents panel's App starter uses it. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * Link the omarchy-app skill on existing installs Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * Let Cancel stop a waiting login, and list a lone account's limits Bash holds a trap until the foreground command finishes, so a login waiting on the browser outlived Cancel. It now runs behind an interruptible wait. With one account the usage record keeps its limits at the top level, which omarchy agent account list now reads for the primary. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * Re-read identity even when signed out, and count Grok as set up A home whose login is gone no longer keeps the identity the registry saved, so it reads as signed out and can be added again. The agents panel leaves its setup screen once any agent is signed in, not only once one has usage to show, since Grok has no usage collector. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * Save each account's current identity before collecting usage A home signed in to someone else since it was added kept the old email in the registry, which the usage records name accounts by. The usage update now refreshes the registry from each home first. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * Offer Claude's paste field from the start, and keep providers with accounts Claude's login prompts for a pasted code without a newline, so the panel's line reader never saw it; the field is simply there for Claude sign-ins. A provider with accounts stays listed even when the active one's limits are unavailable, so its other accounts can still be switched to. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * Never autoswitch to a signed-out account, and report Codex's missing sign-in An account nobody is signed in to is left out of switching, however much room its cached numbers show. Codex answers a home without a login with an error, which now reads as Waiting for auth, so the panel offers Sign-in required for Codex accounts too. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * Link shared files before they exist, and only call every account over when it is A shared file the primary home didn't have yet was left unlinked, so an added account made its own copy and the two diverged for good; it's linked up front now, and written in the primary home by whichever account writes it first. The all-accounts-over notice waits until every account's limits are actually known. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * Make the agent starter tiles compact and call the section Make something cool Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * Hold the agents hero line until its next fade Every usage record that landed rebuilt the summary phrases, and the hero indexed that live list, so opening the panel could swap the line several times between fades. It now keeps what it shows until the timed swap. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * Start the default agent from the agents panel hero A console button beside the add button runs the same launcher as the right click: the default agent, or the picker when none is set. It hides while adding a subscription, where the add button becomes the way back. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * Put the add button before the agent launcher in the hero corner Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * Bring keyboard navigation back to the agents panel The one-page redesign dropped left and right with the per-agent pages, leaving the arrows only to scroll. They now walk everything that does something, row by row: the hero's add and launch buttons, each switchable account, and the starter tiles. Enter acts on the cursor, the cursor scrolls into view, and hovering moves the same cursor so only one thing is lit. Number keys still jump to an account. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * Let adding an account take over the agents hero While adding, the hero's line reads "Add an account" in place of the rotating summary, and the X in its corner is the only way back, so the add view drops its own title and Back link. Each agent to add is just its mark and name; one that can't be added is dimmed and says why on hover. The arrows walk that list too, and Enter picks one. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * Show the agents to add as large marks, three across Each agent is a large mark over its name with no box around it, in one row the arrows move along. The chosen one turns accent and grows a touch. The reason an agent can't be added is shortened to fit inside the panel. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * Move between Autoswitch and Use on an account with the arrows An account that isn't active is now two stops, Autoswitch then Use, so left and right move between them and Enter acts on the one that's lit. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * Land on Use when moving up or down onto an account Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * Hide Use on an autoswitching account until you're on its line With Autoswitch on, the line shows only Autoswitch. Use appears when the line is hovered anywhere or the keyboard is on it. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * Start adding an account with the first agent focused Opening the add screen puts the keyboard cursor on the first agent, so Enter picks it straight away. A focused agent lights up even before the check says whether it can be added. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * Drop the Sign in link under the account name field Enter in the field already starts the sign-in. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * Key the primary Claude account's limits cache by its subscription The primary home always used claude-limits.json, so signing it in to another subscription could carry the old one's numbers over when the first probe failed, and autoswitch would act on them. Once the home says who it's signed in as, its cache is keyed by that like every other account's. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * Leave Qt Multimedia out of the base packages again Quattro dropped it once the shell no longer needed it, and the app skill already has an app that plays audio or video add it and list it in its own depends. The compiler and the rest of Qt stay. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * Render the agents panel's dynamic text as plain text Sign-in status, help text, and provider names come from outside the shell, so none of them should be read as markup. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * Keep an account's plan and trouble clear of its Use and Autoswitch links The details beside an account's name grew to their full width, so a long plan or warning could run under the links on the right. They now shorten with an ellipsis instead, and Sign-in required keeps its whole width. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * Let primary name the first account, as the manual says `omarchy agent account rename primary Hey` was documented but failed, since the primary account's id is main and lookups took exact ids only. primary now reaches the account marked primary, whatever it's been renamed to, and no new account can take primary as its id. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * Install everything an Omarchy app needs by default The app skill builds with Qt Multimedia, SVG icons, and ffmpeg as well as the compiler and the rest of Qt, so qt6-multimedia, qt6-svg, and ffmpeg join the base packages and the migration. The shell still plays no video through Qt Multimedia, which is what its test now checks. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * Dim agent limits kept from an earlier check, with their age on hover When a Claude probe fails, the last numbers carry on and looked just like fresh ones. Those meters now dim, and their tooltip says how old they are. The record carries limitsStale and limitsFetchedAt for this. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * Give the app skill templates for every file its build needs The templates named src/backend.{h,cpp} and a test file without showing them, so each app had to invent its own. They're now templated, with a starter icon, a note that LICENSE and the icon must exist for package(), and qt6-wayland in the PKGBUILD's depends. Scaffolded from the templates alone, the app builds and its tests pass. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * Start the agents panel cursor over when the agent list comes or goes Right after the shell starts, the panel can briefly look like a first setup and focus the first agent to add. When the records land the rows change under the cursor, which then lit an account nobody had picked. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * Drop dead hover and limit checks from the agents panel The tiles and hero buttons fell back on their own hover when there were no keyboard rows, but the hero always has one, so hover only ever moves the cursor. Stale meters only exist when there are limits, so neither the panel nor the record needs to check for some. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * Drop the a hotkey for adding an agent subscription The + in the hero is the way in, by mouse or by arrowing to it. The plugin README also catches up with the panel: the launcher, dimmed stale limits, the new add screen, and the arrows walking a cursor rather than scrolling. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * List a command family that sits under a flags-only command `omarchy agent account` reached omarchy-agent, which takes only flags, and failed on the word instead of listing the account commands. When the command matched so far takes only flags and the next word names visible commands, the router now lists them. `omarchy update aur` likewise lists the update aur commands rather than running a full update. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * Refuse to remove an agent account a running session uses Removing an account deleted its home at once, pulling the login out from under any session started in it, though running sessions are never meant to be touched. It now says to quit that session first. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * Prefer an account checked just now when switching near a limit Autoswitch weighed numbers kept from an earlier check like fresh ones. A parked account's sign-in lapses within hours, so a stale one stays a candidate, but an account checked just now wins over it. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * Leave a user's own omarchy-app skill in place when linking ours Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * Reach Sign-in required with the keyboard in the agents panel Both the link on a lapsed account and the one on a single-account agent are now stops in the cursor's walk, and Enter signs in again. A picked link scrolls into view. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * Bring the manual's agent accounts paragraph up to date Near-limit checks run every three minutes, not every minute, and Autoswitch now appears on hovering an account's line and takes Use's place while it's on. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * Keep the agents panel cursor on things you can act on The active account was a stop with nothing to do, so the cursor seemed to vanish there. It now only stops on it to sign in again. The hero's buttons also show the cursor plainly, with an accent border and a deeper tint instead of a shade's difference. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * Never leave an added agent account's login outside the registry Registering moved the login into its home before carrying settings over and saving the registry, so a failure in either stranded a home holding a sign-in that the account commands couldn't see. Settings are now carried while the login is still pending, and a failed save moves the home back there for the add command to clean up. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * Put room before freshness when picking an account to switch to Preferring accounts checked just now outright could pick a fresh one at 94% over a stale one at 10%. An account within 15 points of the threshold now counts as near its limit however fresh, so accounts with room come first and freshness only decides among them. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * Check agent limits more often relative to the switch threshold Faster checks started at a fixed 80%, so a threshold set lower could be crossed and wait out the 15-minute interval. They now start 15 points below the threshold, which is still 80% at the default 95%. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> --------- Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com> |
||
|
David Heinemeier HanssonandClaude Opus 5.5
|
b421b1b479 |
Read Codex limits without waiting on account/read (#13733)
Codex 0.158's app-server can leave account/read unanswered, and asking it first lost the limits whenever it did, leaving the agents panel showing "Codex limits unavailable". The limits name the plan themselves, so they're asked first and account/read is only a short fallback when they don't. Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com> |
||
|
|
4cd8a081cb |
Fix Codex usage collection on 0.149 (#7649)
* Fix Codex usage collector approval policy * Capture codex argv with boundaries in the scanner test The stub joined its arguments with "$*", so the assertion compared one flattened string and could not tell five arguments from fewer containing spaces. Passing "-s read-only" and "-a on-request" as single arguments -- which codex rejects as an unexpected argument -- passed the test. NUL separation and an array comparison keep the boundaries the assertion is about. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> Co-authored-by: Codex XHigh <noreply@openai.com> --------- Co-authored-by: Omabot <omabot@omarchy.org> Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com> Co-authored-by: Codex XHigh <noreply@openai.com> |
||
|
|
25c2b3aed2 |
perf(agents): cut codex usage collector memory with SQL filter and cache (#6780)
* perf(agents): cut codex usage collector memory with SQL filter and cache The codex collector scanned every row of opencode.db (1.7 GB, 55k+ rows) with Python-side json.loads, peaking around 716 MB of RSS on every run -- including the panel's refreshLimits() call, which passed --limits-only that the collector silently ignored. Filter rows in SQL (LIKE gates + json_valid + json_extract authority, mirroring the old Python filter semantics) so giant blobs are never parsed, and cache the local stats scan in XDG_CACHE_HOME following the claude collector's pattern (atomic writes, flock, schemaVersion). --force rescans, --limits-only and normal mode reuse a fresh cache and fall back to a full scan when it is missing, stale, or corrupt. Measured: cold scan 716 MB -> 158 MB peak; warm --limits-only ~85 MB and ~1.4 s. Output record schema and values are unchanged for the same data (parity verified against the old filter, including malformed rows). * Scope the codex scan cache's 15-minute reuse to --limits-only A no-flag run is the widget's periodic refresh, and refreshIntervalSec is configurable down to 30 seconds; holding every mode to a 15-minute cache meant stats could lag far behind the interval the user asked for. Mirror the claude collector: normal runs reuse a scan for ~20 seconds purely to dedup concurrent collectors, and only --limits-only, which promises just fresh limits, may reuse a scan for up to 15 minutes. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * Invalidate the codex scan cache across day boundaries The cached stats embed date-dependent fields (todayPrompts, todayTotalTokens, recentDays), but only the file's age was checked, so a cache written at 23:58 served yesterday's numbers as "today" for up to 15 minutes past midnight. Stamp the envelope with the scan's local date and treat any other date as a miss. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * Reject codex scan caches with a future mtime A cache whose mtime is ahead of the clock has a negative age, which the freshness check accepted forever: setting the clock backwards froze the stats until real time caught up with the file. Require a non-negative age before trusting the cache. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * Never cache an interrupted opencode scan A transient lock, schema migration, or corrupted database aborts the opencode scan mid-flight; the partial numbers still serve the current run, but persisting them let a single bad read suppress opencode usage for every cache reader until expiry. The claude collector already skips its opencode cache write on a database error; do the same here. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * Make the json_valid guard order explicit in the opencode query The query relied on json_valid(data) evaluating before json_extract(), but SQLite does not promise that AND terms run left to right; a reordered plan would let json_extract raise on a malformed row and silently truncate the scan. Wrap each json_extract in a CASE so the guard is structural rather than positional. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * Drop a claude-collector comment that is false for codex "These caches were world-readable before" was copied from the claude collector; codex had no caches before this one existed. Explain the chmod on its own terms. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> --------- Co-authored-by: markbusking <marcosbustos.dev@gmail.com> Co-authored-by: David Heinemeier Hansson <david@hey.com> Co-authored-by: Claude Fable 5 <noreply@anthropic.com> |
||
|
David Heinemeier HanssonandClaude Fable 5
|
77cf58ccfe |
Add Fireworks balance usage panel (#6488)
* Add a Fireworks balance collector and teach the agents panel prepaid ledgers The omarchy-agent-usage-fireworks collector reads serverless token usage from the Fireworks billing API, grouped by day and model for the last 30 days, and reshapes it into the shared record contract. Fireworks does not expose its prepaid ledger through the documented API, so the record carries an estimated balance instead of rate limits: credits configured in ~/.config/omarchy/agents/fireworks.json minus rated account costs since the funding date. Credentials come from FIREWORKS_API_KEY/FIREWORKS_ACCOUNT_ID, the auth.ini that firectl set-api-key writes, or — last, so an explicit login wins — the key opencode stores for its fireworks-ai provider. The panel gains two generic capabilities any agent record can use: a balance object draws a BALANCE section — remaining credit, a fuel-gauge meter that drains toward empty and lights the bar alarm below 10%, and funded-versus-spent detail — and hasPromptStats: false keeps prompt and session counts out of today's tooltip for agents whose billing API only ever reports tokens, on this machine and through synced snapshots. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * Feed Claude and Codex usage from pi, omp, and opencode sessions A subscription burned entirely through another coding agent leaves no native Claude Code transcripts and no Codex session files, so the panel showed nothing for it. pi and omp write compatible JSONL sessions, and opencode records per-message provider, model, and token usage in its message database; the claude and codex collectors now scan all three — filtered to Anthropic and OpenAI providers respectively — and merge those numbers into their local stats. Fireworks stays out on purpose: its billing API already sees that traffic server-side, and a local scan would count the same tokens twice. The collector tests pin XDG_DATA_HOME so a developer's real opencode history cannot leak into fixture runs. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> --------- Co-authored-by: Claude Fable 5 <noreply@anthropic.com> |
||
|
David Heinemeier HanssonandClaude Fable 5
|
bb8d2f2cb3 |
Split agent usage into data files and rename the plugin to omarchy.agents (#6603)
* Add agent usage collectors that write display-ready data files
One omarchy-agent-usage-scan-<agent> collector per AI coding agent prints a
complete display-ready usage record — identity, tier, status, rate limits,
and today/week/all-time stats. omarchy-agent-usage-update runs every
collector it finds and writes the records atomically to
~/.local/state/omarchy/agents/usage/, so anything that displays usage only
ever reads JSON from there.
The Claude collector absorbs what the shell previously did in-process:
transcript scanning, the stats-cache/history fallback, credentials parsing,
and the OAuth limits probe, now with a probe throttle and last-good limits
kept across network failures. The Codex collector is the existing scanner
reshaped to the shared record contract.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
* Redo the model-usage plugin as omarchy.agents watching usage data files
The panel is now strictly a display. It discovers the JSON records that
omarchy-agent-usage-update maintains under
~/.local/state/omarchy/agents/usage/, watches them for changes, and draws
whatever appears — so adding an agent means shipping a collector, never
touching the panel. Marks resolve by convention (assets/<id>.svg with an
optional -light twin), the limits meters read a generic limits array, and
the per-provider QML adapters and in-plugin scanner scripts are gone.
Cross-device sync aggregation stays in the shell and keeps the snapshot
field names older versions wrote, so mixed-version fleets still merge in
both directions.
With the provider fan-out gone, the widget takes its real name: the plugin
id becomes omarchy.agents. A migration renames it wherever a user's config
mentions it — layout entries keep their settings and position, a disabled
widget stays disabled — then primes the data files once and drops the old
scanner cache. The migration test also drops a stale assertion that expected
migrations to restart the shell themselves, which
|