Commit Graph
4952 Commits
Author SHA1 Message Date
jason.woltje a7362a8fc5 test(webui): tasks-board: the fixture task is due in 2099, so the card still reads due, not overdue, after 20 Nov 2026 passes; the expected label carries the year (#1873, lane chat-fail) 2026-11-19 18:20:43 -06:00
jason.woltje 5a07f9e1f9 test(bus): chat-fail pickup-launch: the fixture runs its Claude model on claude-code, since a legacy Claude entry counts as OAuth and runs there only; the host launcher's env carries a stand-in session login, because claude-code's preflight needs one and this launcher boots with no proxy, and a stand-in adapter path skips only the binary check, so each launch still stops at the model gate with model-provider-unavailable and no assertion changed (#1873, lane chat-fail) 2026-11-19 17:08:06 -06:00
jason.woltje 07eca4a670 fix(chat-fail): review round 3: the link, email and mixed-run redaction patterns start only where a run without spaces starts, so redacting the whole kept stdout stays linear (a 128000-character run took 10 to 15 s, now a few ms) and matches the same text; a long text that redacts below the read limit is not cut and keeps its first word, and a cause that is only [redacted] is dropped; the proxy page reads relogin, plural and past forms of an expired sign-in as a sign-in (#1873, lane chat-fail) 2026-11-19 16:36:16 -06:00
jason.woltje fc9e4dfca8 fix(chat-fail): review round 2: the F1 cause is redacted over the whole text before the tail cut, so a cut that lands between a key word and its value no longer shows the value; the account sign-in words in the proxy page match whole words only ("assigning" and "logging" no longer read as a sign-in) and a prefixed or "is expired" token counts (#1873, lane chat-fail) 2026-11-19 16:36:16 -06:00
jason.woltje 10a35ce65a fix(chat-fail): review round 1: the F1 cause decodes %XX, HTML references and NFKC forms before the secret and redaction checks, redacts the value after a quoted or space-separated key word and Windows and UNC paths, and is read from the tail of the output, with the cause on the retry log line too; a launch-failed start posts only when the same message is refused twice, project-pending and processing-halted are waits, and a refused post is tried again and logged once; the proxy reason refuses an impossible date, checks quota_blocked before the message words and no longer reads a bare "expired" as a sign-in (#1873, lane chat-fail) 2026-11-19 16:36:16 -06:00
jason.woltje 3b9cac5452 fix(bus): chat-fail F3: an account the proxy reports as unavailable or in error carries one plain reason with the next step, built only from the fields of the existing account read (requires_sign_in, refresh_error, quota_blocked, next_retry_after, and status_message matched against fixed patterns and never shown); with nothing more it says the model service gave no reason. No new proxy call; working, turned-off and unknown rows keep their exact fields (#1873, lane chat-fail) 2026-11-19 16:36:16 -06:00
jason.woltje bc6d3565b0 fix(bus): chat-fail F2: a refused thread start and a proxy-stall stop each put one plain line in the chat, with the cause and next step, once per unanswered message and not once per retry; a start that only waits (a free place, a run still ending, a held lock, a lost race) puts none. The line comes from the launcher through a trusted launch op, not a socket verb, from the thread's instance with from_run bus, so no schema change and no new refusal code (#1873, lane chat-fail) 2026-11-19 16:36:16 -06:00
jason.woltje 72ea7272ec fix(harness): chat-fail F1: the turn-failure code moves into the mounted harness failure-notice.mjs, so a sandboxed runner loads it; the separate file was outside RUNNER_FILES and every sandboxed run died with ERR_MODULE_NOT_FOUND (#1873, lane chat-fail) 2026-11-19 16:36:16 -06:00
jason.woltje 27a359590e fix(harness): chat-fail F1: a failed turn's chat line carries the harness's error as a plain cause and next step, cleaned, redacted and capped, with the technical reason under it (#1873, lane chat-fail) 2026-11-19 16:36:16 -06:00
jason.woltje 6c5ec23b47 test(model-proxy): settings-p3 ride-fence: the launch fixture runs its Claude model on claude-code, since a legacy Claude entry counts as OAuth and runs there only; a stand-in adapter path skips only the binary check, so the before launch still reaches the limit checks and the fenced launch after Ride all is still refused with proxy-uncertain (#1873, lane settings-p3) 2026-11-19 15:44:31 -06:00
jason.woltje dd0a8612c8 fix(webui, model-proxy): settings-p3 review P2-1, P3-1 to P3-3: the done line counts an organization that can still ride, so an owner who can't send is not told every organization can see the models; an organization that took cliproxy during a Ride all run reads as already using it, not refused; the stopped page's refusal repaint is tested before the page renders again; a stand-in organization with no bus status (the single-business Console) reads that moving is not available there, with no Move button, from the status summary as well as after a run (#1873, lane settings-p3) 2026-11-19 15:16:59 -06:00
jason.woltje 9276277970 test(webui): settings-p3 shots: the Model setup shot shows the Ride all box below the restart list, with one organization that can ride, one skipped and one on stand-in models; a full-page shot grows the window to the page so the fixed footer covers nothing (#1873, lane settings-p3) 2026-11-19 15:04:56 -06:00
jason.woltje 517a3fe467 fix(webui): settings-p3 P3-1 to P3-11, rebase re-check 1 to 4, Routing help: Model setup says what to do for a held init lock or a partial init, with the README path under Technical details, and both lines are tested; the triage-drawer fixture case says its pair is synthetic; a model-key collision names a step; a fixture refusal leaves Restart on the stopped page but not primary; a ready catalogue the bus picked up late keeps its organization on Model setup's restart list with a Restart, holds the next step, and gets a Restart beside its notice on Routing; the README names the Model setup page; the NO_GUESS pin counts both uses; a reason that is not a stop reads "Last check:"; the why line sits under the headline; orgDownReason tells every waiting caller, each with its own name; the done line does not say every organization while the Ride all box lists one; Ride all is the next step only when the owner can send; the step button reads "Go to Ride all"; Ride all text says model service; the single-business fixture move says nothing about a bus it can't see; the Routing help names no harness (#1873, lane settings-p3) 2026-11-19 15:04:56 -06:00
jason.woltje 149712816b fix(model-proxy): settings-p3 R1, R2: Ride all reads each organization's file and change log again before its add, not only the receipts, so an add and remove saved on its Settings page during a run is not put back; a list that can't be read again, a change log that can't be read or an organization no longer listed skips the add as unreadable; with no bus status (the single-business Console) a fixture move answers not-available instead of bus-running, and whether its bus runs is unknown (#1873, lane settings-p3) 2026-11-19 15:04:56 -06:00
jason.woltje 4f63176676 test(triage): review round 5: a host bundle file replaced with a FIFO is refused before the host reads it (#1873, lane harness-matrix) 2026-11-19 14:43:26 -06:00
jason.woltje 1e0aba8b37 fix(triage): review round 4: the host refuses a pi command value (a string that starts with "!") in any file the module wrote, checks each host bundle file after env() whether or not the scan reaches it, and records ctime for the rest of the run directory (#1873, lane harness-matrix) 2026-11-19 14:43:26 -06:00
jason.woltje 680bb90af3 fix(triage): review round 3: the host checks every URL in a file the module wrote under bundle/, home/, work/ or tmp/, as written or as a URL parser reads it; refuses a backslash in a file JSON.parse refuses; refuses a change to its own bundle files or the rest of the run directory; and a refused start removes work/ and tmp/ too (#1873, lane harness-matrix) 2026-11-19 14:43:26 -06:00
jason.woltje 6a24cf743d test(triage): a declared harness file with a second hard link is refused (#1873, lane harness-matrix) 2026-11-19 14:43:26 -06:00
jason.woltje 06ee172a9d fix(triage): review round 2: the host reads every file under the run's bundle/, home/, work/ and tmp/ after bundle() and env(), refuses a link or a key there, and reads a JSON file decoded too, so an escaped key or URL is seen as the harness sees it (#1873, lane harness-matrix) 2026-11-19 14:43:26 -06:00
jason.woltje 66676804eb test(triage): write() refuses a bundle file that holds the key before the file exists, so the key never rests on disk until the scan (#1873, lane harness-matrix) 2026-11-19 14:43:26 -06:00
jason.woltje a3ad6578c4 fix(harness): review round: a legacy model entry (a provider or route name, no module) carries no account data, so a Claude-named model on it is treated as OAuth and held to claude-code; the Triage host refuses a key in any bundle/ file and any URL but the proxy's in a harness file; the models.json comment says what route() can and cannot see (#1873, lane harness-matrix) 2026-11-19 14:43:26 -06:00
jason.woltje 77a951cfe7 fix(harness): the harness rule keys on the account kind: a Claude model from an API key runs on claude-code and pi; an OAuth Claude model, or a group model whose name an OAuth Claude account offers, runs on claude-code only (#1873, lane harness-matrix)
Jason 2026-11-19: Claude models through an API key may run on any harness; OAuth Claude models only
on claude-code. The name check is gone. The catalogue reads every Claude OAuth account's model names,
whatever the account's state, and holds a group model with one of those names to claude-code, since
the proxy routes by name across every account. route() has no account data, so it holds upstream
claude only; admission and Triage hold the rest through the catalogue.
2026-11-19 14:43:26 -06:00
jason.woltje cd58ac23cd feat(harness): an OAuth Claude model runs on claude-code only; Triage runs on claude-code or pi through the proxy (#1873, lane harness-matrix)
cliproxy holds a model to claude-code when its upstream is claude, when an OAuth Claude account
offers the same name, or when the name contains "claude"; every other proxy model runs on
claude-code and pi. The catalogue, the launch (sessionModel and the models.json list) and the
provider route apply the one rule and refuse with model-harness-incompatible, so a stored
Claude-on-pi route is refused, never rewritten. The other route's pi provider is mosaic-proxy-other.

Triage: the route asks the provider for the transport on the stored harness and holds the
harness's env() to one contract (the key, unchanged, under exactly one of ANTHROPIC_AUTH_TOKEN or
ANTHROPIC_API_KEY; the URL only as ANTHROPIC_BASE_URL), with no list of harness names. The host
gives the module the real proxy and transport, checks each file it writes under the session's HOME
(regular, one link, no key) and binds it read-only after the run directory, and refuses a route env
with no key name or two. PATH starts with node_modules/.bin, as a bus session's does.

Tests: claude-code runs a z.ai and a codex proxy model against a stand-in (real claude binary);
pi answers Triage through a stand-in under bwrap; every new fence has a test.
2026-11-19 14:43:26 -06:00
jason.woltje fbaae25d06 fix(webui): settings slice 1: no harness names written into the page: Settings > Triage shows each harness by the id its module is installed under, from the route read, and its help line says the list holds only the programs the model runs with; the drawer's harness refusal reads the route and names its harness by that id, or keeps a plain line without a name when the route can't be read (#1873, lane settings-s1) 2026-11-19 14:38:04 -06:00
jason.woltje c9d64a341e fix(webui): settings slice 1: the failed settings read and failedView share one copy of the line every failed read ends with, so the shipped-strings pin in shell.test finds it once again (#1873, lane settings-s1) 2026-11-19 14:04:14 -06:00
jason.woltje 4cbf3f07b0 fix(webui): settings slice 1 review round 2: a model-key-collision catalogue says two models share a short name and offers no Restart, so it no longer holds Model setup's next step; the dev fixture line says it can't be changed from Console yet instead of asking for a Routing change that does not remove the fixture; Model setup owner text says model service in the init lock, repair and fence steps, and the fence clear command sits under Technical details with where to run it; the Health list and the stopped-organization page show the reason the workspace read already carries; the stopped page links to Model setup by its name; the hire start refusals point at Settings > Routing for models (#1873, lane settings-s1) 2026-11-19 14:04:14 -06:00
jason.woltje 48a2ed675c fix(webui): settings slice 1 review round: Model setup reads each organization's own routes and offers Restart only where the bus names no cause or one a restart clears, so the dev fixture gets its plain line and a Routing link and only a restartable organization holds the next step; the bus's cause codes read as plain lines with its words under Technical details, and the lock cause gets a primary Restart; a set fence reads as a restart to finish; a failed status read offers Check again; Save routes steps back while Restart is the primary; the page is named Model setup and owner text says model service; Routing explains harness and preflight; the drawer's harness refusal asks for a model that runs with Claude Code; the Triage page says daily token limit and that Triage also needs its key (#1873, lane settings-s1) 2026-11-19 14:04:14 -06:00
jason.woltje e70b6bd684 feat(webui): settings slice 1: Routing says why an organization can't see your models: the model service is not set up or not running (a link to its page), the bus names its cause (that line and Check again, plus Restart only for the lock cause), or the service runs and the bus names none (Restart); Model proxy keeps a Restart per organization whose catalogue is not ready, rereads the whole status after a sign-in, works out the steps after accounts from accounts, organizations and the Triage route, and shows only the next step's button with the rest under Technical details; Triage refusals are plain lines with a link to the page that fixes them, and the harness refusal names Claude Code; a stopped organization keeps its refused restart on screen and keeps the instance pages reachable; at phone width a Routing model select takes the whole row (#1873, lane settings-s1) 2026-11-19 14:03:27 -06:00
jason.woltje 2db4f24fd8 fix(model-proxy): re-check 3: mosaic org update that takes cliproxy off an organization records the removal in the proxy's receipts first and saves nothing if it cannot; Ride all reads the receipts again before each add, so a removal saved during a run is honoured; the bus-running and unreadable skip lines say what to do (#1873, lane ride-default) 2026-11-19 13:13:05 -06:00
jason.woltje 9bfe101176 fix(model-proxy): round 3: a Console save that takes cliproxy off an organization records the removal in the proxy's receipts first, and Ride all skips as removed any organization with any sign it was or may have been on cliproxy; a receipt line that does not parse fails closed; each add records an intent before it starts and an unknown outcome stays recorded; the fixture move and listing read stopped from the bus status and its process id; receipt and change-log reads check owner, mode and no symlink (#1873, lane ride-default) 2026-11-19 13:13:05 -06:00
jason.woltje 869d259036 fix(model-proxy): review round: an organization the proxy once put cliproxy on and that lacks it now is skipped as removed, read from the proxy's own receipts; an organization riding now is already riding whatever its history; a repeated catalogue key keeps the usable row; the page paints a Ride all answer while a load waits, names org-forbidden in plain words and gives a neutral fallback (#1873, lane ride-default) 2026-11-19 13:13:05 -06:00
jason.woltje 8200ea238c feat(model-proxy): Ride all puts the owner's existing organizations on the instance model proxy, and a cliproxy preset starts a new one on the owner's models (#1873, lane ride-default)
A cliproxy organization preset offers a model from the owner's catalogue and says in plain words
why it can't be used: no proxy, no riding bus answered, or no catalogue model is available. The
organization service passes an optional model to a preset's overlay; other presets ignore it.

Ride all, an instance-owner action on the Model proxy page, adds cliproxy to each organization
through that organization's own plugins.set. It skips and names an organization whose change log
shows cliproxy removed, whose bus is down, or whose bus has no proxy client. Proxy admin status
names it as the next step. A fixture organization is listed as using stand-in models; one more
confirmed click moves a stopped one to the owner's models through the organization service's
existing whole-document update at the revision read.

A fenced proxy then refuses each riding organization's launches with proxy-uncertain, a direct
anthropic route included.
2026-11-19 13:13:05 -06:00
jason.woltje 196505e6ef fix(bus): review round: the restart notice shows only on a ready catalogue, since a fenced, colliding or down proxy still refuses after a restart (#1873, lane pickup-launch)
A bus booted with the proxy on a fenced instance reads the same unavailable catalogue, so the
notice sent the owner to a restart that stops running agents and starts none. The test proves
that and shows the notice again once the fence is gone. The disabled check goes away with it;
the disabled test now says the routes verb answers disabled itself.
2026-11-19 12:08:41 -06:00
jason.woltje f7c0e4f46c fix(bus): a bus that picked up the shared proxy says on its catalogue that a restart is needed to start agents on these models (#1873, lane pickup-launch)
A picked-up client serves accounts and the catalogue, but a launch on it is refused at bus
admission (model-provider-unavailable): the host launcher keeps its boot system with no proxy,
and the bus has no ModelKeys. While the bus holds a late client and the catalogue is not
disabled, the catalogue appends one notice after its own: "Restart this organization to start
agents on these models." A boot proxy, a disabled catalogue, refused settings and a dropped
client carry no such notice. Tests cover the launch refusal, refused settings, a dropped
client and a key scan over the launch path.
2026-11-19 12:08:41 -06:00
jason.woltje 5c975faddd fix(model-proxy): review round: the preflight comment says every v8 proxy applies a key PUT by an async reload; a re-probe waits about the time left, not the client default; the README says the proxy unit reads PIN from the tree it was installed from and names .codex-refresh; two test headers stop naming only 8.0.42 (#1873, lane proxy-pin) 2026-11-19 11:08:23 -06:00
jason.woltje 462215b4bd chore(model-proxy): pin CLIProxyAPI 8.0.79; proxy.json from an earlier PIN stays valid; preflight re-probes a just-issued key's 401 briefly (#1873, lane proxy-pin)
The PIN moves from 8.0.42 to 8.0.79 with the upstream archive checksum and the
extracted binary checksum. The panel stays 1.26.7.

proxy.json keeps the binary path of the PIN it was created under. A path of the
form <dataRoot>/opt/cliproxyapi/<version>/cli-proxy-api now validates and reads
as the current pinned path, so an instance created on 8.0.42 needs no edit. The
runner still runs only the verified pinned install.

8.0.79 applies a key-list PUT by an async reload a few milliseconds after it
answers, so a just-issued key answers 401 at first. Preflight now probes again
on 401 only, for at most ISSUE_PROBE_TIMEOUT_MS (2 s). Any other non-2xx still
refuses on the first answer.
2026-11-19 11:08:23 -06:00
jason.woltje d0293d6602 fix(bus): review round: proxy init creates the lock before it installs the client, so a lock that refuses leaves no client; a default-instance bus test proves the catalogue refreshes at once on pickup and on drop (#1873, lane proxy-pickup) 2026-11-19 09:51:39 -06:00
jason.woltje 912138ff61 feat(bus): a running organization bus picks up the shared proxy without a restart (#1873, lane proxy-pickup)
A bus started before the instance had a proxy checks once at the next accounts, page, catalogue
or fence request, and loads the client through the start path's checks, loader and lock. A client
got that way, or by the bus's own init, is checked again at each such request and dropped when
proxy.json goes away, no longer loads, moves port, the key stops loading, the lock refuses, or a
dev fixture the host would refuse is in use. While there is no client, the bus names the cause in
one plain line on not-configured and on the unavailable catalogue, and the Console shows it.
2026-11-19 09:39:41 -06:00
jason.woltje 4a8d310cd5 docs(webui): the Triage settings comment says the host reads the stored route at its next admit, now that the host has a route (#1873, lane triage-wiring) 2026-11-18 16:01:19 -06:00
jason.woltje c28ca57af6 fix(triage): review round: the host writes claude-code's bundle and sets its MOSAIC_ names, so a real harness answers; pi is refused as incompatible; proxy.json is read at each admit; a persistent marker makes the start after an unknown end fresh across a restart; the session key and key-shaped strings are removed from Triage lines (#1873, lane triage-wiring) 2026-11-18 16:01:12 -06:00
jason.woltje 255c1575bb feat(triage): the Console resolves Triage's model route at each admit
The Triage host got route: null. It now gets a route read fresh at each
admit from the route store, the key file, each active organization's
model catalogue (held to checkRoutes) and the harness module. The spawn
env adds only ANTHROPIC_BASE_URL (the instance proxy) and
ANTHROPIC_AUTH_TOKEN (the Triage key); the host sets every MOSAIC_ name.

The key file reader moves to packages/cli/src/triage-key-file.mjs so the
Console reads the key with the key verbs' checks without loading the
proxy management client; triage-key.mjs re-exports it. Each refusal (no
route, no key, unsafe key, proxy down, model not offered, no catalogue,
harness incompatible) launches nothing and answers 409 with one plain
line. Every admit runs the host's fences on the answer, a session
running or not.

The cap fails closed: two turns in a row with unknown usage end the
session with a host line; the next message starts a new run with a
fresh session directory, and its usage counts again.

#1873, lane triage-wiring
2026-11-18 16:01:12 -06:00
jason.woltje 0ba50111f6 test(webui): the Triage drawer's no-Triage-request checks skip the static triage-settings.js script every page loads; it carries no data, and the Settings panel's own test shows a non-owner reads no route (#1873, lane triage-settings) 2026-11-18 15:06:03 -06:00
jason.woltje dab07f0ff5 fix(webui): Triage route review round: a save sends the revision its fields were read at, so a route saved elsewhere answers as stale instead of being written over, and an untouched page follows the store; a read follows the repo GET convention, so a LAN dev host with no Fetch Metadata reads the route while a write still needs one matching Origin and same-origin; a zone in another case is stored in its canonical spelling; focus lands on the saved line; the organization catalogues are read in parallel (#1873, lane triage-settings) 2026-11-18 14:40:22 -06:00
jason.woltje 561ff77e48 feat(webui): Settings > Triage reads and saves the Triage route (#1873, lane triage-settings)
GET and POST /api/triage/route, owner only behind the proxy-admin gate, with one
Origin and Sec-Fetch-Site same-origin on a write and Fetch Metadata on a read.
GET returns the stored route or says it is not set, plus the merged catalogue
choices (model keys, availability, compatible installed harnesses) with no
provider, upstream, module, key or path. POST takes model, harness, cap, time
zone and revision, checks the route with checkRoutes against the merged
catalogue (offered, available, compatible, installed), requires a known time
zone with no default, and writes through the store's writeRoute, so the lock,
the revision compare and the receipt are the store's. A stale revision answers
409 with the stored revision. updatedBy is the signed-in owner.

The Settings > Triage page shows only to the owner /api/config marks proxyAdmin.
Model and harness selects (the harness list is the chosen model's), a cap in
tokens, reset 00:00 read only, a time zone, Save. Not set before a first save;
a refusal is one line and keeps the fields; a stale save offers Reload route; a
second Save while one runs does nothing; one line says saving does not start
Triage. cli.mjs keeps route: null for the Triage host.
2026-11-18 14:40:21 -06:00
jason.woltje e1642be760 test(webui): drawer Stop review round: two Stops run at once in two drawers; coder's Stop the bus takes ends and repaints the page while ideas' still runs, and ideas keeps its running state; then ideas' is refused and each drawer keeps its own line (#1873, lane sp20-drawer-stop) 2026-11-18 14:33:52 -06:00
jason.woltje b23ccdfc12 fix(webui): Stop in the agent drawer: an archived or unavailable organization says so in Restart's plain line ending Nothing changed, not Outcome unknown, with one test each; each Stop keeps its own state per agent, as each restart does, so a second Stop or question in another drawer no longer drops a running Stop's line or its running state, with a two-drawer test (#1873, lane sp20-drawer-stop) 2026-11-18 14:33:52 -06:00
jason.woltje 57af33630d test(triage): a repeated usage.begin keeps the first baseline (review round, #1873, lane triage-cap) 2026-11-18 14:03:50 -06:00
jason.woltje 17b972b92d feat(triage): the host holds the daily token cap
Each turn's usage goes to <dataRoot>/triage/usage.jsonl, one line per
usage.end, 0600 and owned by the running user; a link, wrong mode or
owner, or a non-file refuses. The host holds the stored cap through
daily-quota mayLaunch and report before each start and each queued
message, lets a crossing turn finish, drains the queue with a host
line, sums the file again after a restart (a torn line counts as an
unknown turn), and GET /api/triage adds usageToday, cap and resetsAt.
With no route stored the fields say "not set". cli.mjs keeps route
null. (#1873, lane triage-cap)
2026-11-18 14:03:50 -06:00
jason.woltje d531d7181d fix(triage): review round for the triage key verbs
- The grace is the earliest retire time, not a timer: rotate, status and the
  README say the old key stays live until retire runs after it.
- A temp file left by a kill before the rename is removed under the lock by
  the next issue, rotate or retire; status counts it.
- The receipts file is checked under the lock before any change
  (triage-key-receipts-unsafe); a receipt that fails after the change is
  triage-key-unrecorded.
- A failed directory sync after the rename is triage-key-unsynced, after the
  receipt is written, not triage-key-unwritten.
- proxy-lock-busy points at mosaic proxy fence status.
- Tests: a sentinel lost on the read-back, FIFO key and receipts files, a stale
  lock reaped, a crash before the rename, the receipts checks, the sync
  failure; the no-op assertion is gone. (#1873)
2026-11-18 13:54:13 -06:00
jason.woltje 67e1ff193d feat(triage): mosaic triage key issue|rotate|retire|status, Triage's own model proxy key
issue adds one mst-<tag>-triage- key under the instance proxy lock (read,
sentinel, re-read, fence, PUT, read back, confirm) and writes it to
<configDir>/secrets/triage/model (directory 0700, file 0600, the running
user's; a link, a non-file, a second link, a wrong mode or owner refuses).
rotate adds a new key and keeps the old one until now plus the bus's
rotateGraceMinutes from proxy.json; retire removes retiring keys once due.
status reads the file and the list, takes no lock and writes nothing.
Outputs, errors and receipts carry key ids only. A crash at each step
leaves a state status names and the next verb clears. Org and no-org bus
reconciles leave the key alone. Nothing wires the key into the host (#1873)
2026-11-18 13:54:13 -06:00