`spawn_worker <name> local` started the one agent the model was
installed with. The server now lists every agent the model can be run
with, so the skill passes the choice on:
- `local_model` adds `agents` (each with its status, whether it is the
default, and the readiness mark the server measured) and `models`
(what is on the Mac; only the active one runs). A new reason,
`no-agent`: installed, but nothing that can run the model is there.
- `spawn_worker <name> local:<cli>` (`local:pi`, `local:deepseek`,
`local:opencode`, `local:claude`, also as `name:local:pi` in
`spawn_workers`) starts an agent whose status is `ready`. Anything
else fails the spawn before a session exists and names the agents
that are startable. No agent is ever installed from here: that stays
a person's click in the window. A Codeman that lists no agents knows
one, and only that one can be named.
- The readiness mark comes from the server when it names one, else from
what this skill measured (OpenCode `┃`, Pi `extended-keys`), and
`LOCAL_READY_MARK` overrides both.
- A Pi worker on the local model is started with
`piConfig.approveProjectTrust: false` (--no-approve). On a custom
model Pi keeps no home, so it asks to trust the folder in every
session with Trust preselected, and the task prompt's Enter would
answer that.
Pi's mark is the last line of its startup, a warning about tmux's
`extended-keys` option. Its composer rule is painted in the first frame,
before Pi takes Enter: a prompt sent at the rule stayed unsent, 2 of 2,
and one sent at the warning was submitted, 4 of 4 (pi 1.5.2, Linux). It
is a side effect, not a contract, so a Pi that stops printing it fails
the spawn and the worker is deleted, never handed back on a guess.
The preamble changed, so CODEMAN_PREAMBLE moves to 1.51.0 in the
preamble, the SKILL.md copy of it and every version check, and the
plugin mirror is synced.
Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>
The installed screen gets one more block, "Start a session with": a chip
per coding agent the model in use can be run with, as a radio group, the
server's default selected. A chip that is not on the Mac says so.
The main button follows the selected chip:
- an agent that can be started: "Start a session", with that agent;
- an agent that is not on the Mac: its literal install command above the
button, and "Install <agent>". The click is the consent; the server
runs its own stock command, and the window shows "Installing" on the
chip while the model stays ready and another agent can be started;
- an agent that needs a setup of its own (DeepSeek Harness without a
terminal profile): one line that says where, nothing to press.
A failed install shows the server's sentence under the chip with the
installer's last words folded away; the model's own headline is
untouched. On a stopped model Start stays the one main button. The
choice is remembered on the device. A server that lists no agents gets
the screen it had.
Rendered in Chromium at 1440, 390 and 320 px on a dark and a light
skin: the chips wrap, nothing leaves the window.
Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>
An installed model knew one coding agent: the one it was installed with.
The launch check never asked about the agent, only about the endpoint
and the model, so nothing stopped a second one. What was missing is the
list, and a way to get an agent that is not on the Mac yet.
The install view gains `harnesses`: every catalogue harness the model in
use can be run with, each with how it stands (`ready`, `installable`,
`installing`, `setup-needed`), the literal install command, a readiness
mark for scripts, and the actions the server accepts for it. `cli` stays
the default agent, and names one that can be started whenever one can,
so a client that reads only `cli` never starts an agent that is gone.
`POST /api/local-model/agent { cli }` installs one more agent beside an
installed model. It is not a job and not an event of the reducer: the
state stays `installed` whatever the installer does, so a failed npm
install never blocks a session on a model that works. Same install slot
and same compiled stock command as the `agent` step, one at a time, the
outcome in memory only.
The catalogue:
- `pi` is no longer gated. Its two open questions were answered by a run
against a real model (the `name:tag` id reaches the runtime unchanged;
its first request is 2,096 tokens), and Codeman now passes it the
window. Its minimum is 65536 like the others: nothing below was run,
and no model row offers less, so no Mac's pick changes.
- `deepseek` is new. It is a launcher: its install command leaves it
with nothing to launch, so it is never reported installable (read off
`discovery.launcherProfile`, not the id) and a Mac is offered it only
where it was already set up. Its tool round trip was seen on a
scripted endpoint only; the header says how to gate it again.
- a row may carry `readyMark`; `opencode` has the measured one.
A switch no longer refuses for the agent a model was checked with when
another agent that can run it is on the Mac.
Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>
A pi session on a custom endpoint got a model list entry with only the
model id. pi then takes the model for a 128,000-token one (its footer
reads "/128k") and asks for 16,384 reply tokens, so on a 65,536 or
32,768 window it never compacts before the server's real window is full.
The entry now carries `contextWindow` (the full window) and `maxTokens`
(a quarter of it, 8,192 at most) whenever the endpoint's context length
is known: the same two figures the DeepSeek Harness patch already
carries, now from one helper. An unknown length writes neither, as
before.
Measured on pi 1.5.2: with both fields `pi --list-models` shows the
window, and the request carries `max_tokens: 8192`.
Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>
A job started from a model the person had stopped runs the runtime, and
the model it ends on is in use and running. The stored "stopped" flag
stayed set, so that model would not have started with the server again.
Both verbs now set it back once their job started, the way Start does.
Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>
The local model window could hold one model. A second install was refused
(409 already-installed), and a model a Remove kept could never be freed
from the window again. The store can now hold several models, and exactly
one of them is the model in use: the one the endpoint record lists, the
runtime is started for and a session may be launched on.
Server
- LocalModelStateFile.library lists every model that passed the check
(id, label, context, digest, size, coding agent). The service writes
the entry when verify passes; the reducer carries it and remove drops
it. An installed file of an earlier build counts its one model.
- POST /api/local-model/install is allowed from installed: it adds a
model, and the earlier one stays. A replaced job no longer deletes a
model of the library, only an unchecked or partial download.
- POST /api/local-model/use switches to a checked model: a job of the
two steps verify and connect. The stored digest is compared before
anything is loaded (model-changed otherwise). Nothing is downloaded.
- DELETE /api/local-model/models deletes one model by its name, through
the runtime's own delete, never by a path, never the model in use.
- Adding and switching answer 409 sessions-open while a live session
uses the endpoint. Until such a job ends installed, a session start is
refused with the new code job-unfinished.
- The view carries library (the state's entries joined with what the
running runtime lists: ready, changed, missing, unverified) and store
(the size of the blob folder). Both are cached and refreshed in the
background, never read on the one-second poll.
- With nothing installed, remove is offered while the store still holds
models an earlier Remove kept.
Window
- The installed screen lists the models on this Mac with the context
each was checked with, its state and its size on disk, with Use and
Delete per row (Delete asks once, in the row) and Add a model.
- In the chooser a model that is on this Mac already says so and is
switched to without a download or a confirm screen.
- A paused or failed job that has a checked model beside it shows the
way back to that model.
Every route is admin only in multi-user mode, like the existing ones.
Not run on a Mac: the tests use fakes and the fake runtime, and the
listing, store size and delete were checked against a real Ollama 0.13.0
on Linux, not the pinned 0.40.15 build for macOS.
Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>
A 16 GB Mac got no model: the two 9B builds of the catalogue need 10.2 GB at
65,536 tokens, 0.987 of the estimated budget, just past the 0.98 a tight fit
may reach.
New row `ornith:9b-q4_k_m` (Ornith 1.0 9B, MIT), a fine-tune of Qwen3.5-9B for
coding agents whose GGUF build has no vision projector: 5.63 GB against 6.55 GB,
so 9.3 GB at 65,536 tokens, 0.898 of that budget. SWE-bench Verified 69.4,
SWE-bench Pro 42.9, Terminal-Bench 2.1 43.1 on Ornith's card. Its config.json
and its GGUF header are those of Qwen3.5-9B (no prediction layer), so it shares
the figures of the two other 9B rows. Never run on a Mac: `estimated`.
- What changes for a Mac: 16 GB goes from no model to `limited` with this row,
tight. That is the only pick that changes: the row is ranked under MiMo V2.6
9B, so every Mac that could already run a 9B keeps its pick and gains an
alternative. With the budget a 16 GB Mac mini measured (10.7 GB) the pick is
MiMo as before.
- Why under MiMo although its own card's numbers are higher: the two vendors'
tables put their common base 7 points apart, and runtime issue 16992 (open)
was filed on the 35B sibling, which has the same renderer and parser. The
35B stays set aside.
- The 16 GB rule itself is untouched: same budget estimate, same 0.98, same
eight context checkpoints.
- Ranks renumbered 1 to 10; the pins of test/local-model-recommend.test.ts and
the table of docs/local-models.md follow.
Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>
The "Run a local model" chooser showed the pick and four more models. It now
lists every catalogue model that fits the Mac, and the install accepts every
listed row.
- MAX_ALTERNATIVES goes from 4 to 11, above what the catalogue holds. It stays
a bound: test/local-model-catalog.test.ts fails for a catalogue that outgrows
it, so nothing drops off the list unseen.
- New row `muse-glimmer:30b-q4_k_m` (Meta, Apache-2.0): SWE-bench Verified 76.0,
SWE-bench Pro 51.2, Terminal-Bench 2.1 51.7 on Meta's card. Its K/V cache is
the smallest of the 27B class, so it is the first model of that class a 32 GB
Mac holds: 21.2 GB at 65,536 tokens, 0.970 of the estimated budget. Figures
computed from its config.json and its GGUF header by the llamacpp rule (13
full-attention layers, 39 sliding-window layers of 2,048 tokens, no
prediction layer); the sliding-window cache rule was read again in llama.cpp
b12111 (src/llama-kv-cache-iswa.cpp). Never run on a Mac: `estimated`.
- What changes for a Mac: 32 GB is no longer `limited`. On an M1 Max or M2 Max
the pick is Muse Glimmer; on a base or Pro chip the dense model is slow and
the pick stays MiMo V2.6 9B, now with the small-pick note. Every other pick
is unchanged.
- The "set aside" list was read again at its sources on 2026-11-10 and gained
ripple-preview:65b (every build requires runtime 0.40.16, one above the pin),
Ornith 1.0 (runtime issue 16992), nemotron-3.5-super:120b, qwen3.5:4b,
qwen3-coder-next and the MLX and DFlash tags of Muse Glimmer.
- The header and the guide now say that runtime issue 16992 has one diagnosed
cause in the tool parser every Qwen-family row shares (a tool argument that
quotes a whole tool call is cut off; the fix is runtime PR 17021, open).
Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>
Review follow-up. omp declared `hooks: 'always'`, which every consumer reads
as Claude's full hook set, so an omp session picked up behaviour it cannot
back. omp goes back to `hooks: 'none'`; `capabilities.approvalHookFile` is now
a separate, narrower capability with its own predicate,
`approvalHookAvailableForMode()` (local, non-adopted sessions only).
- An explicit `stop` wait on omp is a 400 again; `blocked` still works.
- omp inbox items are alert-only (`answerInTerminal`, no `approvalId` on the
push), and only `permission_prompt` / `elicitation_complete` are accepted for
an approval-only session.
- The module is installed from POST /api/quick-start too (new and existing
cases, not Docker or remote) and on reboot-restore, after the launch
pre-flight on both create paths. Claude's settings.local.json block and the
Docker CLAUDE.md no longer land in omp workspaces.
- The generated module passes Bun's `tls: { rejectUnauthorized: false }` for an
https CODEMAN_API_URL (checked against a `--https` server), clears the alert
only when the last pending toolCallId resolves, sends reports in order,
carries omp's policy `reason` as `message`, and imports HookAPI from
`@oh-my-pi/pi-coding-agent/extensibility/hooks`.
- Every path component and the target file are lstat-checked.
- docs/omp-integration.md, verbs.md and the stock.ts comments updated.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Ctrl+V in Firefox still pasted the clipboard twice some of the time.
Right-click → Paste never did.
`_handleImagePaste()` focuses a hidden contenteditable trap, calls
`document.execCommand('paste')`, and reads the clipboard from the paste
event that lands on the trap. Firefox dispatches that event synchronously,
inside the `execCommand` call. It then delivers the key's own paste as a
separate task, about 10 ms later in Playwright's Firefox. The one-shot flag
from b87bc687 drops that second event, but only if it still reaches the
trap. The trap removes itself and refocuses the terminal on a
`setTimeout(0)` queued by the first event, and that timer races the second
paste. When the timer wins, the second paste lands on xterm's textarea,
xterm's own `handlePasteEvent` sends the clipboard again, and the flag
never sees it.
The trap now cancels the keydown when its paste listener has already run
by the time `execCommand` returns, so Firefox never issues the second
paste. Both callers pass the keydown as `target.keyEvent`: the main pane's
key handler in terminal-ui.js and a tile's in terminal-tile.js. Chromium
and WebKit refuse the command and dispatch nothing from it, so their
keydown is left alone, because its default action delivers their only
paste event.
Measured in Playwright's Firefox with the trap removed before the key's
paste arrived: the paste reached xterm's textarea and the text went in
twice. With the keydown cancelled, the key's paste never fired. End to end
on an isolated instance with a shell session, 20 Ctrl+V presses gave 20
pastes in both Firefox and Chromium.
test/image-paste-trap.test.ts gains a harness option in which
`execCommand` delivers the paste from inside the call, as Firefox does.
New cases cover the keydown being cancelled for the main pane and for a
tile, being left alone when `execCommand` delivers nothing, and a caller
that passes no keydown. The two cancel cases fail without this change.
test/terminal-tile-input.test.ts now expects the tile to pass its keydown
along with its terminal and session.
Docs: the Terminal paste section of docs/architecture-invariants.md and
the paste-trap entry in src/web/CLAUDE.md now describe the race and the
rule. The section's old claim that the key's default action alone always
reaches the trap is replaced, since in Firefox it can miss it.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
An omp pane asking for tool approval raised no tab alert, no desktop
notification and no Approvals Inbox item. omp inherited hooks:'none' from
agentDefaults(), so hooksAvailableForMode() answered false and
hook-event-routes never opened an item; there was nothing to resolve either.
omp is one of the few non-claude CLIs with real approval events
(tool_approval_requested / tool_approval_resolved), so Codeman can install a
hook module into <workspace>/.claude/hooks/pre/ - the directory omp scans -
on session create. It forwards them as permission_prompt and
elicitation_complete, the names the server already routes on, so the existing
alert, notification and inbox pipeline works unchanged.
The file name comes from the new capabilities.approvalHookFile field rather
than a branch on the CLI id (cli-registry-no-id-branching). stop/idle_prompt
are deliberately NOT reported: omp's idle stays screen-read via workDetect,
which the agent skill's verbs.md now says explicitly.
The plain preview already serves a workspace dotenv file in full to the same user, so refusing the edit protected nothing. The edit gate now passes allowDotenv to the sensitive-path check, which skips only the two dotenv patterns; every other credential pattern, a dotenv file inside a blocked tree and Codeman's own data-dir login file stay blocked. Downloads and attachments are unchanged.
_onWebSpeechEnd counted every run that ended without a result toward the
two-retry cap, however long the run lasted. Under 30 s or Manual, each
no-speech window of the engine is one silent run, so the third ended the
recording at about 18 s of silence, before the deadline, and Manual
stopped on its own. Both recognition.start() sites now stamp the run start;
an empty run counts toward the cap only when it ended within a second of
starting (a recognizer that cannot listen), and any other run resets it.
The 250 ms backoff is unchanged.
The vm harness in test/voice-silence-timeout.test.ts now passes the test's
Date into the context, whose own Date ignores fake timers.
Also from review: the VoiceInput stubs of the four settings-ui.js tests
carry SILENCE_TIMEOUT_SECONDS; Settings-Reference lists Stop after silence;
the touchscreen Linux fallback leaves out ChromeOS and says a touch Linux
laptop still matches; the dead _webSpeechStream teardown is gone; the two
voice tests use vitest expect instead of node:assert.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The floating log viewer reconnected every few seconds when its tail-file
stream ended with a startup error (path refused, file missing, too large,
stream limit, tail spawn failure). It now closes the EventSource on an
`error` message that arrives before `connected`; an established stream
keeps its reconnect.
Squashed from #720 (johoja12). Maintainer edits while landing, as the
review said: the onerror close is dropped (it stopped the browser's own
retry after a transport failure, leaving the window on "Connecting"),
the test's check for a button nothing renders is gone, and the stubbed
stream sets `retry: 200` so a reconnect, if one came, shows within the
shorter wait.
Co-authored-by: Codeman maintainer <kay@stutz.science>
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
- session-ui.js: drop the answer when the case-sync chain moved during the read; settings-ui.js calls _applyServerCaseSettings with ?.
- i18n.js: zh-CN for the failed case-options save toast, case name passed through
- tests: regression for the stale read, zh-CN toast row
- sync-case-state browser and stale-refresh tests use lastPickedCase (the stale test now fails without its guard)
- saveLastUsedCase runs on the case-sync chain, and the follow is skipped while this device's own pick was not stored
- _caseStateSyncOn tolerates a harness without loadAppSettingsFromStorage
- lastPickedCase added to the config-backup volatile-key test and docs; rename rollback tests cover caseSettings and lastPickedCase
- wiki row moved to System
Codeman > Keep This Mac Awake in the macOS app: While Agents Work (an
agent at work, plus 5 minutes after the last one stops) or Always While
Codeman Is Open. Off by default, and only for a Codeman on this Mac.
- Idle sleep needs no password: the app holds an NSProcessInfo activity
with IdleSystemSleepDisabled, which the kernel drops with the process.
While it only watches for work, an activity that still allows sleep
keeps App Nap from slowing its look at the server.
- On a MacBook, Only on Power Adapter (on by default) pauses it on
battery, and Also With the Lid Closed installs a small LaunchDaemon
behind macOS's administrator prompt. It sets pmset disablesleep only
while the app keeps a fresh request file with its live pid, only on
the power adapter, and undoes only what it set itself. The app writes
the script and plist as root from the copies built into it.
- A status line at the bottom of the submenu says what it does now.
This brings the macOS part of the keep-awake idea from PR #595 into the
app; the server-side setting and the Linux route are left out for now.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
- follow lastPickedCase (written by saveLastUsedCase while sync is on), not lastUsedCase, so a quick-start or an agent's spawned worker no longer moves other devices
- close the open case-options popover when the follow or an import changes what it shows; follow through selectQuickStartCase(save:false)
- failed save reuses _noteSettingsWriteResult and says the choice lasts until the next sync
- rename undo renames the shared entry back instead of restoring the whole map; lastPickedCase follows a rename
- caseSettings and lastPickedCase added to the redaction, rename and backup-volatile lists and the docs
The fixes#529's review asked for, each with a test.
- No inline handler. The button quoted the path inside
onclick="...decodeURIComponent('<path>')", and encodeURIComponent leaves
' alone, so a file named x'-alert(1)-'.txt ran script in Codeman's
origin and it's.txt was a syntax error. It is now bound in the existing
.file-tree-item loops from the row's escaped data-path.
- Symbolic links are refused, never followed. validateSessionFilePath()
returns the realpath, so deleting AGENTS.md -> CLAUDE.md deleted
CLAUDE.md and left the link dangling. The path as named is lstat'ed,
a link answers 400, and the inode is compared again before the unlink.
- The edit route's confinement whole: validateLocalSessionFilePath()
behind withGuardedOutsideWorktree(), so an absolute path in a followed
outside worktree obeys the same rules, and the sensitive, blocked and
.git checks run on both the realpath and the path as named.
- Adopted sessions are refused like remote ones (the /tmp exception of
the older file surfaces is not copied to a new one); Docker workspaces
are allowed, as on the edit route. The workspace is probed bounded
first (#516). Errors are thrown with their status.
- Deleting from search results re-runs the search, the row buttons stay
visible on touch screens, the confirm and toasts have zh-CN text, and
no button is drawn for remote or adopted sessions.
- Docs: CLAUDE.md, architecture-invariants, api-reference and the
Working With Files wiki page.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The review items offered on #608, applied on top of the contributor's
zoom:
- `object-fit: scale-down` instead of `contain`, so a small image keeps
its own size instead of being stretched to fill the preview. The pan
clamp now uses the size the image is actually drawn at (natural size
fitted with scale-down) rather than the image box, which let a small
or letterboxed image be panned entirely out of view.
- Wheel zoom is proportional to deltaY (d3-zoom's scale: about 15% per
100px notch, lines and ctrl+wheel trackpad pinches scaled up), capped
at sqrt(2) per event, and a wheel event with no vertical delta zooms
nothing. A trackpad scroll no longer jumps to 600%.
- The zoom controls show only once the image has loaded; on an error
they are hidden and the gesture listeners are dropped with the image.
Both handlers now check the request sequence like the text branch, so
a late load or error from a replaced preview no longer resets the
current preview's zoom or paints its error over the current image.
- test/path-picker-image-zoom.test.ts drives the real listeners in
jsdom (focal point under the cursor and the pinch centre, pan clamp,
wheel scaling, error, close and stale events) in place of the
source-string checks in test/path-picker-ui.test.ts. 7 of its 10
tests fail against the contributor's commit alone.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The path picker's image preview gets zoom controls (out, fit/percentage,
in), a two-finger pinch that keeps the pinch centre in place, one-finger
pan while zoomed (clamped to the image bounds), double-click to toggle
2x, and wheel zoom on desktop. closePreview() removes every listener and
drops the zoom state.
Salvaged from #608.
- session-ui.js: no GET /api/settings on a tab return unless this device's cached syncCaseState is true
- system-routes.ts: caseSettings is merged only while syncCaseState is true (as stored or set by the same PUT)
- sync-case-state tests cover both; wiki notes the stored options stay when sync is turned off
jcode (1jehuang/jcode) joins the CLI registry as the `jcode` run mode:
Run menu, phone overview, TUI, quick-start and POST /api/sessions
(`jcodeConfig: { model?, resumeSessionId? }`), local, SSH and Docker
(dedicated agent-image layer), plus codeman doctor and install.sh
through the generated catalogue. Measured on a live jcode 0.98.1 pane.
- Work detection reads jcode's braille status row and `<n>>` composer;
alt screen with mouse strip and PageUp/PageDown paging like opencode.
- Every jcode TUI is a client of one shared per-user daemon whose tools
run with the daemon's env, so no agent skill injection and no JCODE_*
envOverrides prefix.
- New `capabilities.sharedDaemon`: a session kill spares descendants
outside the pane's process group, so closing the tab that started the
daemon no longer cuts every other jcode session off (measured).
- last-response answers an honest empty for an external CLI with no
transcript reader that the pane segmenter cannot read (jcode alone).
- GET /api/jcode/status via the generic registry resolver.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
The install window polls GET /api/local-model every second. A read of the
state file that was still out while the job wrote its last state and let
go of this.job came back with the old `installing`, and settling that with
no job in the process read it as an interrupted job: `paused` with Resume
for one tick before `failed`. Under CI load this made
test/local-model-install-e2e.test.ts fail on a different failure case per
run (master 86502b69, PR #790).
currentState() now reads again when the job ended during the read. The job
writes before it lets go, so the second read has the final state.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Folding a phone on the home screen left the welcome screen up when the page
had loaded wide (#702): the phone overview's breakpoint listener was wired
from its first show, which a wide-loaded page never reaches. Both home
screens' breakpoint listeners are now installed once from init() and act
only while a home surface is up.
A rail render queued just before the fold could also unhide the desktop
rail behind the phone overview; the debounced refresh now asks again when
it fires.
Tests: test/home-breakpoint-listeners.test.ts (gate) and
test/home-breakpoint-swap.browser.test.ts (browser suite: wide-first and
narrow-first across 599/600 and 767/768, plus a height-only resize).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
- Redact caseSettings for non-admins and never apply a shared map on a read-only page (CASE-SYNC-001)
- Send one case per save and merge it inside mutateSettingsFile, so devices cannot discard each other's cases; rapid saves keep their order (CASE-SYNC-002)
- Check HTTP status and the response body, keep the local choice, report an unsynced save and reuse the forbidden-write handling (CASE-SYNC-003)
- Let the server decide whether sync is on: refresh the cached switch, discard stale refreshes (CASE-SYNC-004)
- Carry shared options through case rename and drop them on delete or unlink
- Give the switch an accessible name; add server, client and two-browser tests
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
The previous commit counted 4 rollback copies of the recurrent state for every GGUF with a
prediction layer, the runtime's default `draft_num_predict`. A tag can set its own in its
params layer, and the runtime then passes that as `--spec-draft-n-max`, which is what
`need_n_rs_seq()` turns into rollback copies. Read on 2026-11-08:
- qwen3.6:35b-a3b-mtp-q4_k_m: draft_num_predict 2, so 2 copies (was counted as 4).
- qwen3.8:27b-mtp-q4_k_m: draft_num_predict 4.
- qwen3.8:27b-q8_0: none in its params, so the runtime's 4.
The 35B-A3B GGUF row's overhead drops by 132 MB: 0.971 instead of 0.976 on a 36 GB M3 Pro or
M3 Max, the same pick. test/local-model-catalog.test.ts now carries the drafted tokens per row
instead of a yes/no prediction layer.
Also states what was verified about llama.cpp's checkpoint default: 8 from b6800 to b8200,
32 from b8250 together with a checkpoint every 8,192 prompt tokens (common/common.h at those
tags), in place of an inferred reason. Checked in the b12111 source as well: the draft context
of a Qwen-family MTP model caches the prediction block alone (`mtp_on_hybrid_qwen`, layers at
or above n_layer()), and the main context leaves it out, so the row counts that layer once.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The catalogue that names each row's runner (feat/local-model-catalog-v2: `runner` is `mlx` or
`llamacpp`, and a row's memory figures follow that runner's rule) needs the weekly check to
notice when a tag starts serving a Mac another runner: a tag that gains an mlx child, or a
list that loses it. The size check usually catches that, but not when the two builds come out
about the same size, and the figures are wrong either way.
classifyModel() now raises a `runner` finding when the row names a runner and the build a Mac
selects runs on another one. catalogueRunner() maps the registry's names into the catalogue's:
`mlx` stays `mlx`; `llamacpp`, and `ggml` (what Ollama calls a GGUF manifest that names no
runner), are `llamacpp`, because Ollama 0.40.15 serves every GGUF build through llama-server.
A catalogue without the field (the one on this branch today) is never held to a runner.
Run live on 2026-11-08 against the v2 catalogue's eight rows: exit 0, every runner matches.
Tests: the mapping, the finding (both directions, ggml, an unknown runner, a row without the
field), the report text, and runCheck against the fake registry with a plain GGUF manifest.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The llama.cpp memory rule of the catalogue left out two allocations that the llama.cpp the
runtime bundles (b12111, the runtime's LLAMA_CPP_VERSION) makes for every GGUF row:
- Context checkpoints. For a model whose state cannot be rolled back (the recurrent state of
the Qwen-family rows, the sliding window of North Mini Code), llama-server keeps host copies
of that state, two or three per request, up to 32 per slot by default
(tools/server/server-context.cpp create_checkpoint). For the 27B that is 32 x 157 MB = 5 GB,
none of which was counted. The runtime now sets LLAMA_ARG_CTX_CHECKPOINTS=8 (read by
llama-server from its environment like LLAMA_ARG_CACHE_RAM; 8 was llama.cpp's own default
before it went to 32), and every llamacpp row counts 8 copies of its state.
- Multi-token prediction. The two -mtp tags and qwen3.8:27b-q8_0 carry a prediction layer
(nextn_predict_layers 1, read from the first bytes of each GGUF; the 9B tags and North Mini
Code have none). The runtime then always passes --spec-type draft-mtp --spec-draft-n-max 4,
so llama.cpp keeps 4 rollback copies of the recurrent state (need_n_rs_seq(),
llama-memory-recurrent.cpp) and the prediction layer's cache in the draft context's cache
type, f16 by default. The rows counted one state and that layer at q8_0.
What moves (test/local-model-recommend.test.ts, "every memory size", re-pinned):
- 16 GB: limited -> none. MiMo 9B at 65,536 tokens needs 10.2 GB, 0.987 of the 10.4 GB
estimate. With the budget the Mac mini measured (10.7 GB after the reserve) it is offered,
0.954, tight; that test is kept. With 4 checkpoints instead of 8 it would be 0.966 on the
estimate: this size sits on the edge, and only a run on such a Mac settles it.
- 18 GB: the same pick, 0.964 instead of 0.928.
- 36 GB: the same picks, 0.966 (27B on M4/M5 Max) and 0.976 (35B-A3B GGUF on M3).
- 48 GB on a Max chip: the 27B at 262,144 tokens is tight (0.853) instead of good.
- 32 GB with a measured budget of 0.74: North Mini Code at 131,072 tokens (was 262,144), the
27B at 65,536 on a Max chip (was 131,072).
- Every MLX row, and everything from 64 GB up, is unchanged.
test/local-model-catalog.test.ts recomputes the new rule from the architecture table and holds
the catalogue to LLAMA_ARG_CTX_CHECKPOINTS. The catalogue header, runtime.ts, the types,
docs/local-models.md, CLAUDE.md and architecture-invariants say the same. Read in the source,
not measured.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The new GGUF rows carry lowercase quantization tags (`-q4_k_m`); pin that
the launch path's model id test (`cliAcceptsCustomModelId()`) accepts every
stock id for every harness the catalogue offers, so a row the chooser shows
can always be started.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The catalogue had seven Qwen-family rows that all carried the memory
figures measured on one model (Qwen3.6 35B-A3B on the MLX runner), so a
9B or a dense 27B was judged with figures that do not describe it, two
rows had the wrong Mac download size, and every Mac below 32 GB got
"none".
Now eight rows, each with the runner a Mac really uses for its tag (a new
`runner` field: `mlx` or `llamacpp`) and memory figures of its own:
- the 35B-A3B MLX row keeps its measured figures and the tested pair;
- every llama.cpp row is computed from its model's config.json: a q8_0
K/V cache for the whole context (what the runtime now sets), its
recurrent or sliding-window state, 1.0 GB of runner and the 1 GiB host
prompt cache;
- the one other MLX row (Ornith 1.5 122B) is computed the same way,
scaled by what the measured row showed, plus the runner's 8 GiB of
prefix snapshots.
Sizes are the registry's Mac pull totals (2026-11-08, lowercase ids
verified there). What each memory size gets: 8 GB none; 16 to 32 GB a 9B
(limited); 36 GB Qwen3.8 27B or the 35B-A3B as GGUF; 48 GB the tested
pair (or the 27B on a Max chip); 64 and 96 GB the 27B or the 35B-A3B;
128 to 512 GB Ornith 1.5 122B. Nothing but the one row was measured.
The catalogue test recomputes every computed figure from a table of
architectures and lets two rows share figures only when they share the
architecture and the runner. The rule tests of recommend() now run on a
frozen copy of the 2026-11-06 catalogue, so a weekly catalogue update
changes only the per-memory table and the real-Mac tests.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
runtimeEnv() now also sets OLLAMA_KV_CACHE_TYPE=q8_0, OLLAMA_FLASH_ATTENTION=1
and LLAMA_ARG_CACHE_RAM=1024. Read in the v0.40.15 source of the pinned
runtime: the first becomes --cache-type-k/-v of llama-server (17/32 of the
f16 cache), the second --flash-attn on (a quantized V cache needs it, and
"auto" or "off" would fail such a load), and llama-server inherits the third
from ollama serve's environment, which bounds its host prompt cache to
1 GiB instead of letting it grow by gigabytes in a session. The MLX runner
reads none of them.
The catalogue's llama.cpp rows are computed for these values, so the two
change together. Not yet measured on a Mac.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Every App Settings Save up to 1.49.0 wrote `agentSkillEnabled: false`, chosen
or not (the switch loaded an absent key as off and Save sends the whole form),
so the new default never reached an install that had ever saved. On its first
start the server now replaces a stored false with true, once per data dir, and
logs one line when it does. It cannot tell a deliberate off from the old
default; that is accepted and stated in the release notes. An off set after it
ran stays off.
It stores true rather than deleting the key: the page's merge overwrites only
the synced keys the server has, so every device that saved on 1.49 kept its own
false, showed the switch off and wrote false back on its next Save (measured in
the JSDOM harness). True wins on every device, including a browser still on a
cached settings-ui.js. For a server with no key the page now drops the device's
copy as well, so the switch shows the default.
The "done" marker is its own file, `agent-skill-default-migrated.json` in the
data dir: it is not a setting, so GET /api/settings never returns it, no Save
can send it back (the strict schema would refuse it) or drop it (which would
re-arm the migration), and a config-backup restore leaves it alone. It is
checked before anything is read and written only after the settings write
landed. The migration runs in WebServer.start() before the first settings read
and before listen (skipped in test mode, like the other boot-time writes),
writes only through mutateSettingsFile(), drops the 2 s settings cache, and
leaves a settings.json that does not parse alone with no marker, so the next
start after the fix runs it.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
This reverts commit 0aa34da8, which made the codeman agent skill default on here as well. That change lives in feat/agent-skill-default-on, together with the self-ignoring copy and the reference-folder symlink check it needs.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
session.test.ts created a session in the fixed `/tmp/test`, so with the agent
skill on by default every `npm test` wrote a skill copy into
`/tmp/test/.claude/skills/codeman/` and left it there. It now uses a temp
folder of its own and removes it afterwards.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Measured on opencode 1.3.0 in a private tmux socket, with a fake
OpenAI-compatible endpoint streaming the reply a character or two at a
time: the marker was on the screen, but the stripped client stream read
D, spinner, O, spinner, N, ... because opencode redraws its spinner
between the streamed pieces, so wait-output never matched it. sendmark
then waited its whole timeout and answered 2 for a reply that had
finished, with the answer file already written (opencode's write tool
wrote it without asking).
The marker wait now runs in rounds of at most 10 s, and between rounds
the answer file is a way out too: a non-empty regular file whose
checksum held still for one round is read, with matched false. The
marker stays the fast path. The answer-file wait after a marker no
longer spins when the exit wait gets no answer, and treats an ended
wait as the worker gone.
The same run measured the readiness mark: with a fresh home and a
custom provider, the first composer bar came with the composer and a
prompt pasted with its Enter at that moment was submitted, so the docs
say so instead of "not measured". The example keeps its input in a
subfolder, apart from the case's CLAUDE.md.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>