ab2336516ffef555fd9f23f5b8c6895f7e3ee58d
658
Commits
| Author | SHA1 | Message | Date | |
|---|---|---|---|---|
 pastilhas
|
773f0e1c8b |
never store a model the picker cannot render
The selected model went blank after a turn: visible only by opening the dropdown.
`officerModel` was coming back as `claude-code` — the bare harness. That is `DEFAULT_MODEL`, used when the
client sends no model and the user has no default. It ROUTES fine (isClaudeModel is a prefix test, and
split('/')[1] is undefined so the CLI picks its own), but it names no catalogue entry, so Radix's
`SelectValue` matched no item and rendered nothing.
It fed itself. A blank picker sends no model, so the next turn fell back to the bare harness again and
overwrote the real value — it destroyed a correctly backfilled `claude-code/opus` within one turn. My own
write path did that, added this afternoon.
Two changes. The server refuses to record a model with no tier: a bare harness means "no explicit choice",
and the honest response is to leave the previous answer alone rather than overwrite it with a non-answer.
And the trigger no longer trusts that its value names an item — an unmatchable value renders as its own
string instead of as an empty control, which also covers a retired model or a vanished OpenCode provider.
Repaired the one corrupted row in place by deleting it, so the transcript family fallback answers for that
session until its next real turn; it now reports claude-code/opus again.
|
||
|
pastilhas
|
d59bc0d152 |
one sessions panel with three views
The /chat sidebar was two panels in a vertical 25/75 split — Live above, Sessions below — plus a directory
dropdown in the list header. Three problems that compounded: the Live panel was empty almost always, so a
quarter of the sidebar was permanently blank; the two panels had entirely disjoint indicators, so a
conversation with a turn running looked ordinary in the history; and the dropdown showed one directory at a
time, which turned "where was I working on that" into a search.
Now one panel, three tabs. Tree answers "what have I got, by project". Latest answers "what was I just
doing". Live answers "what is running". Same rows, same indicators, one place.
Tree is an accordion over /chat/pwds, each header carrying the full path from ~, the session count, and a
live count — plus New Chat here, since the tree only lists directories that already have history. A
COLLAPSED NODE ISSUES NO REQUEST: sessions are read by scanning transcripts off disk, and eager mounting
would be one scan per directory for rows nobody asked to see. `useClaudeSessions(cwd, {enabled})` is the
whole mechanism. The per-directory live count comes from the poll's own `cwd`, never from
`liveSessionTitle`, which reads a whole transcript and whose own comment warns against calling it from
anywhere hotter than a ten-second poll.
Latest needs a new server mode, `GET /chat/sessions?all=1` → `listAllClaudeSessions`. It loops directories
rather than scanning flat, because /clear chains are merged per group and a flat scan would let two
unrelated chains in different directories collapse into each other. Fetched only while the tab is showing.
Rows now carry the model, from the batched `getChatSessionModels` — one query per page, which is what that
query was added for.
THE MIGRATION IS THE DANGEROUS PART AND IT IS INVERTED. Every persisted layout still names `chat-live`, an
appType that no longer exists; `normalizeLayout` would collapse it onto the fallback and WRITE THAT BACK,
which is what made /offscale render three server pickers this morning (
|
||
|
pastilhas
|
fc67973876 |
record a resumed session's model when it switches, not when the turn ends
`createMessageHandler` is built per turn, so `transcriptId` started null on every one — and
`applyModelChange` runs before the turn emits anything, so the `model:switched` it produces arrived while
there was nothing to key the row on. A resumed conversation therefore recorded its model only on the closing
`result`, and a turn that died took the switch with it.
Observed rather than reasoned: the sidecar log said `claude-code/fable -> claude-code/opus[1m]` while the
stored row still said `claude-code/opus`.
Seeded from the resume id, which the Claude path already has in scope for the spawn. A brand-new
conversation still waits for `session:claude` — it has no id yet by definition — but resumed is the case
where switching actually happens.
Also lands docs/chat-sessions-list.md: the plan of record for the sessions-list revamp, agreed with the owner
today. Six numbered targets, four decisions with their reasoning, and the landmines — chiefly that merging
the two left-column panels strands `chat-live` in every persisted layout, which is the same stale-appType
shape that made /offscale render three server pickers this morning (
|
||
|
pastilhas
|
813bc200bf |
a session remembers which model it is using
resume a conversation and the model picker showed the first entry in the catalogue instead of the model the session was actually on. that was invisible for as long as the first entry was `opus`, and became obvious the day `fable` went above it — so this is a pre-existing bug my model-list change exposed, not one it introduced. two defects, stacked. the value did not survive a refresh: `initialModel` came from `props.model ?? locationState?.model`, and both are volatile — a panel channel and react-router location state. on a hard load neither exists, the settings default is skipped (it only runs for NEW chats), and ModelSelector's fallback effect fired `onModelChange(providerModels[0].id)`. and even when it did arrive it could not match. a transcript records the API model; the picker offers CLI aliases. `claude-opus-5` is in a different id space from `claude-code/opus[1m]`, and the mapping is one-way: opus and opus[1m] both serve claude-opus-5, opusplan serves whichever model answered. read out of a real 59MB transcript, the distinct values were claude-opus-5, <synthetic>, claude-sonnet-5, claude-haiku-4-5-20251001, claude-fable-5 — note the synthetic marker and the inconsistent date suffix. so the transcript cannot answer "what should the picker show", and a new table does: chat_session_models (session_id pk, model, updated_at) officer's own id space, so opus[1m] and opusplan round-trip. keyed on the TRANSCRIPT uuid rather than officer's session id, because that is what the session list joins on — which is why the write waits for the harness to announce the id rather than firing at turn start. records the EFFECTIVE model, so a refused switch stores what answered. not derived from chat_session_events, which does now hold model:switched: right data, wrong index. that table holds every assistant message and is unbounded, and "latest model for these 200 sessions" would scan far more than the question deserves. not a chat_sessions table either — a session IS its transcript, and a row holding title and cwd would be a second source of truth for two fields that already have one. BACKFILL RUNS FROM postdb:push, so every server picks it up on its next schema push. bun does fire pre/post hooks for colon-named scripts — verified on 1.3.14 before building on it, because a post hook that silently never runs is the worst outcome here. insert-only, exits 0 on every failure path (the push has already applied; a cosmetic backfill must not make it look failed). seeded 11 of 12 transcripts on this machine. it reads the FIRST model, which sounds wrong for a table storing the current one — but switching only became possible today, so no older transcript can hold two, and first costs a bounded head read instead of reading 59MB. also: the family map returns null for mythos rather than a neighbouring family (Project Glasswing only — a picker entry that 401s is worse than none), and ModelSelector no longer picks a default once a conversation has started. an unset picker is the honest state when the server cannot say. tests: the alias map against the five real transcript values, and the selector's default gated on hasStarted. re-broke the guard to confirm it bites. still to do: `model` on the session LIST rows, which the sessions-list revamp wants. the storage and the batched read (getChatSessionModels) are in place for it. |
||
|
pastilhas
|
57bbcca9ae |
fable, and every model alias the CLI actually takes
the claude tier list had three entries — opus, sonnet, haiku — and the CLI accepts nine. `--help` says
`--model` takes either an alias for the latest model in a family or a full name, and the aliases the
INSTALLED cli accepts came out of its own binary with `strings` rather than from memory: default, opus,
sonnet, haiku, fable, mythos, opusplan, and a [1m] suffix on the families with a 1M-context variant.
verified against 2.1.233.
aliases rather than full ids (claude-opus-5) so a new model in a family needs no edit here. the cost is
that what `opus` resolves to moves under you — right for a picker, wrong for a reproducible eval.
mythos is deliberately left out: claude-mythos-5 is Project Glasswing only, and a picker entry that 401s
is worse than a missing one. the comment says how to add it.
opusplan is the one entry that is a MODE rather than a model — opus plans, sonnet executes.
NOT derived from the harness, though it could be: Query.supportedModels() returns the real list with
display names and effort levels, but it is a control request and needs a live session, and this endpoint
is what the composer calls before there is one. noted as open in the file.
the test pins the alias set and the property everything else rests on: claude-manager derives what it
hands the SDK with `split('/')[1]`, so a bracketed alias only works while the id has exactly one slash and
the brackets are unencoded. tidying an id into claude-code/opus/1m would fail as the harness quietly
running the wrong model, not as an error.
|
||
|
pastilhas
|
4c24df93c1 |
a chat can change model mid-conversation
the selector locked on `hasStarted`, so the tier was fixed for the life of a conversation. nothing justified it — the commit that added the flag says only "default model" — and the hazard it resembles is narrower than what it blocked: crossing HARNESS hands an opencode `ses_…` id to `claude --resume`, which reads as a dead agent rather than a wrong dispatch. changing TIER was always safe. so tier is free, harness is pinned, and the server refuses a crossing rather than relying on the UI to. the claude side does it the way the CLI's /model does: `Query.setModel()`, a control request that needs streaming input mode, which is the only mode this file uses. nothing is torn down and nothing is re-resumed — cwd, permission mode and transcript all continue and only the next response changes. an earlier plan here was to kill the session and re-create it with the stored resume id; that would have worked and was unnecessary. opencode already switched per turn (serve-runner POSTs the model to the live session), so only the UI was in its way. FIXES A SILENT MISMATCH while it is here. model reached the SDK once, in createSession's query() options, and every later turn was pushed onto that same call — so a changed model was dropped while session:init, the cost fold and the durable log all reported the new tier. reachable today by leaving a chat idle past the 30-minute GC: the UI said sonnet, opus answered, nothing logged it. the fold now tracks what the harness actually did rather than what the client asked for, and a refused switch reverts the picker. the switch is a durable `model:switched` event and a centred marker in the transcript, because the useful question is the BOUNDARY — which half of this conversation is opus. per-message model was already stamped server-side and shown nowhere, and claude-sessions.ts reads the FIRST model in a transcript, so a switch was invisible in history. tests: turn-stream proves the marker is durable and lands after the previous model's last words; the selector tests pull in opposite directions, so unlocking everything fails one and restoring the blanket lock fails the other. verified by re-breaking both. not done: opencode's `steer` path (a send mid-turn) returns before the model POST, so a switch there applies on the next full turn. session.model also stays stale for opencode tier changes — harness routing is unaffected. |
||
 pastilhasandClaude Opus 5
|
0d29e242a3 |
themes are installable, like plugins
`platform/themes/<id>/` — untracked, same rule as `plugins/`, only the instructions
tracked. Drop a directory in and the theme appears in the picker on the next page load.
themes/my-theme/theme.css required — `[data-theme='my-theme'] { … }`
themes/my-theme/theme.json required — { name, family, dark }
themes/my-theme/assets/ optional — wallpaper, logo
── Why this is cheaper than installing a plugin ──
A plugin is code: it needs a manifest module, mounting, and a frontend rebuild to reach
the bundler. A theme is a block of CSS custom properties, so it needs none of that. The
server reads the directory at request time and the client adds a `<link>`. No install
step, no rebuild, no restart — which is the whole reason this is eighty lines of
discovery rather than four hundred.
The id is the DIRECTORY NAME, not a manifest field, for the same reason a plugin's
`appName` is not one: an id that can disagree with where its files sit eventually will.
Discovery checks that the CSS actually targets its own directory name, because the
failure is otherwise silent — the theme appears, is selected, and nothing changes.
── Problems are reported, not swallowed ──
A malformed theme returns a reason rather than vanishing:
broken-probe: theme.css does not target [data-theme='broken-probe']
The client logs those, so an author has somewhere to look instead of wondering why
their theme never appeared.
── Boundaries kept ──
The `theme` package still performs no IO. It gained an `extraThemes` prop; Officer
fetches and injects, exactly as it supplies the panel registry to `workspaces`. A
failure to fetch is not an error — no server, no session, no directory all mean "you get
the built-in themes", which is a complete product.
`/api/themes` is a `core` capability: read-only, and not deniable, because an account
that cannot read the theme list renders unstyled. Installing remains a directory on
disk; an upload endpoint would be a way to write arbitrary CSS into the page from a
request.
Verified in-process: listing, the problem report, CSS served as text/css, and a
traversal attempt answered 404.
── Needs a restart ──
Server-side change. Until then the client's fetch 404s and silently keeps the built-ins,
so nothing breaks in the meantime.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
|
||
|
pastilhasandClaude Opus 5
|
997f7e7a87 |
the account stores a theme, not a colour mode
Your theme now follows the account instead of living in one browser's localStorage.
Choosing Plain on a laptop reaches the phone.
`appearance.colorMode` is superseded by `appearance.theme`. Along the way this
resolves a mismatch that had been sitting there: the server's default was
`appearance: { theme: 'light' }` while the client read `appearance.colorMode`, so that
default had never once been read by anything. Both sides say `theme` now.
The Light/Dark item is gone from the user menu. Light and dark are two themes, both
already in the list above it, and that toggle was the last writer of the legacy key.
── Two bugs found by testing this, both mine, both silent ──
**A default that looked like a choice.** `useSettings` returns `DEFAULT_SETTINGS`
until the request lands — `settings` is never null, it is a complete valid object. So
"wait for settings" fired on the first render against the defaults, marked itself
done, applied the default theme and never saw the real one. Then the persist side
wrote that default over the account's actual preference. It overwrote the owner's own
`dark` while I was testing it; restored by hand from the `colorMode` it had not yet
destroyed. `useSettings` now exposes `settingsLoaded` so the difference is expressible.
**Defaults masking a migration.** With that fixed, an account holding only
`colorMode: 'dark'` STILL came up light. The client merges `DEFAULT_SETTINGS` over
whatever arrives, so a missing `theme` is filled in with the default before any
consumer sees it — and a default is indistinguishable from a choice. The migration
therefore belongs on the SERVER, which sees the stored object before anything is
merged over it. It is read-only: a GET that rewrites the row on every call is a
surprise, and the client persists on its next change anyway.
Verified by putting the account into the pre-themes shape and asking the new handler
directly, rather than by restarting to find out:
stored: {"colorMode":"dark"} -> returned: {"colorMode":"dark","theme":"duck-dark"}
── Needs a restart ──
The first server-side change in this work. Everything until now was frontend and a
rebuild was enough.
950 pass / 7 fail unchanged.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
|
||
|
pastilhasandClaude Opus 5
|
690a726f55 |
a member's turn ran as the owner after any idle gap
This is the cause. Everything before it was symptom.
`handleChat` had two spawn call sites. The first, for a new session, passed `member`.
The second — "session already live" — was a separate object literal that did not.
That branch is taken whenever OFFICER still holds the session, and officer holds it
long after the SIDECAR has let go: the sidecar reaps a session after 30 minutes idle,
`_claudeKill` stays set here regardless. So a gap sent a turn with no identity to a
sidecar with nothing to push it onto, and it created a fresh session from those
params. No member means the owner.
[claude:spawn:…] resume=49527b33… cwd=…/data/jg@pertento.ai/home as=owner
green is a Developer with their own Linux account, their own `claude` and their own
login. Every turn after a break ran as the OWNER: owner's HOME, owner's Anthropic
credential, owner's MCP config — which carries a JWT that signs as the owner. The
entire per-member boundary, undone by an omitted line in the branch nobody reads.
It surfaced as breakage rather than as silent escalation only by luck. The owner's
`claude` was handed the member's cwd — a 700 home it cannot enter — and the member's
transcript id, so it exited 1 instead of quietly working. Reported as "he went for a
walk and it stopped working", which is the idle timeout to the minute.
── The fix, and what it then found ──
Both paths now build from one `turnParams`, so they cannot drift again.
Then `member` was made REQUIRED-though-possibly-undefined in the protocol, and the
compiler immediately found three more sites that had left it out:
- `agent-handoff/deliver.ts` — the same bug, worse. No member AND a cwd from
`getOwnerHomeDir`, which ignores the email it is given. A member's agent panel ran
as the owner in the owner's home. Now resolves identity like chat does.
- `agents/agent-runner.ts`, `tasks/pipeline-executor.ts` — correct as they were:
both are `kind: 'execution'`, owner-only and never grantable. They now say
`member: undefined` explicitly, because a field you must fill in is a decision and
a field you may omit is an accident.
`resolveTurnIdentity` is exported so "whose turn is this" has one implementation
rather than three.
── What I got wrong on the way ──
Three wrong answers first — the credential, the working directory, then a recovery
whose condition could never fire — each from reading code and reasoning instead of
making the system say what it was doing. The log line that found this took five
minutes to write and should have been the first thing I did, not the fourth.
Also: `--resume` by hand appeared to work twice, which is what sent me down the wrong
paths. It proved nothing. The id I tested was not the id the platform passed.
tsgo clean, 950 pass / 7 fail unchanged.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
|
||
|
pastilhasandClaude Opus 5
|
d771736a0d |
the preview panel follows the app's navigation
The panel knew one URL: the one you typed. Navigate three pages into an app and the header still showed the front door, "open in a new tab" still went there, and a reload — which fires by itself whenever a dev server restarts — threw away where you were and started again at `/`. Now the app reports where it is, and a reload composes it back. ── How the panel finds out ── The frame is same-origin for a proxied preview, so `contentWindow.location` is readable. It is PUSHED rather than read, because a dev server is an SPA: a route change is `history.pushState`, which fires no event and never reloads the document. Polling would have to be constant and would still lag. The script the proxy already injects now patches `pushState`/`replaceState`, listens for `popstate`/`hashchange`, and posts the path — with the proxy prefix stripped, since that is plumbing and the panel should not have to know about it to show a URL. Only proxied previews can do this. A cross-origin frame reports nothing, and grafting a remembered path onto someone else's host would be inventing a URL, so `targetUrl` composes only when there is a port. ── Two fields, and the loop that forced them ── `url` is what you pointed the panel at; `current` is where it has got to. The obvious design is one field — a browser has one address bar — and it feeds itself: the stored value IS the iframe's `src`, so recording a route change hands the frame a new `src` for the page it is already showing, which browsers treat as a navigation. Every click would cost a reload. Belt and braces, because one of the two would be enough only until someone changed the other: the fields are separate, AND `PreviewFrame` captures its `src` at mount and ignores later changes. Remount-by-key is still how a reload happens. `ev.source` identifies which frame is talking. `message` is a window-level event and two same-origin previews hear each other; comparing the posting window to this iframe's `contentWindow` is exact, where matching on URL would break the moment two panels pointed at the same port. Writes are debounced 800ms. `useDashboardState.setValue` PATCHes the server on every call and an SPA emits a route change per click; the value is only read at mount, so being a second behind costs nothing. Also: the header shows `localhost:3001/settings` rather than `api/dev-proxy/3001/settings`, and editing starts from that — one is a thing you can retype, the other is not. ── Tests ── The reporter is RUN, not string-matched: `new Function` against a fake window, which also means a syntax error in the injected script fails the suite rather than silently breaking every proxied page. Covers prefix stripping, the app root reporting `/` rather than empty, query and hash surviving, the report on load, pushState, and popstate. Plus the composition helpers, including that an ordinary URL is never composed with a path. ── One stale comment corrected while in there ── The file claimed `allow-same-origin` "is present, and has to be". There is no `sandbox` attribute at all, so nothing was present — and a sandbox with `allow-same-origin` plus `allow-scripts` is worth roughly nothing anyway, since the frame can remove its own sandbox attribute. Now says that. Not verified against a live dev server through the real proxy — that needs the server restarted with this code. tsgo clean, 934 pass / 7 fail unchanged, frontend builds. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> |
||
|
pastilhasandClaude Opus 5
|
430346e14b |
a restarted job outlives the server that restarted it
The other half of the same report. The pill came back after a refresh, and then a
`pm2 restart officer` killed the dev server outright — because `restartBackgroundTask`
spawned it as a plain child of the platform. A job the AGENT starts is a child of the
`officer-claude-code` sidecar, a PM2 peer, so those were never affected. The tray
therefore showed two chips that looked identical and behaved differently, and the
restart button's own help text promised the opposite of what happened.
Measured against a throwaway PM2 app rather than reasoned about, because the two
plausible causes call for different fixes — a signal to the process GROUP, or a walk
of the process TREE:
plain `sh -c` killed
`setsid sh -c` killed <- a new session is NOT enough
`setsid --fork sh -c` survived
PM2 walks the tree by ppid, so a new session leaves the job a child and it is found
anyway. `--fork` is what matters: setsid forks and its parent exits at once, so the
job is reparented to init before any restart can enumerate it.
── What that costs, and how it is paid ──
The handle `Bun.spawn` returns is now the setsid parent, which exits immediately.
Waiting on it reported every job as finished the moment it started, so exit detection
moved to `watchForExit`: poll who holds the output file, and when nobody does, read
the status the wrapper recorded. That is the same liveness question Stop and Pause
already ask, so it adds no second notion of "running" — and a paused job still holds
its file, so SIGSTOP correctly reads as running rather than as an exit.
Two details that would each have been a bug:
- A GRACE PERIOD. Between the fork and the job opening its output there is a moment
with no holders; polling into it announces a completion for a job that never ran.
- The command is a POSITIONAL PARAMETER to the wrapper, not interpolated into it.
Interpolation would let a command containing quotes terminate the wrapper's own
quoting. Tested with `echo "quoted 'both' ways"`.
A missing status file is reported as a clean finish, not a failure: that is what Stop
looks like, and "you stopped it" is nearer the truth than "it crashed".
Also corrected the button text. It said background jobs survive "because they are
their own PM2 processes" — true of agent-started jobs, false of restarted ones, and
now the wrong reason for both. They survive because they are outside this process's
tree.
── Verified ──
End-to-end through `buildHonoApp`, on a synthetic task:
POST /restart -> 200
holders: pid 906186 (ppid 1, session leader), pid 906188 (ppid 906186)
probe process was 906173 — in neither chain
quotes survived the wrapper
durable events: task:started -> task:notification (completed), exit file "0"
My own assertion in that probe was wrong and the output is what corrected it: it
demanded ppid 1 of EVERY holder, but only the session root is reparented — its own
children hang off it, which is the point.
Not in `bun test`: nothing in the suite touches Postgres or spawns processes.
tsgo clean, 919 pass / 7 fail unchanged. Needs a server restart.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
|
||
|
pastilhasandClaude Opus 5
|
b7d26a3350 |
a restarted background job survives a refresh
Reported as "I refreshed the page and the pill was gone, but the server is still
running and now I have no way to control it". Not a design choice — a bug, and it
came in with the restart button.
The tray is not stored anywhere. It is a fold over the durable chat event log:
`task:started` minus `task:notification`, per `collectRunningTasks`. That is what
rebuilds it after a refresh.
The restart route announced its respawn with `emitToSession`, which folds into the
in-memory session and writes to open sockets — and never called `appendChatEvent`.
So the pill existed only in the RAM of the browser tab that clicked the button. The
log still read:
seq 10979 task:started "Run dev server"
seq 11362 task:notification failed <- the ORIGINAL dying
Folded on refresh that says "not running", so the chip vanished — while the process
it described was up and serving on 3001. And every task route takes an id, so the
job became unreachable from the UI at the moment the browser forgot the id. Stop,
Pause and Restart would all still have worked; nothing could name the task.
The second half was quieter: `emitToSession` returned false when the session was not
in memory, and the route ignored the return. A platform restart drops the in-memory
session while the job keeps running, so the announcement went nowhere AND was never
written down.
Now it appends first and unconditionally — persistence does not need a listener,
only delivery does — and sends with the seq the log assigned, so a client's replay
cursor stays correct. Both events: the `task:started` is awaited so it is durable
before the client is told the restart worked, and the exit notification is
fire-and-forget because it lands long after the response.
This makes officer a writer of `chat_session_events`, which the note in websocket.ts
said it no longer was. `cut-off` had already been that exception. The real rule is
narrower and is now written down: officer writes events it ORIGINATES, because no
sidecar will. Turn output still belongs to the harness that produced it.
── How it was verified ──
End-to-end, out of band: a synthetic task with a real process holding a real output
file, a transcript in the harness's shape, and the two durable events that reproduce
the reported log. Driven through `buildHonoApp` so the permission gate and auth
middleware were in the path — the last two bugs of this kind both passed tests that
mounted the router alone.
BEFORE restart: 2 durable events -> tray would show 0 running (bug reproduced)
POST /restart -> 200
AFTER restart : 3 durable events -> tray would show 1 running
Not added to `bun test`: nothing in the suite touches Postgres, and this claim is
untestable without it. So this is a probe that ran once, not coverage.
Needs a server restart. tsgo clean, 919 pass / 7 fail unchanged.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
|
||
|
pastilhasandClaude Opus 5
|
1426c3a856 |
rebuild the frontend without restarting, and restart when you must
Settings → System → Maintenance. Two buttons that are deliberately not a pair. ── Rebuild: real, and free ── The same `rebuildFrontend` a plugin install already runs. `Bun.build` reads source from disk on every call and writes `build/`, and the server serves that directory by reading each file per request — so a rebuild lands on the next page load with no process restart, no dropped websocket and no chat session lost. ── Restart: the only way to load server code, and it costs ── Bun caches modules for the life of the process. Measured rather than assumed: editing a file and re-importing returns the ORIGINAL export, and a cache-busting query picks up that one specifier while everything it imports stays cached. So `rebuildHonoApp` — which install already does — re-wires the same module instances into a new tree and cannot see a backend edit. Busting the whole graph would duplicate singletons: two database clients, two session maps, one process. There is therefore no third button, and the screen says why. A "remount the API" control would look like it reloaded the backend and would not. The restart answers BEFORE it schedules itself, because the process serving that request is the one being replaced. It refuses outright when PM2 does not know this process — in development the platform runs under `bun --watch`, where the restart would fail into a promise nobody awaits while the button reported success. Verified: `POST /rebuild-frontend` → 200, 3.0s, 3 outputs; anonymous → 401. The restart button is deliberately untested from here — pressing it would end the session this was written in, which is the same reason it needs a label that says so. tsgo clean, frontend builds, 914 pass / 7 fail unchanged. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> |
||
|
pastilhasandClaude Opus 5
|
92bac490d7 |
attach a PDF and the agent reads it; the microphone button is gone
Two requests, one file apart. ── The microphone ── Removed from the chat composer. The HOOK stays: `useAudioRecording` and `blobToWav` are also the file browser's DictateDialog, which is not what was asked to go. ── PDFs, by not extracting anything ── A PDF entry sat in the attach menu once with no `onSelect` at all — the menu offered it and did nothing, because nothing here could extract PDF text. It is back and real, and it still extracts nothing: the file is uploaded, left on disk, and the agent is given its PATH. Its own `Read` handles PDFs by page, better than any extraction this could have shipped, and it can go back for a section later without paying for it twice. The alternative was base64 in the message. That is megabytes in the turn, re-sent for the rest of the conversation, to produce worse text. ── What this uncovered ── `attachmentIds` has been on this websocket and in `types.ts` for a long time and NOTHING on the server read it. Images work because they travel separately as base64, so the ids were dead on the wire — and a PDF is exactly the case that needs them. They resolve to paths now, appended to the prompt as `Attached files (read from disk)`. Ids arrive in a websocket message and become a filesystem path, so they are checked rather than trusted: only the `<uuid>.<ext>` shape uploads actually generate, and only in the two directories uploads actually write to. Verified — a valid id resolves, an unknown one yields nothing, and `../../../../etc/passwd` and `; rm -rf /` yield nothing. Both upload directories are searched because the first message of a chat carries ids written to `tmp/` (no session yet) and later ones carry ids under the session. PDFs may be 25 MB against an image's 5: nothing carries them into a prompt, so the ceiling is storage rather than context. tsgo clean, frontend builds, 914 pass / 7 fail unchanged. `[open]` For a MEMBER the file lands under `data/<email>/attachments/`, and whether their Linux user can read it depends on the confinement ACLs. Untested — the owner path is what was verified. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> |
||
|
pastilhasandClaude Opus 5
|
c7749308e1 |
a path that climbs out of the proxy is collapsed before it is prefixed
The blank preview, and it was never the restarts. Bun's dev server emits
href="/../../../chunk-qmw09gd9.css"
Root-absolute AND climbing. Prefixed naively that is
`/api/dev-proxy/3001/../../../chunk-…`, which the BROWSER normalises straight back
to `/chunk-…` — the platform root, where it 404s. Two failed asset requests, a
white pane, and nothing in the network tab suggesting a proxy was involved. The
rewrite looked correct in isolation and was undone by whoever read it.
`..` is resolved by the consumer, so it has to be resolved before the prefix goes
on. Both halves do it now: `rewriteHtml` for what is written in the document, and
the injected script for URLs built at runtime, which Bun also emits in this shape.
Verified through the whole stack against the real app on 3001: document 200, both
assets 200 with `text/css` and `text/javascript` — where before they were 404s at
the platform root.
13 tests, and one of them asserts the failure rather than the fix: that
`/api/dev-proxy/3001/../../../chunk-x.css` really does resolve to `/chunk-x.css`.
Without that line the other tests look like arbitrary string manipulation.
── Separately, and not a bug in this code ──
Two processes were listening on 3001, holding DIFFERENT task output files
(`b0oe99a47` and `b6zat0x7e`) — two independent jobs, not one restarted. Bun allows
both to bind through SO_REUSEPORT, so requests were split between them and stopping
one legitimately left the other serving. That is worth knowing before reading a
Stop as broken, and it is the second time today SO_REUSEPORT has made a duplicate
invisible.
tsgo clean, frontend builds, 914 pass / 7 fail unchanged.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
|
||
|
pastilhasandClaude Opus 5
|
bda5eb15de |
a job's controls can no longer freeze the platform
The server went down. It had not: it was STOPPED — eleven threads in state T, port 9000 still listening with connections queueing behind a process that would never accept them. Indistinguishable from a crash unless you look, which is why it took a terminal to see. `kill -CONT` brought it back with nothing lost. I did it, with one missing line. `restartBackgroundTask` opened the job's output file to hand the child a descriptor and never closed it, so the PLATFORM became a holder of that file. `pidsHoldingFile` asks `fuser` who holds it — that is how Stop finds a job with no recorded pid — so the platform was in the answer. The next Pause sent SIGSTOP to it. Every part worked as designed. Pause froze exactly what was holding the file, and one of those things was the thing doing the freezing. Two fixes, and the second matters more than the first: **Close the descriptor.** The child gets its own duplicate at spawn, so closing costs it nothing. That removes this instance. **Never signal our own pids, whatever `fuser` says.** `pidsHoldingFile` now filters `process.pid` and `process.ppid`. The leak is fixed; this is what makes the next one survivable. Nothing that manages a job should be able to name the process doing the managing — a rule worth having independently of the bug that suggested it. Verified: after a restart this process holds the output file zero times, `fuser` returns only the spawned shell and its child, and our own pid is filtered even when present. tsgo clean, frontend builds, 901 pass / 7 fail unchanged. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> |
||
|
pastilhasandClaude Opus 5
|
19e75651e7 |
a restart runs the job in its own directory, or not at all
Restarting a `bun dev` belonging to a project in ~/dockers/officer-atelier started the PLATFORM's dev server instead. The respawn passed no cwd, so it inherited the server's own working directory, where `bun dev` means something else entirely. The result was not an error. `bun --watch src/server.tsx` came up and bound port 9000 ALONGSIDE the running platform — Bun allows that through SO_REUSEPORT — so requests were split at random between two instances of the platform. Confirmed on the machine by ancestry: `sh -c bun dev` → `bun dev` → `bun --watch src/server.tsx`, with two listeners on 9000. Stray killed; the real server was unaffected. The information was there all along: every transcript entry carries `cwd`, and the entry holding the tool call is the one that says where it ran. `commandForTask` returns both now. And when there is no cwd it REFUSES, rather than falling back to the server's own. The fallback is what produced this, and it produced the worst outcome available — not a failure, but a different program running successfully. A command is only meaningful with the directory it was written for. Verified with a job whose transcript says it ran in /tmp/cwd-probe-app: the restart reports that directory and the respawned process prints it, reading a file that exists only there. tsgo clean, frontend builds, 898 pass / 7 fail unchanged. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> |
||
|
pastilhasandClaude Opus 5
|
fccabfcea3 |
the preview authorises itself before it mounts the frame
401 on every asset, white pane. The server was never wrong — replaying the exact browser sequence with a minted owner token gave HTML 200, `set-cookie` present, and the sub-resource authorised. So the fault was in how the frame got its credential, and both mechanisms it relied on were guesses. Putting the token on the iframe URL guesses which key the browser stored it under. Letting the frame mount and hoping guesses that a cookie from some earlier load is still there. When either guess misses, the HTML renders, every sub-resource 401s, and the pane is white — which is the shape that makes this hard to read, because the document loading looks like success. So neither is used now. The panel asks `/api/dev-proxy/session` with the ordinary API client — which already knows how to authenticate, because it sends the header on every other call it makes — and mounts the frame only once that returns. The iframe URL carries nothing. `iframeSrc` is deleted rather than left as a fallback: a second path that runs when the first fails is how you get a bug that only appears sometimes. `Authorization` rather than a query parameter, so the token never reaches an address, a log or a Referer. Re-run on every reload, so a preview left open overnight recovers by itself instead of showing a wall of 401s. ── Also fixed: a lie attached to every binary asset ── The non-rewritable branch returned the upstream's `content-encoding` and `content-length` unchanged, but `fetch` had already decompressed the body — so the browser was told to inflate plain bytes, with a length that no longer matched. The rewrite branch had always deleted both. Images, fonts and wasm were affected; nothing had hit it yet because the test app serves none. Verified: `/session` 200 and sets the cookie with a token, 401 without; the frame 200 with the cookie, 401 with nothing. `[open]` A request that escapes the prefix hits the SPA catch-all and returns `200 text/html` instead of 404, so a `<script>` gets HTML, the MIME check refuses it silently, and nothing appears as an error. That masks this entire class of bug and is worth closing separately. tsgo clean, frontend builds, 890 pass / 7 fail unchanged. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> |
||
|
pastilhasandClaude Opus 5
|
73eb128fb1 |
pause, resume and restart a background job without asking the agent
Three buttons in the expanded task panel, beside Stop. Not on the chip: the panel is where you decide what to do with a job, and a row of unlabelled glyphs on a pill is where you misclick. ── Pause / Resume ── SIGSTOP and SIGCONT — the only pair that means "stop doing things and keep everything". The process keeps its memory, its open files and its bound port, and picks up mid-instruction. Signals go to every pid holding the job's output file, because for a shell job that is the shell AND what it spawned, and freezing one of them would be worse than freezing neither. Verified by watching a job that writes a line a second: 3 lines, frozen at 3 across three seconds of pause, 6 after resume. Worth knowing before pressing it: a paused server does not REFUSE connections, it stops answering them, so a browser pointed at it hangs rather than erroring. That is what pausing is, not a shortcoming here. ── Restart, and the cost it accepts ── The command is recovered by joining the transcript to itself: the harness records the `tool_use` carrying `command`, and separately its result — the text `Command running in background with ID: <id>`. Neither alone is enough; together they are a join on `tool_use_id`, and that is the only place the command survives. The transcript is found FROM the output path rather than searched for, since a task file at `<tmp>/claude-<uid>/<slug>/<session>/tasks/<id>.output` names both. The respawned process appends to that same output file. That is what keeps everything else working untouched — the log view reads that file, and Stop finds the process by asking which pids hold it open. The real cost, chosen deliberately: the platform becomes a SECOND owner of a task. The harness stops watching when its own child dies, so the platform supervises what it spawned and emits the same `task:started` / `task:notification` pair everything downstream is built on. `emitToSession` is the seam, and it is the only one — nothing else learns that a task can have two kinds of parent. A restart re-announces an id already in the transcript, so `task:started` now RESETS an existing task row rather than appending a second one. Without that the tray showed the job twice, once finished and once running. Restart is offered for a finished job too, which is the case that matters most: a dev server that fell over is exactly what you want back without typing a sentence at an agent that may be mid-turn. Verified end to end against a planted task: command recovered, old process killed, new one spawned, output APPENDED (`first run output` then `RESTARTED-OK`), exit reported through the callback, supervision registry released afterwards. `[open]` A restarted job does not survive an `officer` restart — the platform is its parent now. The harness's own jobs never did either. tsgo clean, frontend builds, 890 pass / 7 fail unchanged. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> |
||
|
pastilhasandClaude Opus 5
|
7d64b0a0d6 |
paste localhost:3000 into the preview and see the server's port
No proxy host, no certificate, no port opened. `/api/dev-proxy/5173/…` forwards to `http://127.0.0.1:5173/…` over the platform's own origin, and the panel maps what you type onto it. ── Why a proxy at all, when the ask was "just point at localhost" ── Neither direct route works from a browser on https://<platform>. `http://localhost:5173` resolves on the VIEWER's machine, not the server's, and `http://<server>:5173` is mixed content inside an https page and is blocked before it is even attempted. Both fail looking exactly like the dev server being down. The bytes have to arrive over an origin the page already trusts, which is the platform. A dev server can therefore stay bound to 127.0.0.1. The workflow this replaces — one NPM proxy host per app — needs it on 0.0.0.0 plus a firewall rule for the docker bridge, so this is both less setup and less exposure. ── What came back from `188b45c1^`, and what did not ── The 46-line HTML/asset rewrite, which is the hard-won part: absolute paths in the document, plus an injected script that intercepts fetch/XHR/WebSocket for the URLs built at runtime. Its JWT injection is deleted — those apps called the PLATFORM's API; a dev server proxied here talks to itself, and handing its process the owner's bearer token would be giving credentials to whatever is on that port. The other 314 lines are not back: spawning, port discovery, idle timers, log collection, a proxyId registry. The port is the id, and the background-job button already starts dev servers. ── HMR is bridged, because "see it" without "follow it" is half the feature ── The injected script rewrites the dev server's socket under the prefix; the relay dials `ws://127.0.0.1:<port>` and forwards frames both ways. Same shape as the vault relay including the non-obvious part — state registered synchronously and frames queued until the upstream opens, because Bun requires the upgrade to be synchronous and an HMR client speaks immediately. An upstream close closes the client, which is what makes a dev-server restart reconnect by itself. ── Access, which was not in the ask and is not optional ── This route reaches any loopback port on the machine, on a host that answers from the public internet. An iframe cannot carry a bearer token — it is a navigation — so the panel puts it in the URL once, the proxy verifies it is the OWNER through `resolveAuthToken`, and issues a short-lived cookie every sub-resource then carries. Anonymous and non-owner requests get 401, verified. Owner-only rather than a permission: reading an arbitrary loopback port is the `execution` power the registry says can never be granted. Both exemptions are registered with reasons, and the boot check earned its keep — it refused to start until the socket provider was claimed too. ── Verified against a real loopback-only server ── Confirmed unreachable from outside, then through the proxy: index served and rewritten, `/_bun/` chunk resolved, JSON passed through untouched, 401 anonymous, 401 with a bad token, "nothing is listening" on a dead port. Websocket: refused without the cookie, target parsed, frames relayed in both directions. One bug found by testing rather than reading: the upstream path was derived by slicing a hardcoded prefix, so every asset request returned the index page. It is found from the `/<port>` segment now. A test asserting `http://localhost:5173` is kept verbatim is now inverted, with the reason written beside it — that was correct when the panel only framed real hostnames. tsgo clean, frontend builds, 890 pass / 7 fail unchanged. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> |
||
|
pastilhasandClaude Opus 5
|
0230e62a71 |
the platform stops explaining what Headscale is
Correcting my own reasoning from the last commit. I left "Headscale's API", "the Headscale admin credential" and similar alone on the grounds that they name the thing the plugin drives, so renaming them would make the sentences false. True as far as it goes, and it misses the point: those sentences are in the PLATFORM. It has no business knowing what Headscale is at all. Correct language in the wrong repository is still a remnant. Ten of them, each replaced with the platform's own vocabulary rather than the plugin's domain: - `crypto.ts` and `secret-store.ts` listed "the Headscale admin credential" among what one key used to protect. It is "the tailnet's admin credential" — the purpose, which is a word this repository owns. - `catalogue.ts` and `availability.ts` explained `alsoServes` — a LIVE mechanism — entirely through a Headscale example that left the repository. Restated as the general shape, with one parenthesis saying which case produced it. - The photos types file explained another plugin's sidecar. - `App.tsx` cited `/headscale/:section` as the example of a route pair; that route does not exist. `/photos/:section` does. - `plugins/router.ts` pointed at `useCompanionLogStream`, a hook that lives in a plugin now. - `registry.ts` said in the present tense that `headscale` is `admin` — the permission is the plugin's and it declares its own, so a member-facing enrolment surface would be the PLUGIN's to add, not this file's. - `proxy.sh` cited `headscale/schema.ts` as a style reference. That file is not in this repository. - machine-setup said "Its Headscale app" — the app is the Offscale plugin. `machine-setup/lib/tailscale.sh` keeps all ten of its mentions: it exists to explain tailscale, headscale and offscale to somebody choosing between them, on a core install, before any plugin exists. That is machine setup's subject, not the plugin's domain leaking upward. What survives elsewhere is only dated history — "was here until 2026-08-15", "It was: music, vault, slskd, headscale, …" — plus the notes about the real `headscale_servers` orphan and the secret-store key removed yesterday. Those record that something called headscale used to be here, which is a different claim from explaining what it does. tsgo clean, frontend builds, 873 pass / 7 fail unchanged, setup scripts parse. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> |
||
|
pastilhasandClaude Opus 5
|
e0d3f59b9b |
name the tables nothing on this machine owns
The marketplace screen now lists tables in the database that no code here claims, with a row estimate and, where the prefix says so, which absent plugin they look like they belong to. This is the information that cost an evening. `bun db:push` saw tables the schema did not declare, asked an interactive question no non-TTY caller could answer, applied nothing and exited 0 — and the install reported success while the plugin was unusable. The cause was two tables from before an extraction, found by reading `\dt` by hand. Nothing in the platform would ever have mentioned them. ── What counts as an owner ── The core schema, or a plugin DIRECTORY — not a plugin INSTALL. A plugin that is present but uninstalled still owns its tables, because that is the promise uninstall makes: keep every row so a reinstall puts you back where you were. So this reports only what no code on this machine describes at all, which on a healthy install is nothing, and the block does not render. A plugin whose schema will not import has its prefix treated as claimed rather than having every one of its tables reported as unowned — a broken manifest should not read as data with no home. ── No delete button, deliberately ── Dropping a table is not something to offer beside a list of plugins, and the right action differs per row: `music_favorites` wants the music plugin installed, `headscale_servers` wants dropping. Naming them is the part that was missing. Verified by planting `headscale_servers` and `music_playlists_old` and reading them back through the same call the route makes: the first attributed to nothing, the second to music, which is not on this machine. Row counts come from `pg_class.reltuples` — an exact count on a large table is not worth blocking a page render for a number that only answers "is there anything in here". tsgo clean, frontend builds, 873 pass / 7 fail unchanged. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> |
||
|
pastilhasandClaude Opus 5
|
c98b5643aa |
show what the commands are actually saying
Install, uninstall and purge narrated what the platform was doing and hid what it was running. Every spawn captured stdout and stderr, held them, and surfaced the last 600 characters only if the exit code was non-zero. That is enough to explain a crash and useless for the two things people want while watching an install: knowing a slow step is still moving, and seeing the warning a command prints on its way to exiting 0. The second is not hypothetical — `db:push` reported success on a machine where it created no tables. Its output was captured, discarded because the exit code was fine, and the failure surfaced minutes later as 42P01 in a sidecar log. Whatever drizzle said about that push, nobody ever read it. `runLogged` streams both pipes as they arrive, so they interleave the way they do in a terminal. Following only stdout would have been silent for exactly the commands worth watching: git writes everything to stderr, drizzle-kit writes progress there and results to stdout. Two pieces of noise control, because a log nobody can read is not an improvement: `git clone` gets `--progress` instead of `--quiet` (git says nothing when it thinks it has no terminal, and that is the step where a slow link looks like a hang) with per-percent rewrites collapsed to the last fragment per chunk; and drizzle's spinner, which is eight lines saying the same thing with a different braille glyph, collapses to the first — otherwise "Changes applied" scrolls past underneath it. Steps and output share ONE ordered list rather than two. Their order relative to each other is the whole information: "which step was this output from" is the only question the log exists to answer. Output renders dimmer and indented under the step that produced it. Verified by running a real purge and reinstall: the clone's object counts, the barrel generation naming both plugins, drizzle's push, then "schema: applied in 2029ms (1 table verified)". tsgo clean, frontend builds, 873 pass / 7 fail unchanged. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> |
||
|
pastilhasandClaude Opus 5
|
3c03e61c5e |
delete a plugin and everything it owns, and stop trusting exit codes
Two halves of the same problem: a plugin installed onto a fresh machine, reported success at every step, mounted, started its sidecar, registered — and then failed its first query with 42P01, because `offscale_servers` had never been created. Nothing said so, and there was no way back to a clean machine to try again. ── The install now asks Postgres ── `pushSchema` only checked the exit code, and `db:push` is a compound command: generate the barrel, cd, run drizzle-kit. It can exit 0 having created nothing. So the install now reads the plugin's schema for the tables it declares and verifies they exist, failing the install with a message naming what is missing instead of leaving a stack trace to appear in a sidecar log minutes later. Tables that are NOT prefixed with the app name are reported as a warning rather than a failure — they still get created, but `purge` will not remove them and a reinstall would find them already there. ── `purge`, next to an uninstall that deliberately keeps everything ── Uninstall is careful on purpose: every table and row survives so a reinstall puts you back where you were. That is what makes it safe to try, and it left no way to get to a CLEAN machine. A second install finds the directory (so the catalogue hides the plugin) and the tables (so push has nothing to do, and a broken schema stays broken). So a different verb, a different word, its own confirmation. It reads the table list from the schema BEFORE deleting the source, uninstalls (stopping the sidecar — nothing may be running against tables about to vanish), drops, deletes the directory, then regenerates the barrel unconditionally, since a barrel naming a directory that no longer exists is what stopped a server booting yesterday. The table list comes from the plugin's own schema, not from `like '<app>_%'` — that query eventually drops somebody else's table, because the prefix is a convention nothing validates and `music` is a prefix of `music-videos`. The prefix is used as a GUARD instead: a schema module could export `users`, and anything outside the plugin's own namespace is refused rather than dropped. ── Verified as a full loop on this machine ── purge offscale → table dropped, directory deleted, barrel regenerated → the marketplace offers it again → install → cloned, ssh checked, "schema: applied in 2072ms (1 table verified)", dock pinned, mounted, sidecar started. I left offscale INSTALLED here rather than restoring the half-state it was in (no install row, table present, sidecar running). tsgo clean, frontend builds, 873 pass / 7 fail unchanged. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> |
||
|
pastilhasandClaude Opus 5
|
ba20155a09 |
the store's README shows in the platform, not just on the website
A plugin's page in the Marketplace screen now renders the README the marketplace holds — the same text the public site shows, in the pane where somebody actually decides whether to install it. Fetched by the SERVER and proxied, though CORS would have allowed the browser to ask directly. An Officer client is routinely somewhere its server is not: a phone on the tailnet, a laptop with the store blocked. The server already had to reach the marketplace for the catalogue, so routing through it means the README appears in exactly the cases the plugin list did — rather than in a strictly smaller set, which is the kind of difference nobody reports and everybody notices. Separate request from the list, because the list endpoint deliberately carries no READMEs: three of them is ~38KB of markdown on the request that draws the tiles. This one fires when a plugin is opened. Absence renders nothing at all — no error, no empty state. A plugin without a README is normal and a store that is down must not put a failure message on a page whose job is installing things. Under `/plugins` so the existing permission covers it: reading the store and installing from it are the same privilege, and a second key would be a second thing to grant for no separate decision. tsgo clean, frontend builds, 873 pass / 7 fail unchanged. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> |
||
|
pastilhasandClaude Opus 5
|
bf44a4cfab |
resolving a member's task symlink needs their identity too
One layer below yesterday's fix, and it looked identical on screen: the job was
found, then declared to have written nothing.
`loadBackgroundTask` resolved the located path with `realpathSync`, as the
platform. Resolving a link needs `x` on every directory above it, and
`/tmp/claude-<uid>` is `drwx------` — so it threw EACCES on an INTACT link, and
the catch below it reported that as a dangling symlink and returned null. The
tray polls, gets `{kind:'pending'}`, and shows "Nothing written yet" forever.
The lookup fix could not have exposed this: it made the file findable, which is
precisely what let execution reach the line that then threw it away.
`realpath -e` rather than `readlink -f`, because `-e` requires every component to
exist — that keeps a genuinely dangling link reporting as one instead of
resolving to a path that is not there. Verified against symlink, plain file,
dangling link and missing path: `realpath -e` and `realpathSync` agree on all
four, so the owner path is unchanged.
Audited the rest of claude-sessions.ts for the same shape. Every remaining
mention of `statSync`/`readdirSync`/`Bun.Glob` in that file is a comment
explaining why it is NOT used.
Live coverage extended: the member resolves their own link, the platform gets
null. Skips without OFFICER_TEST_MEMBER.
tsgo clean, frontend builds, 865 pass / 7 fail unchanged, 7 skips.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
|
||
|
pastilhasandClaude Opus 5
|
6bee6b856e |
a member's background job can be read and stopped
A member started `bun dev`, watched it run, and got no logs at all — then
`{"stopped":false,"reason":"no such task"}` from the Stop button. Two symptoms,
one cause, and it is the same one `read-as-user.ts` was written for.
`findTaskOutput` globbed `/tmp/claude-*/…` as the platform's own uid.
`/tmp/claude-<uid>` is `drwx------` (checked, not assumed), so for a member that
directory cannot even be traversed. A glob that cannot traverse does not throw —
it returns an empty iterator. Both callers read that as "no such task": the log
view showed "Nothing written yet" forever, and Stop refused an id that was
perfectly real.
Every other filesystem access in that module already goes through the as-user
helpers. This one path never did, and it was invisible for as long as only the
owner used chat.
So the lookup moves into `findPathAs`, and Stop runs as the member too — both
halves of it. `fuser` reads `/proc/<pid>/fd`, which only the process's own user
and root may open, so as the platform it would have found no pids and reported
the job already stopped; the kill after it would have been cross-user EPERM,
swallowed by a catch. Neither would have said anything.
── The exit code is ignored, and that is the point ──
The first version of this fix located the file and threw it away. `find` exits 1
when it cannot descend into something, and searching `/tmp` from ANY identity
hits another account's `drwx------` — it prints the match on stdout first, then
returns 1. Treating that as failure reproduced the original bug through a new
mechanism. Caught by testing the owner path against a real task file rather than
assuming the refactor was behaviour-preserving; a missing root still exits 1 with
empty stdout, so absence still reads as absence.
Owner and member now take the same `find`, deliberately: one set of matching
semantics, rather than a glob for us and a find for them, which is how a bug ends
up reproducing on one account and not the other.
Verified on the owner path here against a live task file. The member path is
covered by `read-as-user.live.test.ts` — same opt-in shape as
`spawn-as-member.live.test.ts`, asserting the two identities DISAGREE, which a
world-readable tree would not satisfy. It cannot run on a machine with one
account, so it skips here and needs OFFICER_TEST_MEMBER on the host.
tsgo clean, frontend builds, 865 pass / 7 fail unchanged, 3 new skips.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
|
||
|
pastilhasandClaude Opus 5
|
048e718b21 |
a stop button on running background jobs
Killing a job meant asking the agent in words — a whole turn to say "please stop
that", and only possible while the agent is responsive. Now it is a button on the
chip.
── Finding the process ──
The platform does not own these processes. Claude Code starts them and nothing
records a pid. What it has is the file the job streams into, and the job is still
holding that file open — so `fuser` gives the pid, and the file is resolved
through the SAME scoped lookup the read path already uses: the id is validated
against TASK_ID_RE, then matched only under this caller's own directories. An id
that resolves to nothing kills nothing.
Verified end to end rather than reasoned about: a stand-in job holding its output
file open, `fuser` finding it, `{"stopped":true}`, process gone.
SIGTERM, not SIGKILL — a dev server with children should get the chance to take
them down with it. SIGKILL orphans them and leaves the port bound, which is worse
than a stop that takes a second.
── Two things it deliberately does not do ──
It does not dismiss the chip or write a `stopped` status locally. The real end
arrives the way every other end does — the harness sees the exit and the
transcript gets the row — so faking it would be a guess that usually agrees with
reality and sometimes does not. A stop that fails leaves the chip running, which
is the truth.
It does not offer to stop a SUBAGENT. Those write a `.jsonl` transcript rather
than holding a log open, so there is no fd to trace; it says so instead of
pretending.
`Square`, not `X`. X on the same chip means "close this", and confusing "hide the
chip" with "kill the process" is the one mistake worth designing out — which is
why stop appears only on running tasks and dismiss only on finished ones.
tsgo clean, frontend builds, 851 pass / 7 fail unchanged.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
|
||
|
pastilhasandClaude Opus 5
|
5c40d4423d |
app preview: a panel that points an iframe at a running app
A dashboard can now hold chat, the code editor, the file browser and the app itself — the last piece of working on a project without leaving Officer. ── Why this is ~200 lines and not 560 ── The version deleted on 2026-07-31 spawned `bun dev` in a project directory, found a free port and proxied it under `/api/dev-server-proxy/<id>/`. That prefix is what cost the money: HTML, JS and CSS rewritten; a script injected into the page to monkey-patch fetch/XHR/WebSocket so runtime URLs got the prefix too; a Referer fallback for chunks that resolved at the wrong depth; a websocket bridge for HMR. All of it faking an origin. On a machine with a proxy in front — nginx on edge.pertento.ai, wildcard subdomains — the app is ALREADY served at the root of its own origin, and the browser is already on the tailnet. So there is nothing to fake and no server side at all: no route, no socket, and therefore no permission-registry entry. The old code is not obsolete and is not being thrown away. An install with no proxy in front still has an app on localhost that the browser cannot reach, and that is exactly what it solves. It is at `188b45c1^`; the note is in TODO.md. ── What it does ── Holds a URL per panel, renders it, reloads it, opens it in a real tab. The URL lives in dashboard state so it survives a reload, and is shared LIVE between header and body — the header writes and the body reads, so claim-once would mean typing a URL changed the header and nothing else. The reload counter is deliberately not persisted: it is a signal. Reload remounts the iframe by key rather than calling `contentWindow.location .reload()`, which throws cross-origin. Replacing the element re-requests the URL, which is the same thing from the outside and the only move the parent is allowed. `allow-same-origin` is set and has to be: without it the framed app gets an opaque origin and no localStorage, and it authenticates with a JWT it keeps there. `normalizeUrl` accepts a bare host as https and refuses javascript:, data: and blob: — an iframe src with `javascript:` executes in the PARENT's context. The owner typing in their own panel is not an attacker; it costs one line and removes the category. 5 tests. ── Two iframe facts worth knowing before testing ── Storage is PARTITIONED by top-level site, so the login inside the panel is separate from the login in a tab: expect to sign in once, in the panel. A page can refuse to be framed (X-Frame-Options, CSP frame-ancestors) and renders blank with only a console error. If that happens it is nginx's config, not this code — which is why "open in a new tab" is in the header rather than left to copying the URL out. tsgo clean, frontend builds, 851 pass / 7 fail — +5 new, same 7. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> |
||
|
pastilhasandClaude Opus 5
|
465435ccf1 |
a chat panel picks from the conversations in its dashboard's directory
A dashboard rooted in ~/Tests/platform now has a session picker in its chat panel's header: every conversation started in that directory, newest first, plus New chat, rename and delete. ── Almost all of this already existed ── `GET /chat/sessions?cwd=` returns conversations for a working directory, merged across Claude and OpenCode. It is not a registry Officer maintains: Claude writes 每 transcript to `~/.claude/projects/<cwd-slug>/`, so "the conversations belonging to this dashboard" is how they already sit on disk — the same ones the terminal `claude` sees. `useClaudeSessions` already wrapped list, rename and delete for the /chat screen. The panel header is new; nothing under it is. ── The one design correction ── Which session a panel is on is NOT `usePanelMemory`. That hook claims its saved value once and then ignores the bag, because the editor and file browser are the only writers of their own state and hearing it echo back would tell them to reopen what they just closed. Here the header writes and the body reads — two components with no props between them — so the value has to stay LIVE. Claim-once would have meant picking a session changed the label and nothing else. It is the same dashboard-state bag, read live, so it still survives a reload. Switching remounts the chat on its key, which is what the existing comment calls "the honest way to say this is now a different conversation": `useChat` seeds from the first `initialSessionId` it sees and the resume handshake sends whatever is in that state on connect. A NEW chat is `null` until the harness names it — only the harness can — and the id is recorded on arrival, so "new chat, say something, reload" returns to that conversation rather than a third one. Deleting the conversation on screen points the panel at a new one rather than at a transcript that no longer exists. ── Left alone, and one thing to know ── `agent_panels` is untouched: a named agent still wins the session choice outright, since it is one conversation forever by definition. The picker is for the ordinary case. The email screen mounts ChatPanelWrapper too, so its chat now also remembers its session instead of starting fresh each visit. That is a behaviour change to a screen this work was not aimed at — better in my view, but say if not, and it scopes to dashboards in one line. tsgo clean, frontend builds, 846 pass / 7 fail unchanged. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> |
||
|
pastilhasandClaude Opus 5
|
74495c25ab |
a file browser panel reopens where you left it
Navigate into a subfolder of a dashboard's directory, refresh, and you were back
at the top. The /files screen keeps its folder in the URL and never had this; a
panel had nowhere to put it.
── Made shared, because chat is next ──
The editor's tab memory from the last commit was one panel's version of a pattern
the terminal had already invented. Rather than write a third, `usePanelMemory` now
holds the mechanics — claim once, key per panel, drop on real close — and each
panel supplies only a payload. The editor moved onto it and lost 50 lines; the
file browser stores `{ path }`; chat will store whatever chat needs.
Restoring is SYNCHRONOUS, in the initial state rather than an effect, and can be:
`WorkspaceView` renders nothing until dashboard state has loaded, so the
remembered folder is already there on the first render. An effect would list the
base folder first and visibly jump to the right one.
── Two ways it could strand you, both closed ──
A remembered folder OUTSIDE the panel's base is discarded on read. A dashboard
scoped to ~/Tests must not reopen at ~/other because it was rescoped or the folder
moved. The existing guard catches that after the fact; this stops it being
requested.
A remembered folder that no longer EXISTS was the new failure this feature
introduced — the listing throws, a toast appears, and every reload lands in the
same dead folder, because that is what was remembered. The first listing now falls
back to the base and rewrites the memory. Deliberately first-listing-only: later
in a session a failure is far more likely transient, and bouncing someone out of a
deep folder over one bad response is worse than the error they can already read.
`files` registered in PANEL_STATE_PREFIXES, without which the PATCH is a 400.
tsgo clean, frontend builds, 846 pass / 7 fail unchanged.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
|
||
|
pastilhasandClaude Opus 5
|
95b83fb29b |
a code editor panel remembers its tabs across a reload
Refreshing the page emptied it. You came back to "Open a file from the explorer to
start editing" with no trace of what you had been doing — two tabs open, one of
them the notes you were writing in.
The terminal solved this shape years ago: a `panelId -> sessionId` map kept in
dashboard state, keyed `ws-{prefix}-{dashboardId}`. This is that map with a
different payload, deliberately the same shape so there is one pattern rather than
two to compare. `editor` is registered in PANEL_STATE_PREFIXES — an unmatched key
is a 400, not a silent 200, which is the branch that exists precisely so a new
prefix cannot vanish quietly.
── Paths, not contents ──
The disk is the truth. Restoring a remembered BODY would show a version that may
be hours stale, or resurrect a file something has since deleted. Each path is read
fresh on restore, in the order you left them, and one that no longer reads is
dropped rather than opened as an empty buffer named after a missing file.
The cost, stated rather than hidden: unsaved edits still do not survive a reload.
They did not before either — the whole tab went. Keeping them means persisting
buffers, which has its own hard question (the disk AND the buffer have both
changed — now what?) and is not smuggled in here.
── The two ways this goes wrong, both guarded ──
Reading the saved set more than once would fight the user: every tab they open
writes into that same bag, so a later read arrives as an update telling them to
reopen what they just closed. It is claimed once, on the first render that has it.
Writing before restoring would persist an empty list over the state it is about to
read — first render, no tabs yet. The write is guarded on the restore having run.
Also: a panel closed for good drops its entry, via `usePanelClose`, which fires on
a real close and not on a drag, a swap or a mobile panel switch.
Per PANEL, not per dashboard, so two editors side by side keep separate strips.
Not wired on the /code-editor screen: `?file=` already survives a reload there.
tsgo clean, frontend builds, 846 pass / 7 fail unchanged.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
|
||
|
pastilhasandClaude Opus 5
|
a84eedd126 |
zip from the folder, extract on double-click, show location
Three things from testing. ── The download zip carried the path to itself ── Selecting files in ~/Tests/test_folder and downloading produced an archive whose contents were `Tests/test_folder/…`. `POST /download` spawned zip with `cwd: rootDir` and root-relative paths, so the whole chain from home went in. Nobody selecting three files is asking for the route to that folder to be part of what they downloaded. Now it zips from the items' own folder using bare names. A MIXED selection keeps the old form, because two `notes.txt` from different folders would collide without the paths keeping them apart — the UI only selects within one folder, so that branch is a safety net rather than the case. `/compress` already did it this way; this is the same fix on the endpoint that predates it. Verified against real zip: `Tests/a.txt` → `a.txt`, `Tests/test_folder/b.txt` → `test_folder/b.txt`. ── Double-clicking an archive extracts it ── It opened a pane that rendered "Archive file" and the name you could already see in the listing. That is a step on the way to the only thing anyone opens a zip for. Extract resolves collisions, so a second double-click makes a sibling rather than overwriting. The menu's Extract stays — a double-click is not a discoverable way to learn what will happen. ── Show location ── A search result answers "here is the file"; its path underneath raises "where does this live", and clicking only ever answered the first. This goes to the containing folder, clears the search and SELECTS the row, so a folder of forty files does not leave you hunting. A directory hit reveals its parent — where is this means the folder it sits in, not the folder it is. tsgo clean, frontend builds, 817 pass / 7 fail unchanged. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> |
||
|
pastilhasandClaude Opus 5
|
cc889a864c |
file browser: close the accidental gaps in the context menu
The audit split the menu's omissions in two. Three were deliberate — OCR, Transcribe and Extract Audio were removed in July because the task system does them better, and those stay gone. The rest were nobody's decision. This is the rest. ── One menu body instead of two ── `DropdownMenuItems` and `ContextMenuItems` were character-identical apart from the component prefix: 159 lines duplicated, including the same comment twice. Radix gives both families the same props, so the components are now a parameter and there is one list. That is why this diff removes more than it adds while adding seven items — and why the next item only has to be written once. ── Added to the row menu ── - Open — double-click was the only way. Every file manager puts it first. - Open in editor — `/code-editor` is on the SAME `files` permission as `/files`, so anyone who can browse could already edit; there was just no way to get there from a file. Narrower than Read Aloud on purpose: that one uses the `text` FALLBACK type, which matches .exe and .bin. - Paste into folder — a move was only expressible as cut → navigate → paste. Right-clicking the destination is the obvious gesture and did not exist. - Duplicate — one call away the whole time; `/copy` already resolves collisions to " (copy 2)". - Compress — the missing half of Extract. `/download` shelled out to zip but streamed the result away, so nothing could make an archive and keep it. New `POST /file-browser/compress` writes it beside the original. - Pin / Unpin — pinning existed only in the widget, which is not where you meet a file worth keeping to hand. ── Added to the background menu ── - Clone repository and Show/Hide hidden files. Both were toolbar-only, and the toolbar hides them under `md:` — so on a phone neither existed. All four desktop-only toolbar actions are now in this menu, which long-press reaches. - Paste is gated on the clipboard, as both toolbars already were. Ungated it fell through to the OS clipboard path and either toasted a secure-context error or silently did nothing — an item that looked available and was not. ── Removed ── `pcm-worklet-processor.js`, orphaned when cliamp left for the music plugin this morning. Verified unreferenced; the plugin ships its own copy inline. The `Play` hole itself is NOT closed and cannot be from here: there is no way for a plugin to register a context-menu item. Tasks and agents are the only extension mechanism and a task can only contribute a task entry. That is a platform gap and wants its own decision. Verified the zip invocation by hand for all three shapes — single file, directory (recursive, contents included), and a multi-item selection. Archive naming strips the extension only for a file, so a directory called `my.photos` stays `my.photos.zip` rather than becoming `my.zip`. tsgo clean, frontend builds, 808 pass / 7 fail unchanged. SelectionActions.tsx and usePipelineRunner.ts are prettier reflowing long lines — no semantic change, caught by formatting the folder. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> |
||
|
pastilhasandClaude Opus 5
|
dd26e4a688 |
example leaves too; plugins/ is documentation now
Done with `tea` as admin: plugins/offscale made public, plugins/example created public, both verified anonymously over https — which is the check that matters, because fetch.ts clones with no credentials and a private plugin simply cannot be installed from the marketplace. plugins/example 8 files, cloned back and diffed identical plugins/offscale now public (was private, so was uninstallable) `.gitignore` is one line now — `/plugins/*/`. The trailing slash is the whole design: it ignores DIRECTORIES, so files at the top of plugins/ stay tracked. The documentation about plugins belongs to the platform; the plugins do not. All three are in the marketplace catalogue, which is what makes this reversible: before this, untracking offscale meant a fresh install could never get it back. Now the store is the way in for all three. CatalogueEntry gains `icon` (a lucide name) alongside `iconUrl`. Both new entries ship glyphs rather than artwork, and without it they drew the generic puzzle piece in the store and their real icon after install — which reads as the icon changing rather than the store not knowing it. EXTRACTING-A-PLUGIN.md pointed at four directories a fresh checkout no longer has. It now gives repository URLs, says up front that this directory holds documentation rather than plugins, and carries a table of where each one went. Checked first that no tracked source statically imports a plugin directory — none does, so a fresh clone still builds. tsgo clean, frontend builds, 787 pass / 7 fail (unchanged). Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> |
||
|
pastilhasandClaude Opus 5
|
29170dd922 |
installing a plugin pins its dock tile, with no reload
Installing Music left the dock unchanged. The tile existed server-side the moment the plugin was enabled, and the only way to see it was Profile → Dock, by hand, after a page refresh. Installing something is already the decision to use it; saying so twice is the bug. Three separate causes, which is why it looked like one stubborn one: 1. `dock_configs.paths` is a COMPLETE ordered list, not a diff — useDock renders the saved list rather than everything available, so anything not named in it is invisible. Nothing in the plugin system had ever written to that table. installPlugin now appends the new route, for the caller (threaded through the route) rather than a guessed owner id. Only for a user who HAS a row. A user without one is left alone deliberately: there is nothing to append to, and writing one would freeze today's defaults into their account permanently. That case is (2). 2. The fallback for a user who never customised was DEFAULT_DOCK_PATHS, which is core-only — so a plugin tile was hidden there too. `defaultDockPaths(plugins)` is now that list plus a tile per installed plugin, and the dock and its settings screen both call it so they cannot disagree about what "default" means. 3. useDock holds ['DOCK'] at staleTime: Infinity and the plugin verbs invalidated only ['plugins'] and ['self-permissions']. So even once the server wrote the path, nothing asked again — that is the part that made a reload necessary. Also fixes a guard I broke reaching for this: typing VERBS as Record<string, VerbFn> widens keyof to `string`, which makes `isVerb` prove nothing and lets an unknown verb index the map. `satisfies` instead, and there is now a check that Verb is still the four-member union. Uninstall deliberately does not remove the path. A stale path with no item behind it is dropped by useDock, and keeping it means disable/enable and reinstall put the tile back where the user had it rather than at the end. tsgo clean, frontend builds. 787 pass / 7 fail — 6 identical to before; the 7th is plugins/music/cliamp, which install pulled into test discovery and which the manifest documents as parked. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> |
||
|
pastilhasandClaude Opus 5
|
dadfdc26af |
plugins page becomes a marketplace: catalogue, clone-on-install
The /plugins route showed only what was already on disk, so a plugin nobody had copied in by hand did not exist. It now shows a third list — what a remote catalogue offers — and installing one fetches its source. - plugins/marketplace.ts — a stand-in for the remote store, a JSON array in process. One entry: music, pointing at gitea.officer.dev/plugins/music. It is deliberately a module and not a fetch, so the shape of a catalogue entry gets settled before anything depends on a server answering. - plugins/fetch.ts — git clone --depth 1 into plugins/<appName>/, INSIDE the platform repo because bun resolves the workspace links by walking up and a plugin cloned beside it cannot import anything the platform provides (measured, not assumed). https only, no credentials, GIT_TERMINAL_PROMPT=0 so a repo that turned private fails instead of hanging. A failed clone removes the directory: a half-written one reads as a broken plugin rather than an absent one, which is the worse of the two. - install.ts fetches first when the directory is missing, then RE-DISCOVERS rather than assuming what it cloned. Everything after is the existing path, including the frontend rebuild — so a cloned plugin's panels reach the SPA without a restart. - GET /api/plugins returns `available`, catalogue minus what is on disk. Each entry gets the same live PATH probe a local plugin gets, so music's ffmpeg and ffprobe are answered for this machine BEFORE the clone rather than after. - The detail panel is a separate component for a catalogue entry, not a mode of the installed one: there is no sidecar status, no enabled state and no route to open, and what there IS — the source URL — the installed panel has no reason to show. The streaming log is shared, because the marketplace install is the case that needs it most: it clones before it does anything, and a slow clone behind a blank panel looks like a hang. The catalogue's icon is an absolute URL to the store's copy, which is the [open] question the previous commit left: assets only publish at install, so a plugin you have never installed had no artwork to draw. Not done: no update verb, no signature, no private repos, and removePluginSource has no caller — deleting source drops tables on the next push, and that wants a verb that says so. tsgo clean. 767 pass / 6 fail, all 6 identical on HEAD without this change. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> |
||
|
pastilhas
|
027b10bd6e |
step 4/4: the docs say permissions too, and capability means one thing again
44 files of prose — CLAUDE.md, AGENTS.md, TODO.md, 20 docs, both plugin design
documents, and the comment surface the earlier steps could not reach.
Applied against an explicit keep-list, not swept, because the word turned out to
have SIX meanings in this repository rather than the three the offscale doc
recorded:
permissions renamed (steps 1–2)
$OFFICER_ROOT/capabilities/ KEPT — the item store, and now the only thing
the word means that is ours
sidecar routing keys renamed to `handles` (step 3)
Lightning wallet KEPT — a domain term, and on the wire to the mobile apps
terminfo queries KEPT — XTGETTCAP, in the pty sidecar
InvoiceShelf KEPT — per-resource { write, bulkDelete } flags
The sweep still falsified two things, both caught by checking rather than by
review, and both in prose that discusses more than one meaning at once:
CLAUDE.md began claiming the item store lives at `$OFFICER_ROOT/permissions`.
It does not; that directory is on disk and full of skills and tools.
And the offscale doc's own note about the collision became
"Named `permissions`, NOT `permissions`" — a sentence that had eaten the thing
it existed to warn about.
Both restored, and the note rewritten to say what is now true: capability means
one thing of ours, and three that belong to somebody else's vocabulary.
Verified live after restart: self and admin permission endpoints 200, gated
route 200, agent-status 200, 9 grants intact with 6 permissions offered.
tsgo clean, 797 tests, 787 pass, same 7.
The rename is done. Four steps, no data lost, no client break that survived
the step it was introduced in.
|
||
|
pastilhas
|
5ec354cfcb |
step 2/4: role_capabilities becomes role_permissions, without losing a grant
The dangerous one. `drizzle-kit push` does not understand renames — it sees a
table gone and a table added, and with --force it resolves that by dropping and
creating. The 9 live grants would have vanished silently and every member would
have lost terminal, chat and files until someone noticed and re-granted by hand.
So the database moved FIRST, by explicit ALTER, and the code followed:
ALTER TABLE role_capabilities RENAME TO role_permissions
ALTER TABLE role_permissions RENAME COLUMN capability TO permission
ALTER INDEX uq_role_capabilities_role_capability RENAME TO uq_role_permissions_role_permission
ALTER TABLE ... RENAME CONSTRAINT ck_role_capabilities_{role,level,not_owner} TO ck_role_permissions_*
ALTER INDEX role_capabilities_pkey RENAME TO role_permissions_pkey
Seven objects, not one, and all in a single transaction: a partial rename would
leave drizzle-kit seeing a table it half-recognised, which is the same drop.
The names were read out of pg_indexes and pg_constraint rather than assumed.
`bun db:push` then reported **No changes detected** — which is the whole proof.
It means the ALTERs matched the schema code exactly, so there was nothing for
push to reconcile and nothing for --force to destroy.
Grants verified against a JSON backup taken before the first ALTER: 9 rows,
byte-identical, `role_capabilities` gone from the database.
No client noticed. Step 1 had already moved the wire to `permission` by mapping
in the route while the column still said `capability`, so this step deleted the
mapping rather than changing any response. The two call sites that carried a
`capability` key with a comment explaining why now say `permission` and the
comments are gone.
Also renamed officer_db/src/capabilities/ → permissions/, roleCapabilities →
rolePermissions, CapabilityLevelValue → PermissionLevelValue.
Verified live after restart: self 200, admin list 200, gated route 200, no
token 401, grants read back as {role, permission, level}, and a PUT round trip
left 9 grants intact.
tsgo clean. 797 tests, 787 pass, same 7.
|
||
|
pastilhas
|
2e8ec845c8 |
step 1/4: the permission engine is called permissions, not capabilities
The word meant four different things in this repo, not the three the offscale
doc records:
1. the permission registry → RENAMED here
2. $OFFICER_ROOT/capabilities/ items → kept; this is what capabilities are
3. sidecar routing keys → step 3, becoming `handles`
4. Lightning wallet features → kept; a domain term, and on the wire
to the mobile apps
The fourth was not in the doc and a global find-and-replace would have broken
the mobile wallet, which reads `{ kind, capabilities: Capability[] }` from the
wallet sidecar. So this renamed against an explicit file allowlist rather than
by sweeping the tree, and `CapabilityPage.tsx` — the UI for the item store, and
correctly named already — was left alone.
Moved: servers/capabilities/ → servers/permissions/, capability-gate.ts →
permission-gate.ts, users/capabilities-routes.ts → permissions-routes.ts,
hooks/useCapabilities.ts → usePermissions.ts. Identifiers follow.
Three breaks the typechecker could not see, all found by exercising it live.
The route paths moved with the prose sweep, so the server served
/user/permissions while the frontend still called /user/capabilities. A 404 on
every page load, and tsgo clean throughout.
The response FIELD moved too. `client.get<SelfPermissions>()` is an unchecked
cast, so `data.capabilities` became `undefined` at runtime with no compile
error — `can()` would have answered "no" to everything and the dock would have
emptied itself.
And the grants list was passed straight out of the database, so it arrived as
`{ role, capability, level }` while the screen read `grant.permission`. Every
role would have rendered as holding nothing. It is now mapped in the route:
the wire says `permission`, the column still says `capability`, and step 2
therefore changes nothing any client can see.
The stale react-query keys were the quiet one: two files still invalidated
['self-capabilities'] after the hook moved to ['self-permissions'], so
installing a plugin would have silently stopped refreshing the dock.
The database is untouched — `role_capabilities` and its `capability` column are
step 2, and the two call sites that cross that boundary say so in a comment.
Round-tripped the 9 live grants through the admin endpoint to prove the PUT
contract survived: 9 before, 9 after, Member's three intact.
Also reverted prettier churn on five landing-page files that a broad --write
picked up. Second time today; the lesson is not sticking.
tsgo clean. 797 tests, 787 pass, same 7 pre-existing failures — two of which
now read "path → permission" rather than "path → capability".
|
||
|
pastilhas
|
f78abbe05a |
a plugin ships its own dock icon as a file, not a lucide name
The manifest's `icon` was a lucide NAME, resolved by `resolveIcon` — which knows 106 glyphs out of lucide's ~1,500. A plugin naming one outside that set silently rendered a neutral box, and a plugin from a marketplace had no way to see the ceiling coming. So the icon is a FILE now: `plugins/<name>/assets/icon.png`, discovered by presence like everything else here. `manifest.icon` stays as an optional fallback for a plugin with no artwork — example and offscale still use it — and the file wins when both exist. No new mechanism was needed. The app store already published sidecar assets: `<dir>/assets/` → `public/plugins/<id>/`, served by a dynamic `/plugins/*` route, with `DockItem.image` rendering an <img>. `pluginDockManifests()` simply never emitted `image`. Install now publishes and uninstall unpublishes — the one thing uninstall is allowed to delete, because these are copies whose originals are still in the plugin's source. Base64 in the manifest was considered and dropped. It would ride in every /api/user/capabilities response for every user on every page load, can't be cached separately, and puts a 5KB string literal in a source file — against the rule this manifest keeps: what a directory listing can say, it says. It also needs no support: `image` goes straight into <img src>, so a data: URL already works for anyone who wants one. Music ships OffMusic.png, resized 1254² → 256² (1.6MB → 108KB) with alpha intact, sized for 3× DPI at the dock's 32px render. It also drops `icon` — the artwork is a voxel duck in headphones, not a glyph. The plugins page showed a generic Puzzle for every plugin; the list and detail header now show the plugin's own icon when it has one. Two bugs found while testing. Dock.tsx imported 26 lucide icons and used 14. The twelve dead ones — Music, Bitcoin, Receipt, Images, CalendarDays, Contact, Clapperboard, Mail, Network, ArrowDownUp, FileText, FolderKanban, MonitorSmartphone — were residue from when every feature had a hardcoded tile. Deleted. And uninstalling a plugin left its icon URL answering 500, not 404. server.tsx globs ./public at BOOT into one exact route per file, each holding a Bun.file handle, spread into the route table AHEAD of the /plugins/* wildcard. So an icon present at boot got an exact route that outlived the file, returning ENOENT on every dock render with the error logged each time — exactly what the wildcard's own comment says it exists to prevent. The comment covered the ADD case; this is its mirror. `plugins/` is now excluded from the boot glob, so the wildcard owns that prefix alone. Verified: 200 installed, 404 uninstalled, 200 reinstalled. [open] A plugin's icon cannot be seen BEFORE installing it, which is the one place an app store most wants to — assets are published at install by design, and an authenticated icon route is no use to an <img>. |
||
|
pastilhas
|
a9bf51407e |
cliamp moves into the plugin, and the platform loses its last music file
The owner read the code and asked why `plugins/music/api/router.ts` was three
lines importing `@@/api/music/router` — platform code that knows the string
'music'. He was right, and tracing it found the justification was hollow.
The chain: server.tsx:20 imported the cliamp relay's two exports, which are
used only on commented-out lines; so the relay's functions were never invoked;
so its call to getMusicServerWsUrl never ran; and the file's other export,
getMusicServerUrl, had no consumers at all. A dead import held a music-named
file in the platform, and I documented that as a "seam" last night after
checking the import existed and stopping there.
Everything cliamp now lives in plugins/music/cliamp/:
sidecar/music/{cliamp-ws,pulse-audio}.ts, asoundrc, the test
api/cliamp/relay.ts
apps/FileBrowser/{CliampPanel,AudioStreamPlayer}.tsx
src/servers/sidecar/music/, src/servers/api/cliamp/ and src/servers/api/music/
are gone. server.tsx has no cliamp import, provider name, handler entry or
route. The platform contains no file named for music or cliamp.
Two of the things that moved were live, not inert.
The file browser's `Play` context-menu item, on any audio file or folder, set
?play= and rendered a cliamp terminal pointed at /api/cliamp/ws — a route that
upgraded into a handlers entry that was commented out, so handlers[provider]!
asserted non-null on undefined. Using that menu item crashed the socket
handler. Removed: the action, the layout, the panel wiring and both menu
entries. Verified the routes now 404 rather than crash.
That closed the totality drift as a side effect. server.tsx's route table and
its handlers map agree again for the first time since 2026-08-13, and
registry.test.ts now asserts it rather than pinning the hole.
The proxy is built in the plugin now, and its prefix is DERIVED. It was the
literal '/api/music', which the proxy uses to strip characters off the path —
correct only because mountPrefix returns /music for a first-party publisher.
The same plugin published by anyone else mounts at /api/p/<publisher>/music and
would have forwarded /alice/music/stream to a sidecar expecting /stream. A
latent bug only third parties would ever hit, and a quiet violation of the rule
that mountPrefix is the one function allowed to know about provenance. Offscale
has the identical hardcode and still needs it.
Still open there: appName is passed as a literal, because a plugin's router
cannot see its own directory name — the platform imports the module and reads
`router`, so there is nowhere to inject it. The fix is a factory the installer
calls with the plugin's identity.
Plugin backend coupling is down to 7 imports, all of them "a plugin talks to
its host": data-path, sidecar/connect, sidecar/protocol, officer-url, the
manifest type, officerdb/db and the users.id FK. Nothing music-shaped left.
bunx tsgo clean. 797 tests, 787 pass, same 7 pre-existing failures. Verified
live: manifest 200, favorites 200, stream 206, /api/cliamp/ws 404.
|
||
|
pastilhas
|
e930586878 |
plugins declare the host binaries they need, and the installer checks
Offscale was self-sufficient. Music is not — it shells out to ffmpeg and ffprobe — and the way it fails without them is the reason this is a check rather than a line in a README. It does not fail. Missing ffprobe means the indexer catches the spawn error and returns a track carrying its filename and nothing else: no title, artist, album, duration or embedded lyrics. It then walks the whole library, writes a complete cache tree and reports success. Five swallowed catches, no log, no counter, and the only tell is coversSaved: 0 in a report nobody reads. So `osDependencies` is a manifest field: the binary to probe on PATH, why it is needed, and a package name per package manager. The shape is taken from scripts/setup-old/setup.sh rather than invented — probe the binary, case on $PM — and the names are per-manager rather than canonical-with-overrides because lib/packages.sh already recorded why that indirection was rejected. Probing the binary is what makes "built-in on this OS" free: on PATH means the package map is never consulted. Four decisions worth naming. Missing and uninstallable REFUSES the install, first, before a table is created or a row written — so there is nothing to undo, and the alternative is a plugin that installs, answers 200 and quietly produces nothing. The status is on GET /api/plugins and rendered before the button, because the owner is deciding whether to let the server run a package manager as root and that needs answering first. Installing by hand and watching it flip to present is the escape hatch on a machine without passwordless sudo. Package names get a deliberately narrow regex and reach Bun.spawn as an argv ARRAY, never a shell. Both halves are load-bearing: the regex means a metacharacter cannot get there, argv means it would be an argument rather than syntax if it did. Narrower than package managers actually accept — no `:`, no `+` version pins — because a plugin needing one wants a conversation. Success is OBSERVED, not inferred: after installing, the binaries are re-probed. A package manager exiting 0 having installed something that does not provide the binary is exactly the failure this exists to catch. installCommand mirrors lib/packages.sh's pkg_install_now exactly, including apt's non-interactive environment, so there is one definition of "install a package" rather than two that drift. sudo always gets -n: under PM2 a password prompt is not a slow path, it is a hang. brew never escalates. Verified live. ffmpeg and ffprobe were absent on this machine all evening; the page showed both missing with the exact root command, the install streamed `dependencies: installing ffmpeg with apt` then `ffprobe, ffmpeg now on PATH`, and X-Audio-Duration appeared on a stream response for the first time. The refusal path was exercised against a temporary probe dependency: HTTP 400, steps: [], reason named. THIS CHANGED THE MACHINE: ffmpeg 6.1.1-3ubuntu5 is now installed via apt. Found on the way: a manifest is read once per process. Discovery does `await import()` and the module cache holds it, so editing a manifest changes nothing until pm2 restart officer — including `outdated`. Cost ten minutes and is now in the runbook. bunx tsgo clean. 797 tests, 787 pass, 7 fail — the same seven, +25 new. |
||
|
pastilhasandClaude Opus 5
|
8b6cb34ae0 |
plugins contribute dock tiles, and the doc says what is actually built
offscale had no dock tile: that field comes from the app store's catalogue, so a plugin installed through the plugin system was reachable only by typing its url or following the link on /plugins. built at runtime rather than baked into the generated bundle, deliberately — WHO sees a tile is a permission question, and a grant takes effect on the next request rather than the next build. presentation comes from the manifest and the route from mountPrefix, so there is one source for both, and is the plugin's first permission so the endpoint can filter a tile out for an account that cannot reach the screen. two sources for tiles today, because the app store still has its own catalogue. one when it is rebuilt on this. the doc now records the system as complete rather than half-stubbed, including the three bugs the extraction found — the sidecar-before-mount ordering, the missing tailwind, and the build that could delete its own shell — and what is genuinely still open. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> |
||
|
pastilhasandClaude Opus 5
|
e13128846b |
offscale is a plugin
headscale leaves the platform. 45 files move to plugins/offscale/ and the
platform stops knowing it exists.
api/router.ts the thin auth-gated proxy, now at /api/offscale
sidecar/ 18 files, the whole headscale contract and its admin keys
db/ schema + queries, offscale_servers
web/ 26 files as panels and a layout — no screen, per the rule
removed from the platform: the hono mount, the `headscale` capability, the
App.tsx route pair, the screen and its barrel, the AppRegistry spread, the
officerdev re-exports, the dock tile, the page-title rule, and both database
barrels. tsgo is clean and nothing references it.
the imports tell the story of what the plugin↔host API actually is. the sidecar
takes @@/sidecar/protocol, @@/sidecar/connect, @@/data-path and
@@/officer-url.mjs; the queries take officerdb/db and officerdb/crypto; the
schema takes officerdb/auth/schema for the one reference a plugin may make; the
web half takes useClient, copyToClipboard, WorkspaceView and TerminalView from
the officerdev barrel. all of it resolves because a plugin lives inside the repo
— no publishing, no version negotiation.
AND IT FOUND A REAL BUG IN THE INSTALLER. createSidecarProxy learns its port
from a one-shot `<name>:server` event and subscribes when the plugin's router is
first imported — at mount. install started the sidecar BEFORE mounting, so the
announcement fired into a void: process online, routes mounted, every request
answering `503 sidecar not available` until something forced a reconnect. it
would have hit every plugin with an http sidecar. `example` never caught it
because it has no listener to announce.
install and enable now mount before starting; disable still unmounts before
stopping. neither direction leaves a mounted route in front of a sidecar that
cannot be reached.
verified live: /api/offscale/_officer/servers answers {"servers":[]}, /offscale
and /offscale/nodes serve, the old /api/headscale is 404, the offscale
capability is registered from the manifest, and officer-offscale is online.
757 pass, same 10 pre-existing failures.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
|
||
|
pastilhasandClaude Opus 5
|
4d4606d4a2 |
close the owner's dotfiles once somebody else has a shell
ubuntu's default umask is 002 with user-private groups, so everything the owner
creates lands 775/664. alone on a machine that is harmless. it stops being
harmless the moment a member has a login — and member homes are NESTED inside
the owner's, so the owner's home must stay traversable AND readable (the
ancestor-read requirement bun exposed today) and every dotfile in it is legible
by default.
measured as green before writing this: ~/.pm2/logs (all 12 files, every log the
platform has written), ~/.pm2/dump.pm2, ~/.claude/projects (names every
directory the owner works in), ~/.config, ~/.local, ~/.cache, ~/.npm, ~/.bun,
~/.opencode — all listable. assertSecretsClosed was already holding the line
that matters: .env, .ssh, .zsh_history, .claude.json and the credentials are
denied, and dump.pm2 turned out to hold no secret values because bun loads .env
at runtime rather than through pm2.
so this is the tier below fatal: not tokens, but logs and the shape of the
owner's work.
it runs from PROVISIONING, not from setup, and that is the point. ~/.claude does
not exist until the agent has run once; a chmod at install time finds half the
list missing and silently does nothing — the same failure mode as the ACL mask
earlier today. every member's arrival re-closes whatever appeared since.
two directories are left open on purpose, and both are the same latent bug:
/usr/local/bin/bun -> /home/pastilhas/.bun/bin/bun
/usr/local/bin/gh -> /home/pastilhas/.local/bin/gh
system-wide tools installed into one user's home, so every member resolves them
through it. i found this by closing them and breaking bun and gh for green.
~/.local/share and ~/.local/state ARE closed; only the bin directory is
reachable. the honest fix is installing them outside the owner's home.
verified both directions on this host: green is denied .claude, .pm2, .config,
.local/share, .local/state, .cache — and still has working bun, gh, psql, their
own claude, and their own project tree.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
|
||
|
pastilhasandClaude Opus 5
|
a220342b22 |
stream the install, so it reads like a log instead of a spinner
each verb now reports its steps as they complete, over server-sent events, and the detail panel renders them arriving. POST rather than GET, so EventSource is unavailable — it sends no Authorization header and these routes are owner-only. The client reads the body and parses frames by hand, which is what useCompanionLogStream already does for the headscale container logs; the parser only has to understand what our own endpoint emits. the runner does not know whether anyone is listening. it takes an optional onStep and calls it, so the non-streaming path is the same code with no callback rather than a second implementation of the same four verbs. there is a 220ms beat between steps and it is cosmetic — worth saying out loud. pm2 start genuinely takes ~770ms, measured, but writing a row and rebuilding the router do not, and four lines landing in one frame look like a stall followed by a jump. small enough not to matter to a script, long enough to follow. writing to a closed stream is caught rather than fatal: navigating away mid-install must not abort the install, because by then it is the server's work and half an install is the one outcome the ordering was designed to avoid. verified over the wire with timestamps — frames arrive incrementally, the sidecar step showing its real duration rather than the beat. afterwards pm2 holds the five core apps, plugin_installs is zero, and ecosystem.config.cjs is byte-identical. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> |
||
|
pastilhasandClaude Opus 5
|
2634df7a04 |
the install runner: ecosystem entry, sidecar, row, mounts
closes the hole app-store/pm2.ts has carried since 2026-08-13 — "installing a
plugin has to append its entry here before starting it, that is the plugin
system's job and it is not built". this is that job, and it is why nothing in
the app-store catalogue installs end to end either.
verified against a running server with a sidecar in the tree:
install ecosystem added · sidecar started · recorded · mounted /example
route 200, pm2 online
disable sidecar stopped · unmounted
route 404, pm2 stopped
enable sidecar started · mounted
route 200, pm2 online
uninstall record removed · unmounted · sidecar stopped, deleted, entry gone
route 404, not in pm2, tables untouched
afterwards ecosystem.config.cjs is byte-identical to before, pm2 holds the same
five core apps, and plugin_installs is back to zero rows.
the ecosystem file is edited rather than regenerated: the core entries come from
officer-setup's shell array, so the platform does not know that list and a copy
here would be a second thing to drift. the header above module.exports is
preserved verbatim too — officer-setup's explains that bun auto-loads .env from
the working directory and that data-path derives the install root from its
PARENT, so a wrong cwd relocates the whole install rather than failing. losing
that to a plugin install would be a poor trade.
order is the design. bringing up goes outside-in, taking down goes inside-out,
so the worst intermediate state is "recorded but not running" — visible, and
fixed by a retry — never "running but forgotten", which nothing can see.
each verb returns what it actually did, in order, and the detail panel shows it.
an install that mounted routes but could not start a sidecar is a different
outcome from one that worked, and a spinner that stops cannot say which.
the schema push is still deliberately not wired, and the reason is now in the
code: db:push DROPS tables absent from the schema it is given, so an uninstall
that regenerated the barrel would delete a plugin's data as a side effect of
stopping it. offscale does not need it — headscale_servers already ships in the
platform schema.
720 pass, same 10 pre-existing failures.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
|
||
|
pastilhasandClaude Opus 5
|
282a64a637 |
plugins install, enable, disable and uninstall at runtime
the rest of the mechanism, and it works end to end. against a real server, with
no restart at any point:
/api/example/ping BEFORE install 404
AFTER install 200 {"plugin":"example","ok":true}
AFTER disable 404
AFTER enable 200
AFTER uninstall 404
core route throughout 200
plugin_installs is a new table rather than a reuse of sidecar_installs. that one
belongs to the app store's model, where installing means provisioning a
container or pointing at a remote instance, and it carries mode, compose_dir and
completed_steps to say so. a plugin install has none of those, and reusing it
would have meant a `mode` that lies about every plugin. the two models coexist
until the app store is rebuilt on this one.
the row is needed because presence is not installation: plugins live in the
repository, so a developer writing one has the directory there and has installed
nothing. the tree says what could run, the table says what does.
mount.ts joins the two and rebuilds. an install row whose directory has gone is
dropped from the snapshot rather than reported — but the row is left in the
database, because deleting it there would turn "somebody moved the checkout"
into silent data loss. a plugin whose router will not load stays unmounted and
says why, rather than taking the other nine down with it.
/api/plugins is owner-only in its own right, like /api/app-store, and its
capability guards the MANAGEMENT surface only — a plugin's own permissions come
from its manifest, so a member can hold one at read without being able to
install anything.
plugins/example is the reference implementation and is meant to be read: the
smallest thing that is still a real plugin, with the directory layout as its own
documentation.
not wired yet, and marked [open] in the router: the schema push and the
sidecar's pm2 entry. a plugin with db/schema.ts or sidecar/ needs both before it
works end to end.
full suite: 719 pass, same 10 pre-existing failures.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
|
||
|
pastilhasandClaude Opus 5
|
327783532e |
reach a member's transcripts by listing them, not just by reading them
yesterday's fix routed transcript CONTENT through the member's identity and
stopped there, on the strength of a comment in ChatIdentity saying enumeration
never needed it — "their directories are 775 and the platform holds an ACL
entry, so readdirSync and statSync have always worked".
there are no 775 directories on this path. claude creates ~/.claude/projects/
and every project group at mode 700, and a 700 directory clamps the ACL mask to
--- exactly as a 600 file does:
user:officer:rwx #effective:---
mask::---
measured against a real member home:
existsSync(projects) -> true (stat only needs traverse on .claude)
readdirSync(projects) -> EACCES
existsSync(projects/<slug>) -> false
statSync(<transcript>) -> EACCES
existsSync answering false rather than throwing is why this was invisible: every
caller read it as "no such session". one root cause, three reported symptoms —
an empty conversation list, no title on a new chat, and a /chat/<id> deep link
that never restored the conversation. a fourth nobody had reported yet: delete
removed nothing and still answered ok, because unlink needs w+x on the group
directory too.
so enumeration goes through the same door as content, as ONE call rather than a
spawn per entry: listTranscriptsAs runs a single `find` as the member and
returns every transcript with its mtime, which readdir+stat could not do without
dozens of setpriv forks per request and a matching pile of auth.log lines. the
owner keeps a fork-free path — that process already IS the owner. removeAs does
the same for unlink, and readTailAs no longer stats a file it cannot stat.
summarizeTranscript now takes the mtime it is given instead of stat'ing again,
which is both the fix and one less syscall per file.
verified against jg@pertento.ai on this machine: 6 conversations listed with
titles from their first prompts, and a deep link by id alone loads 71 messages.
owner path re-checked and unchanged.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
|