Each lane chose its batch's side on its own coin flip. So one person's
machine could be long NQ in the futures lane while its Vest accounts sat
short NDX, hedging that person against themselves across firms.
- Every new entry now takes the side of whatever the machine holds in any
lane (machineSide in trading-logic). That includes another firm's
positions and hand trades on any of the machine's accounts.
- With positions on both sides already, nothing is added: the batch reports
positions-open with the reason, and a manual leg is refused.
- A side chosen with nothing held is reserved while its batch places it, so
a lane deciding meanwhile follows it.
- It works on the side, not the instrument: NQ against an NDX perp is the
same bet, and the indices and crypto are close enough that one side for
everything is the only rule that never hedges.
- Unchanged: one direction per batch within a lane, and the DXtrade and Vest
clients' own guards against opening opposite another of their accounts.
The fleet still doesn't coordinate sides across people.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Vest (next.vestmarkets.com) is a perps prop firm: index, metal and energy
perps on CME hours, plus stocks, FX and crypto. It joins as a fourth
platform, 'vest', on the DXtrade engine paths. Those are now keyed by
isNotionalPlatform() instead of === 'dxtrade': the firm's own symbols,
percent fees, notional sizing from volatility, and a scheduler lane of its
own.
- Login is relayed, not stored. Vest signs people in with Google through
Privy and renews the session with a device key the browser won't export.
Extension 0.17.0's vest-relay.js reads the token the Vest app renews for
itself (~15 min). It renews nothing, so the tab is never logged out. The
background script posts that token to POST /api/vest/session, which holds
it in memory only. With no Vest tab open on a machine that trades Vest,
the extension opens one.
- The firm's credential is the Vest login's wallet address. The dashboard's
form offers the relayed wallet with one click.
- VestClient (lib/vest-class.ts), mapped from the web app's traffic and code
since Vest publishes no API:
- Reads the capital accounts, balances (free amount plus position
collateral), positions and open orders with the user token.
- Mints per-account tokens for orders and transactions.
- Names accounts for their rules (V50K-E-D3-T10-34, V50K-F-D3-34,
V25K-I-7) so account-type prefixes can match.
- Takes Vest's own max-drawdown and daily-loss floors. dayStartBalance =
floor + the type's limit, so the engine's room is never more than
Vest's.
- Uses each symbol's taker fee and margin rate from exchangeInfo, and
hourly volatility over in-session hours only.
- Places take-profit and stop as position legs, which Vest cancels as an
OCO pair.
- Daily P&L and fund rows come from /v3/transactions by Vest's 20:00 ET
day. A books check holds an account, naming the transaction types, when
the balance and the transactions disagree.
- scripts/check-vest.mjs (in npm run check): Vest's day across DST, account
names, parsing, volatility gaps, market hours, the session store.
Not yet run against Vest with a live session: the order and exit bodies
come from the app's code, and the transaction classification from its enum.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Reverts 062ec07 and 5eb0ea7: every firm's broker session stays logged in
around the clock again, renewing its token on the fixed timer as before.
The tree is back to 6dd2528's. A session_rest row left in settings is
unused.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
A firm resting across the trade-date roll kept reporting its last snapshot,
so yesterday's realized P&L and DLL lockouts counted as accounts already
traded today (Nik's Topstep read 5/5 traded at 5:31 PM from 3:29 PM data).
Its report is now the unsynced one until it logs back in; a rest within one
trade date reports as before.
/api/health also counts a firm logging back in after a rest as nothing
wrong, within its 15-minute grace, rather than reading not ok for the
seconds before its first sync.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
A firm's client used to stay logged in around the clock, renewing on a fixed
timer overnight and at weekends. Now, once a firm has nothing open, nothing
pending and nothing left to trade before its next session for a random
15–45 minutes, its client disconnects and the dashboard shows it Resting.
It logs back in a random 20–50 minutes before the next session opens, at a
check-in every 3–7 Central daytime hours, or at once when trading starts, a
manual leg, copy trade, manual order or AutoBuyer run needs it. Token
renewals fall at random points (Tradovate 15–45 min before expiry, TopstepX
every 45–150 min). Settings → Trading → Rest idle broker sessions
(session_rest) turns it off.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Every machine sized a like account in a like state to the same position as
every other person's, so two people's positions could match to the unit.
Each person now draws a size factor in [0.85, 1.15] per trade date, seeded
by the master's uuid for them; all of one person's accounts share it.
The factor moves the hold a DXtrade leg is sized for (hold / factor², since
size goes as 1/√hold), so the instrument pick, the dashboard's next leg and
the Firms page's reach test all see the hold the size was made for.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
PM2's cron for autofirmer-update dies inside a live daemon (six Windows
machines in October), and nothing on the machine noticed. The app now
watches scripts/.last-check and, past 10 minutes, runs
update-check.mjs --heal: a checker PM2 still lists is restarted (which
re-registers its cron), one it lost is registered again, and one whose
errored entry still holds a live pid is left alone, since PM2 would kill
that pid. The heal scrubs PM2's per-app fields from its environment first.
The master's new 'update-now' command runs --heal --now, skipping a
failed update's retry wait. A failed pull, npm install or build records
one line of reason (or the chown to run when a file isn't ours), which
rides on .last-check to the master. A package-lock.json rewritten by npm
is put back before the pull. The fetch timeout drops under the cron's
minute. scripts/.no-auto-update opts a checkout out.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
When the master changes an account type with "New accounts only", it now
serves the type's old rules as kept rule sets (firm.keptAccountRules), each
for accounts dated inside [openedAt, closedAt). This build applies them:
- getAccountConfig, the one rules lookup trading reads, puts a kept set's
rule fields over the live type when the account's rules date falls in
its window. Id, label, prefix, size and the projection fields stay the
live type's, so every report still names the live type.
- account_assignments gains rules_date and rules_date_manual. A new row is
dated when the machine first sees the account, on the master's clock
when this one runs behind and never before the catalog's last change.
Auto dates come down to the broker's evidence: the end of the first fund
day in Chicago, or a DXtrade registration instant. Evidence is applied
before an account's first trade. Rows from before the migration get its
time; unpinning keeps the row and its date. Backups carry both columns.
- An account on a kept set stays on its type while its prefix and cached
size still match, so a tie-break flip can't move it off its old rules.
- Kept entries parse through the same normalizer as live types (now
normalizeAccountType in lib/kept-rules.ts); a malformed one rejects the
sync like a malformed type. The cached catalog records the parser that
stored it, and a cache from an older parser is fetched in full rather
than trusted with ?since= (this also brings in maxConcurrentAccounts).
- Reports add keptRuleIds per firm and live counts per kept set per type.
- The dashboard row and the account page show the rules date next to the
account name, with an "old rules" chip saying what differs. Clicking
the date edits it inline (PUT /api/accounts/:id/rules-date), previewing
which rules the date lands on.
Also: the master's per-firm cap on accounts traded at once
(maxConcurrentAccounts), applied on top of the machine's own max
concurrent when a batch is drawn and in copy trading.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
A sync reads the balance first and each account's positions and trades
after it, so a fill landing in between shows in the trades but not yet in
the balance, and for that one pass the books are off by the whole fill.
reconcileFunds booked that as a fund event, and the next pass booked the
reversal. Same-day rows merge, so the two netted to the commission drift
of the day and left a tiny row:
- 2026-10-05, Express 43788281 and 23275136: +3,001.16 "deposit" (balance
-2.16 against the -3,001.16 stop-out already in the trades), then
-3,002.16, merged to -1.00.
- 2026-10-08, combine 99745612: +2,474.01, then -2,492.01, left -18.00.
- 2026-09-24, five 150K combines on another instance: +3,049.45, then
-3,059.45, left -10.00 each.
A negative row is a withdrawal. It started a new cycle, dropped its day
out of the cycle (the -3,003.32 on 10-05) and counted toward the stage:
both Express accounts traded drain days from 10-06 on a $1 "payout".
- fundReconcileStep (trading-logic): an unexplained balance is booked
only when a check at least 30s after the first sighting still sees it,
to within $5. A transient is gone by the next pass and books nothing.
- A transient that outlasts that, and so books, is taken back whole once
its reversal nets the day under FUND_EVENT_MIN. The day's first booking
saves what it replaced (fund_day_before), and dropFundTransaction
deletes the row and puts last_fund_date and ended_cycle_start back in
one transaction. A row reconcileFunds didn't book (an import) merges as
before.
- funds_reconciled is set only once a check settles, so a first-check
lump that waits on a second pass is still booked as the first check.
Existing tiny rows are left as they are; cleaning them up is a separate,
one-time step.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
TopstepX's gateway reports a fill's exchange and NFA fees only (NQ $1.39 a
side, GC $1.66, CL $1.51, MNQ $0.36). Topstep's own commission, $0.50 a side
on a full-size contract and $0.25 on a micro, comes off the balance with no
trade row carrying it, so our P&L ran $1.00 per full-size round turn above
Topstep's. On 2026-10-18 the combines sat $6 to $8 under their books and an
XFA $27.50 under, and one combine had its $18 booked as a "withdrawal".
Topstep's published round turns (NQ $3.78, MNQ $1.22, GC $4.32, CL $4.02)
are the reported fees plus this commission.
- lib/topstep-fees.ts: the per-product commission table, and TopstepX
contract ids mapped to roots (ENQ, GCE and CLE as seen on real fills,
overridden by contracts the client resolves itself; an unknown product
pays the full-size rate and is logged once).
- Daily P&L, today's realized, and so the measured entry fee a take-profit
is placed from, all net the commission. The 90-day refetch rewrites the
cached days.
- npm run check: the table against the micro flags, the id mapping, and
the published round turns.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Every FirmStats now carries statusVersion 1, the catalog firmId and, once
synced, syncedAt (from lastSyncAt), including the no-accounts branch, which
is now built by emptyFirmStats(). Each accounts[] entry adds status, reason,
balance, floor, accountTypeId, firstFundDate, positiveFunds and brokerStatus,
each left out when null or at its default.
accountStatus() in platform-client.ts is pure: a missing balance is unknown
(cash_missing), never dead; with a balance, dead is exactly the old floor
test (legacyDead); then history_pending, locked, live. Only the reporter
uses it. The aggregates, totals and every trading decision keep their own
dead test, unchanged.
The reporter reads history readiness through a new Tradovate
peekHistoryReady(), because historyReady() there can start a hold and move
the balance baseline. DXTrade exposes accountStatus and registrationTime via
brokerAccountMeta(). A throw in the new fields reports the account as
unknown instead of losing the report.
sendReport skips the socket for a firms payload over 900,000 chars (the
master's frame cap is 1,000,000 bytes) and POSTs it, logging once per switch.
scripts/check-account-status.mjs proves the legacy invariant over the full
matrix and the size budget (worst 228 B added per account, 154 B typical
live; 2000 accounts 592,501 B).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
4217.6 − 87 × 0.1 is 4208.900000000001 in floating point. TopstepX refused
that take-profit ("Invalid price. Price is outside allowed range") on two
GC shorts on 2026-10-02, and with no stop on the leg nothing else was
working: both rode far past their target with no exit. ES and NQ never hit
it because 0.25 is exact in binary; GC and CL ticks aren't.
- onTick puts the measured-fee take-profit and every stop on the grid —
the take-profit away from the entry, the stop toward it — so an averaged
multi-contract fill between ticks is handled the same way.
- A leg's lone take-profit (no stop, or the fallback after a refused
bracket) that the broker refuses now flattens the position and benches
the account for the day, like a refused bracket. An unknown outcome still
places nothing more.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
A firm can cap how many contracts of an instrument its accounts may hold
(the catalog's instrumentLimits, set on the master's Firms page). A leg
trades one contract per $1,000, capped by the account type's max contracts
and that limit. The Auto draw sizes each instrument's hold with the capped
count, so a limit that slows a leg past 12 hours takes it out of the draw.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
They traded the micro of whatever the firm listed: MNQ with NQ, but MGC
for a firm listing only GC and CL. Now every futures firm's tiny legs go on
MNQ, the one micro the bot has always drained on.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The master saw a scheduler's last run time and nothing of what it did. Each
tick now records, per lane, its outcome, a line on what happened (what
fired, or the spread, contract or hold cap that stopped it) and why the
lane's accounts sat it out, counted by reason with a few names each. It
rides the scheduler status as lastTicks.
Each report also carries the auto-updater's state from the files it keeps
and from git: its last check, a failed update waiting to retry, how far the
checkout is behind origin, and tracked files edited by hand, which stop the
fast-forward pull.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Tradovate moves each account's cash record to the new trade date on its own
schedule, an idle one perhaps not until it trades, and the bot holds any
account still on the old day until 5:45. Opening at 5:05 fired the ready
accounts alone (TopstepX, the early rollers) while the rest, released at
5:45, waited an hour or more for them to go flat. A 5:45 open puts every
account in the first batch and skips the open's widest spreads. The day's
0–180 minute delay still comes on top: first entries land 5:45–8:45 PM.
Every-day firms keep their own 5:05 PM end of the roll guard; the equity
hours open is unchanged. The draw's session progress now starts at 5:45.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The day's delay was 1–20 minutes. It is now 0–180, so the evening open
lands anywhere from 5:05 to 8:05 PM CT. The day's roll is saved as a
setting: with a three-hour window, a restart during the wait (an update, a
crash) would otherwise roll again and could start the session early.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
One quote feed per machine: a TopstepX client's market hub, else a
Tradovate client's market-data socket, subscribed to every listed root's
front month and the drain micro. A supervisor keeps it on a working client,
dropping one that fails or goes quiet while the market trades.
The Auto draw weighs each instrument by its live spread (the typical tick
when no fresh quote is at hand). No futures entry pays more than the
master's max entry spread ($15 until set) to cross the book: the draw
leaves such an instrument out, an explicit symbol sits the tick out, and
fireLeg and the extra-day trade check it last before sending. The Settings
page and the master's stats show the live spreads and the feed.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
fireLeg clamped the order to the instrument's max and min size, but the
size the pick ranked by (and the Firms page shows) didn't: an instrument
whose largest order falls short of the ideal notional ranked on a hold it
can't reach. planLegSize now holds the notional to the margin cap and the
broker's largest order, then its smallest, so the ranking, the stats and
the trade agree.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The firm rolls its day at 22:00 UTC; our trade date at 5 PM CT. In summer
they coincide, in winter the firm's rolls an hour earlier. New entries are
already kept out of that hour, but a take-profit or stop that fills in it
was booked to the day before. firmDayOf labels the firm's day the way
currentTradeDateCT labels ours, and the sync's transfer window now starts
at the firm day behind today's trade date.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The rule is: when hitting the day's loss limit would leave the account 15%
of that loss or less above its max-drawdown floor, ride without a
daily-loss stop. It had compared against the stop-out after the 10%
reserve, which made it stricter — $3,294.70 to the floor with $3,000 of
daily room (9.8% left) wasn't a last stand, and now is.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
- lib/instrument-specs.ts (pure, tested): the CME specs — point value,
tick size, one-tick spread, Yahoo ticker — replacing POINT_VALUES; the
draw's maths; the 12-hour hold cap; the micro for drain and extra-day
legs; reading a firm's list (the master's, or the old rules against an
older master: this machine's enabled set less bans, the symbol map).
- lib/instruments.ts: the one module the trading paths ask. Measures every
future's daily move from Yahoo's hourly candles once per trade date, and
shows entry cost (a tick of spread plus Tradovate's fee parameters or
TopstepX's last measured fill) without ever choosing by it.
- Futures Auto draws one root per batch from the batch firm's list: movement
per dollar of spread early in the session with a 10% floor each, sliding
to the fastest movers by the last entry; nothing expected to hold 12h or
more; micros only if nothing else is listed. Every draw logs its odds.
- The batch is chosen before its instrument, and a group that can't trade
anything yields to the next. Copy trade matches the held contract against
the listed roots and won't copy an instrument it can't identify. Manual
legs, manual orders and drain/extra-day legs follow the lists.
- DXtrade firms rank their own list whatever root was drawn.
- Capabilities, the dashboard's trade bar and contract resolves follow the
lists; the trade bar and scheduler default to Auto. The master link
reconnects when master_dashboard_url changes, and the catalog drops
?since after a switch so two masters' versions can't collide.
- Reports carry each futures firm's instrumentStats for the master's Firms
page; GET /api/instruments shows them (?simulate=N runs the draw, no
orders); the Settings page shows the lists read-only.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
- A batch sizes its leg on every mapped candidate, each at the leverage
the broker states for it, and trades the cheapest (spread, commissions,
expected swap) that reaches the target hold — else the quickest. The
dashboard ranks the same way from a half-hourly candidate survey.
- Last stand: when a daily-loss stop-out would leave the account within
15% of that loss above its max-drawdown floor (the account type's static
max loss, or the broker's liquidation floor), the leg carries no
daily-loss stop and is sized against the firm's own breach.
- The daily-loss reserve is 10% of the limit or 0.1% of the position,
whichever is more, since a stop's slippage grows with the position.
- The account type's max notional is gone: the broker's margin rates cap
each symbol. planLegSize holds the one sizing rule fireLeg, the pick and
the dashboard share.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
- Transfers: a close books its SETTLEMENT and COMMISSION under one
transaction code, and keying by the code alone let the commission
overwrite the P&L. Keys now carry the type, so the day-start balance,
daily-loss room, realized P&L, daily history and books check all see
closed-trade profit and loss.
- Volatility from a week of hourly closes (was 48 h), still measured once
per trade date, so one quiet weekend can't triple a hold estimate.
- Each symbol's margin rate from the account's instrument details caps
the position at available funds / rate (Tradeify: 5:1 crypto, 2:1
stocks); the account type's max notional stays as an optional ceiling.
- Firms with overnight financing add the swap a winning leg is expected
to pay (the winners' mean hold, (A²+2AB)/3σ²) to its take-profit, and
the sizing and hold estimate count it.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Tradeify247's API user may not read DXtrade market data (code 34), which
skipped every batch at the instrument pick. Quotes and hourly candles then
come from Hyperliquid's public API (BTC, ETH on the main dex; XYZ100, SP500
and tokenized stocks on trade[XYZ]) — for the pick and sizing only; orders,
fills and exit prices stay DXtrade's.
Closing STOP/LIMIT orders close the whole position by positionCode and must
not carry a quantity (code 33) — the exit bracket was refused, and the
daily-loss guard flattened both accounts. A close's fill price also falls
back to its executions.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The symbol map's fixed notional per contract is gone: every DXtrade leg is
sized from the instrument's recent volatility for the firm's target hold
(blank = 12 h), and a candidate without a volatility isn't picked. Percent
commission and every-day trading are ignored on non-DXtrade firms, which keep
measured per-contract fees and the CME calendar.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
- DXtrade client (lib/dxtrade-class.ts): REST session login with lockout-safe
backoff, 60s sync of accounts/metrics/positions/orders/transfers, history and
fund transactions from cash transactions, day-start balance at the 22:00 UTC
reset, CLOSE-by-positionCode exits (native OCO group, else a stop-first
emulated pair), a cross-account hedge guard, stray-order reaper.
- Instrument selector: the firm's symbol map lists candidate symbols per CME
root; the cheapest tradable one per unit of recent hourly volatility wins
(spread + 2x commission [+ financing] / sigma).
- Fees: a firm's commission % of notional prices every fee off the fill; the
take-profit is solved in closed form so the leg nets its target.
- Hard daily loss limit per account type: every leg carries a stop keeping the
firm's day inside the limit less a 10% reserve and both fees (rounded in);
a green day never widens it; no room or a refused bracket flattens.
- Notional sizing where the platform prices positions: target-hold volatility
sizing or the symbol map's fixed notional, under a per-type notional cap.
- Every-day firms trade weekends/holidays; the scheduler runs lanes (futures,
one per DXtrade firm) with their own hours, flat gate, mutex and batches.
- The catalog sync skips platforms this build doesn't know instead of treating
them as Tradovate. One shared fireLeg replaces the duplicated batch/copy code.
- npm run check: fee/stop/sizing/hours/selector checks; scripts/dxtrade-smoke.mjs
confirms the API against a real login before trading.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Why the autotrader sometimes broke after a weekend until `pm2 restart`:
`next start` only logs uncaught errors, so nothing ever crashed, and a few
paths could leave a client dead for the life of the process.
Tradovate: a closed socket scheduled exactly one re-login, with no catch —
one failure during weekend maintenance (or a rate-limit penalty returned as
a 200 without a token) and the client never came back. connectAndAuth is now
a single-flight loop that never throws: jittered backoff capped at 5 min,
penalty tickets waited out, captcha demands backed off. A 15s watchdog
abandons silent, never-authorized or stuck-closing sockets and stale syncs;
renewal falls back to a fresh login; orders fail fast (OrderNotPlacedError)
on a dead socket and time out instead of hanging; a bad sync answer no
longer wipes accounts and positions.
TopstepX: hub start/stop/invoke and every REST call are bounded; a wedged
hub is discarded and rebuilt; a stuck sync is superseded after 3 min;
re-login no longer tears down the sync loop; a watchdog re-logs when syncs
go stale; a no-fill timeout cancels the order.
Scheduler: one unready client used to hold every firm forever. It now holds
the tick for a 10-minute grace, then trades the other firms without it.
Per-firm try/catch in batch collection, one batch at a time, accounts marked
in flight, legs un-counted only when the entry never reached the broker, and
the scheduler starts before the first batch fires.
Boot and supervision: instrumentation.ts boots the process at server start
(it used to wait for the first dashboard request); a supervisor rebuilds a
client that has been unhealthy for 20 min (not a TopstepX one holding a
position); broker HTTP calls default to a 60s timeout; GET /api/health shows
each client's liveness and the scheduler's last decision. Shared state the
eager boot would otherwise split across bundles lives on globalThis. Logs no
longer carry access tokens or whole AxiosErrors (which include the encoded
Tradovate password).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
PM2 runs update-check.mjs as a cron *restart*, and PM2 7 stops a process
that is still running at the next minute boundary. Any update that needed
npm install or next build (more than ~55s) was killed half-way: the lock
was left behind, HEAD had already moved at the pull so the next run saw
nothing to do, and next build had already wiped .next under the live
server — a box that broke after a weekend push and stayed broken.
- The cron run is now a quick CHECK: fetch, decide, and hand the work to a
detached worker (--apply) that PM2's next tick can't reach.
- "Up to date" is judged against scripts/.built-commit (what .next was built
from), not HEAD, so a pull that was never built finishes on the next check.
First run on a machine adopts a complete build that is newer than HEAD's
last move and rebuilds anything else — that rescues boxes the old updater
already stranded.
- The previous build is set aside as .next-prev and restored if the build
fails, so a broken push never leaves a box without a build; failures retry
after 15 min, then hourly (scripts/.update-failed).
- The lock holds a pid and is atomic; build dirs are deleted without ever
following the links in .next/node_modules into the real node_modules.
- Every instance in scripts/instances.json is restarted, and python/pip
children stay hidden on Windows now that the worker has no console.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
`node -` puts '-' in argv[1], so the registry snippets in add-instance.sh
and remove-instance.sh were one slot off: they wrote the registry to a file
named "-" in the working directory, with the instances.json path as the
instance name, a null port and the port as the db. scripts/instances.json
was never created, so the updater and start-all only ever knew instance 1 —
after an update, instance 2 and up kept running old code against the new
build until someone ran `pm2 restart all`.
Devices that already ran add-instance.sh need their scripts/instances.json
written by hand (and the stray "-" file deleted).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
lib/db.ts switched to WAL at module load before setting busy_timeout.
next build imports it from several workers while the live server holds
the same file, and a journal-mode change needs the file to itself — so
the pragma failed instantly with SQLITE_BUSY and took the whole build
down ("Failed to collect page data"), blocking updates on any box whose
database was still in rollback mode. The timeout is now set first and
the WAL switch is wrapped: a held file keeps its current mode.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Four features:
- Backup: lib/backup-sync.ts pushes the full settings table and firm
credentials (the deliberate exception to "credentials never leave the
machine") to the master's bearer-gated /api/instances/backup — every
5 minutes when the content hash changed, within seconds of a settings
or credential edit, and on the new backup-now command. Enabled by the
master_backup_token setting.
- Restore: a dead VPS's replacement pulls the backup and adopts its
uuid/name, so roster, payouts and card links carry on. Pre-boot via
seed-settings.js --restore-from (setup-linux.sh prompts for it) or
the Settings page's restore picker (lib/backup-restore.ts). Stop the
old machine first; the master 409s a live-name rename.
- Proxy: the proxy_url setting routes BROKER traffic only — TopstepX
REST + SignalR hub, Tradovate REST + websocket, Yahoo quotes — via
lib/outbound-proxy.ts (axios agents, undici dispatcher, ws agent).
The master link, reporter, catalog and backup push stay direct, so a
dead proxy cannot take down the control channel. Restart to apply.
- Multi-instance: AUTOFIRMER_DB moves the SQLite file so one checkout
runs N instances; scripts/instances.json (per-device, gitignored)
drives start-all/update-check, and add-instance.sh / remove-instance.sh
manage the extra PM2 processes. Absent registry = single-instance
behavior, unchanged. Instance 1 keeps port 3000 for the clicker and
extension.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
setup-linux.sh accepted being run as root and simply set SUDO="". That quietly
broke everything user-scoped: PM2 keeps its process list in $HOME/.pm2 and the
login entry goes in $HOME/.config/autostart, so under sudo both landed in /root.
The install reported success, then `pm2 list` as the real user showed nothing -
it was talking to a different daemon - and the autostart entry sat in a home
directory that never logs in.
Both scripts now refuse, and name the account to use when SUDO_USER says how
they were invoked. setup-linux.sh still uses sudo for apt, which is the only
part that needs it.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
The no-change path is a single `git fetch`, so polling master every minute
costs nothing and lines up with the VPS's own 60s deploy timer instead of
lagging it by up to five.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Neither can be done for the running profile: UserChoice is hash-validated and
guarded by UCPD.sys since Feb 2024, and taskbar pinning lost its shell verb
years ago - both to stop installers helping themselves. What is supported is
policy, which applies when a profile is created, so this writes the
DefaultAssociationsConfiguration and taskbar layout XMLs for the next user and
does the two things that help the current one: a desktop shortcut, and opening
the Settings page where the browser choice actually has to be made.
Folded into the enrolment script's existing elevation so the install still
raises a single UAC prompt.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Invoke-WebRequest pushes every buffer through the pipeline and is the slowest
of the three by a wide margin; curl.exe ships in System32 on every supported
Windows and resumes a partial file with -C -, which matters on a 166 MB
download that gets interrupted. Also checks the finished file against
Content-Length, because a truncated MSI fails in msiexec with an error that
says nothing about why.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Windows PowerShell re-renders progress on every buffer, so a 120 MB download
spends most of its time drawing a bar rather than transferring - minutes for
what should be seconds, and indistinguishable from a stall while you watch the
byte count crawl.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Setup asked for a token that is the same on every machine - a token is reusable
and enrols any number of browsers - so it is a default in both scripts now and
step 7 is a plain yes/no. It is a modest secret: it lets someone enrol their own
browser into the org and nothing more, and rotating it is a console click and a
one-line change.
Both scripts also install Chrome if it is absent, from Google's enterprise MSI
and the official .deb, since enrolling a browser that is not there writes a
token nothing reads. Chromium is refused rather than substituted: no cloud
management, so it could never be enrolled - and on ARM, where Chrome is not
published at all, the Linux script says so instead of installing something that
cannot do the job.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Both setup scripts ended by telling you to load the folder unpacked, which is
the one install that cannot update itself - and the policy install they moved
to next does nothing on an unmanaged machine. Enrolment is the step that
actually matters: with it, the extension arrives from the Admin console already
force-installed, and there is nothing else to configure on the box.
So step 7 now asks for an enrolment token instead. The Windows script elevates
itself for it - the token is machine-scoped, and that is the only UAC prompt in
the whole install.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
The managed check asked Windows, but Enterprise Core is browser management by
Google: Windows reports a plain standalone machine however enrolled Chrome is.
The guard therefore blocked the exact path it recommends. It now asks Chrome
first - the dmtoken it writes on a successful enrolment, and the enrolment
token that requests one, in both registry views - and only then falls back to
domain and Entra join state.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Enrolling is what lifts the off-store restriction without joining a domain, and
it is one machine-scoped token in both places Chrome looks: HKLM on Windows, a
file under /etc/opt/chrome/policies/enrollment on Linux. Both scripts take
--remove, and both say plainly that removing the token locally does not release
the seat in the Admin console.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Chrome blanks an off-store ExtensionInstallForcelist entry on an unmanaged
machine exactly as Edge does - the qualifying states are AD domain join, Entra
join, or Chrome Enterprise Core enrolment, and this box is none of them. The
guard now covers every browser and refuses up front, with the three ways
forward, instead of skipping Edge alone and letting Chrome fail the same way.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Unmanaged Edge installs extensions from the Microsoft Store and nowhere else:
it reads ExtensionInstallForcelist, blanks the value to [BLOCKED]<id>, and
reports an invalid extension id forever after. Writing that key buys nothing
and costs a permanent policy error plus a "Managed by your organization"
badge, so the script now checks dsregcmd and skips Edge with the two ways
round it. -Force writes it anyway.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Test-BrowserInstalled probed App Paths under HKLM:\SOFTWARE only. Edge and
Chrome ship 32-bit installers, so on a 64-bit box their App Paths entries live
under WOW6432Node and a 64-bit PowerShell never saw them — the browser was
skipped, silently, and -Browser All configured nothing for it.
Checks the install directories first now, with App Paths (both registry views)
as a backstop, and prints which browsers it is configuring.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
They are Chromium: same CRX, same update manifest, same policy — only the
registry vendor key differs. The script now defaults to every one of the three
it finds installed, which is what the Linux counterpart already did.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Both installers ended by telling you to load the folder unpacked, which is the
one install that can never update itself. They now offer the policy install as
a last step instead, and only fall back to those instructions when it is
declined or fails.
Windows writes to the user's own policy hive rather than HKLM, so the step
needs no elevation and no separate admin shell; -Scope Machine is still there
for every-account installs. Linux gets the counterpart script — a managed
policy file under /etc, which is root-only because Chrome reads policy from
nowhere else on that platform.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
The dashboard-hosting stack from the previous commit is gone; it was answering
a question nobody asked. What was actually needed is Chrome's real update
channel, which an unpacked extension does not have and cannot be given.
So the extension is now signed into a CRX with an update manifest beside it,
served from the droplet, and installed on the trading machines by policy
(ExtensionInstallForcelist). From there Chrome installs it, updates it on its
own schedule, and puts it back if it goes missing. The manifest gained the
`key` that fixes the extension id, so an unpacked development load keeps the
same identity as the published build.
The self-update check now branches on how it was installed: policy installs ask
Chrome to check the update server now rather than waiting out its five-hour
cycle, and unpacked loads keep the reload-from-disk path. chrome.management
.getSelf() tells the two apart and needs no permission to do it.
The signing key lives at /opt/autofirmer-updates/key.pem on the droplet and
nowhere else. It is the extension's identity: a different key is a different
extension, and no browser would take it as an update.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
