Fixes found while reviewing #4547, stacked on that branch so they can be
reviewed on their own and merged into it. One commit per fix.
## Write-path correctness
**Refuse account writes while impersonating.** The five
`dashboardPreferences` writers already no-op for an impersonating admin,
but the three profile writers added next to them did not, and
`requireUserId` returns the impersonated user's id. Both gates now
refuse up front and say so, rather than the preference writers silently
no-opping while the page reports success.
**Preserve unknown keys on a full-blob write.**
`mutateDashboardPreferences` parses the JSON column, hands the result to
a mutator and persists the whole object back. zod strips keys it does
not declare, so a deploy that predates a preference field drops it on
the next write through that path — and
`updateCurrentProjectEnvironmentId` sits on the navigation hot path.
`preserveUnknownKeys` re-attaches them at the write. Note this cannot
help deploys already running, so it makes this the last release able to
strip rather than retroactively protecting the fields added in #4547.
**Scope hidden-sidebar writes to what was shown.** The customize dialog
builds its hidden map from the sections it can see and the write
replaced `hiddenItems` wholesale. The profile page has no org in scope,
so it resolves sections from the most-recently-updated project's org:
confirming there dropped hidden ids belonging to sections that org's
flags exclude. The payload now carries the ids the dialog rendered and
the write only replaces those. Submissions without the list stay
authoritative.
**Consider both addresses when checking email ownership.** The check
only looked at the address the user already had; it now considers the
current and submitted address together, so an org managing either one
governs the change. Validation moved ahead of the check, and
`emailDomainOf` splits on the last `@`.
## Interaction
**Revert unsaved themes, debounce contrast saves.** The theme and
system-theme selects stamp `data-theme` before the write lands. When it
fails, the loader returns the value it always had — so
`useSystemThemeSync`'s effect deps are unchanged and React's vdom diff
sees no change either, and nothing rewrites the attribute. The page kept
rendering a theme that was never stored while the select showed the
stored one. The stored pair is now re-applied explicitly, as the side
menu's switcher already did. The contrast slider is debounced because
Radix commits on every arrow keypress, so a keyboard user crossing the
range fired one write per step.
**Tick More options for themes outside the short list.** The appearance
submenu offers System, Light and Dark; Black and White live on the
profile page. With one of those stored, every row read as unselected.
## Subtraction
**Drop the profile update rate limiter.** It covered one of four paths
that write the same column — `resources.preferences.sidemenu` and
`.favorites` take unlimited authenticated writes and go through the
locked read-modify-write, which is more expensive than the single narrow
`jsonb_set` this capped. It was also what made the contrast slider
unusable by keyboard. If preference writes want limiting, it belongs in
one place covering all of them.
**Resolve email ownership when the dialog opens.** It fans out one SSO
status lookup per organization the user belongs to and ran in the
profile loader on every page view, purely to pick which body the dialog
renders. The action re-derives it before writing either way, so the
check that guards the write now has one call site instead of two.
## Testing
`typecheck --filter webapp` and `lint` clean. New unit tests for
`preserveUnknownKeys`, `mergeHiddenItems` and `emailDomainOf`;
`themePreference`, `mergeHiddenItems` and `ssoManagedIdentity` suites
pass locally (26 tests). The rest of the webapp suite needs
testcontainers and is left to CI.
No changeset or `.server-changes` entry: everything here fixes code on
the parent branch that has not shipped. The one exception worth a
maintainer's call is `mergeHiddenItems`, which also touches the side
menu's own customize path.
## What this does
Rounds out the theme work behind the existing `hasThemeSwitcher` flag.
**Two new themes.** Black and White sit alongside Dark and Light. They
inherit their neighbour's whole token set and only pin their surfaces
flat, so sections are separated by grid lines rather than layered fills.
**`System` is now configurable at both ends.** You choose which theme
the OS light setting lands on (Light or White) and which the dark
setting lands on (Dark or Black).
**Two accessibility toggles.**
- *Stronger colors* — swaps tinted status chips for solid fills, drops
decorative icon accents to monochrome, and darkens chart series that
didn't clear 3:1 on a white plot.
- *Underline links* — underlines body-text links, so an underline always
means the preference is on rather than being a hover style.
**Contrast slider.** Stores a 0–100 position within the active theme's
own range rather than a shared scale, so 35% stays 35% when you switch
themes. Each theme maps it in CSS, which keeps `system` working before
hydration.
**Appearance in the account popover.** A submenu listing the themes with
a check against the current one, plus a link through to the full set on
your profile. Picking one applies immediately rather than waiting for
the write to round-trip.
**Profile page.** Each row now saves on its own — no submit button. Name
and email show their value inline with an edit button; the email row is
read-only when an identity provider owns the address.
**A `/storybook/colors` audit page.** Renders every colour-carrying
pattern in the app once per theme plus once under Stronger colors, and
measures contrast ratios off the live DOM rather than a hard-coded
table, so it can't go stale.
---
## Demo
https://github.com/user-attachments/assets/d56cd4d8-719f-4ec5-a990-e04cdb98def1
---
## Compatibility
The stored preference shape is unchanged (`version: "1"`), and the four
new fields are all optional. The retired `classic` theme falls back to
Dark, whose palette at contrast 0 is what Classic shipped.
One deliberate change worth knowing: the default contrast moves from 50
to 0, so existing users who never touched the slider will see slightly
less contrast than before. That's what makes 0 mean "the base palette".
---
## Testing
Switched between every theme from both the account popover and the
profile page, in the expanded and collapsed rail, checking `data-theme`
follows and survives a reload. Dragged the contrast slider in each theme
and confirmed the percentage label tracks the handle and resnaps if a
save fails. Checked both accessibility toggles across the
`/storybook/colors` page, which is also where the contrast ratios were
read from. Confirmed the Appearance entry stays hidden for a non-admin
while the flag is off.
<!-- conductor-workspace-link -->
---
[Open workspace in
Conductor](https://app.conductor.build/workspace/fee50611-7623-4422-bada-ed1cba317ed1)
---------
Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>
Adds `RedisSnapshotStore` to `@internal/run-store`: a Redis-backed,
append-only store for a run's execution-state log, as an alternative to
keeping that log in Postgres.
Nothing constructs it. No existing code path can reach it, so merging
this changes no behaviour. The store, the wiring that would use it, and
the switch that would enable it are deliberately separate changes.
## Design
Four keys per run, plus one key per wait cycle, all sharing a `{runId}`
hash tag. Every mutation for a run therefore lands in one cluster slot,
and each operation is a single Lua script.
No script mints a key name. Dynamic keys are derived from `KEYS[1]` by
string surgery, because ioredis applies `keyPrefix` only to the KEYS
array: a key built inside Lua would be unprefixed while the client wrote
a prefixed one.
Retention is keyed to run completion. A non-terminal run's keys carry no
expiry at all, since a suspended run can wait indefinitely with nothing
left to refresh a TTL. The terminal transition sets the completion
expiry once, and a write arriving after completion re-applies that same
expiry rather than a live one, so a stale client cannot resurrect a key.
Entry JSON round-trips byte for byte. No script calls `cjson`, and the
values the store assigns itself live in their own hash fields instead of
being patched into the caller's document.
Sizes are observed, never enforced. Entry and cycle-key bytes are
recorded, with a warning above a configurable mark. Nothing rejects,
truncates, or spills.
`append` takes an optional expected-current-snapshot argument. Left out,
it advances the pointer unconditionally, matching the Postgres behaviour
it replaces. Supplied, it advances only on a match and otherwise reports
the conflict without writing.
Covered by 48 tests against a real Redis container, including the
retention transitions, the single-slot guarantee under a key prefix, and
tenant-scoped reads.
Connecting a Vercel project now writes
TRIGGER_AUTOMATIC_SKEW_VERSION_PROTECTION=1
(plain, create-if-absent only - an existing value, including "0", is
never
touched; presence is target-containment aware, branch-scoped records do
not
count, a truncated env listing skips the write). The onboarding wizard
no
longer offers automatic atomic deployments (default off); the settings
row is
labelled Deprecated and enabling it requires confirming a dialog that
points
to task version skew protection and the docs (TRI-13001).
New deployment/version-skew-protection page: the skew problem, the
--external-id primitive and its reuse behaviour, runtime discovery (call
option, configure(), TRIGGER_EXTERNAL_DEPLOYMENT_ID, and the gated
platform/CI/generic commit-SHA variables with the build-time caveat),
the
manual any-platform recipe, waiting/expiry semantics, precedence, and
automatic skew protection on Vercel. Deprecation callouts on the atomic
deployments page and the Vercel integration page; --external-id/--force
added to the CLI deploy reference; redirect from
deployment/vercel-skew-protection so existing webapp links resolve
(TRI-13002).
## Summary
Makes the runs list customizable. A new **Display** control lets you
show, hide, and reorder columns, and add **smart columns** that pull a
single value out of a run's payload, metadata, or output by JSON path
(e.g. `$.failed`, `$.order.total`). Column choices live in the page URL,
so a view can be bookmarked or shared. Applies to the global runs list
and every per-task / scheduled / agent / webhook / error list, which all
share one table.
ID, Task, and Status can be reordered but not hidden. Smart columns are
display-only (no sort or filter, which would defeat the ClickHouse sort
key and cursor).
## How it works
Columns come from a shared registry; the Postgres `select` is derived
from the visible columns, so a run's large payload/output are only
hydrated when a smart column actually references them. All JSON parsing
for smart columns happens client-side, respecting the packet content
type, parsed once per source per row. Offloaded (too-large) values and
paths that aren't present render distinct placeholders rather than
fetching per row. The live poll carries the same sources so smart-column
values update in place.
Scalar columns stay always-selected for now: the shared list presenter
has a fixed output shape consumed by several routes and the live poll,
and narrowing individual scalar fields would add no real query cost
benefit on a single-row read. The select derivation is already
column-driven, so tightening this later is a one-line change.
## Screenshots
<img width="590" height="1028" alt="CleanShot 2026-08-21 at 16 48 17@2x"
src="https://github.com/user-attachments/assets/86b39856-bfcc-47c0-85ed-ee6ccddc3590"
/>
<img width="1924" height="1528" alt="CleanShot 2026-08-21 at 16 48
27@2x"
src="https://github.com/user-attachments/assets/6c766249-6d5b-45be-9330-c6caa75af7f7"
/>
<!-- conductor-workspace-link -->
---
[Open workspace in
Conductor](https://app.conductor.build/workspace/d6911080-2140-4de1-b88a-1b0623593caa)
---------
Co-authored-by: James Ritchie <james@trigger.dev>
Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>
## What
`RoutingRunStore` held two named store fields, `#new` and `#legacy`, and
took its routing policy from the order the statements happened to run
in. It now holds a `Map<ShardKey, RunStore>`, and the three policies
that were implicit are readable data:
- **`#probeOrder`** (`new` → `legacy`) — the sequential probe for a
lookup with no routable id. The first non-null result wins, and the
*last* entry owns the canonical not-found throw.
- **`#precedence`** (`legacy` → `new`) — ascending authority for a
merge, so the highest-authority shard wins a duplicate id.
- **`#idlessRouteShard`** (`new`) and **`#idlessWaitpointShard`**
(`legacy`) — the two id-less defaults, which differ by role and were
previously two unrelated literals in unrelated methods.
The two orders are the **reverse of each other**, which is why they are
separate fields rather than one ordering. Nine sites observe the
result-array order and must iterate `#probeOrder`; five decide a value
by which shard wins a duplicate and must iterate `#precedence`. Five
more sum counts and are order-independent, because addition commutes.
Four helpers absorb the twenty-six hand-written fan-outs —
`#probeFirst`, `#fanOut(order, fn)`, `#fanOutPartitioned`,
`#shardsExcept` — and `#shardKeyOf` replaces the inline
residency-to-store ternaries. `#fanOut` takes its order as an argument
so every call site states which policy it uses.
The constructor keeps its exact options type. No union arm, no `shards`
member: that would loosen the excess-property check and silently retire
the `@ts-expect-error onLegacyRead` lock in the test corpus. N-way
construction is a later change.
## One behaviour change
`findManyTaskRunWaitpoints` merged its edge rows NEW-first into a
last-wins dedupe, so a duplicate edge id resolved to the **legacy** row
— the opposite of the rule the other four merges follow, and the
opposite of what `dedupeEdgesById`'s own comment claimed. No test pinned
it in either direction.
It now resolves NEW-wins, consistent with every sibling merge, and a new
test pins the winner so it cannot drift back silently.
Reaching this case needs one edge id present on both stores at the same
time, with no routable `taskRunId`. That only arises from drain
mirroring. The drain seam is removed (`runOpsStore.test.ts`, "fan-out
spans NEW+LEGACY with no drain seam"), so **no new duplicates can be
created** — but removing the code does not delete rows it previously
wrote, and this class still carries comments treating mirrored rows as a
live data condition. Whether any historical duplicate edge rows persist
is an empirical question about production data, not something this diff
settles.
If such a row is hit, the two copies either agree — in which case the
winner is immaterial — or they have diverged, in which case NEW is the
authoritative copy by the router's own precedence rule. So the corrected
behaviour is at least as correct as the old one in every reachable case.
Everything else is behaviour-preserving.
## How it was verified
- **`internal-packages/run-store`: 69 files, 379 tests pass.** The
corpus is the regression gate for this refactor. 67 of the 68
pre-existing test files are byte-identical; the one that differs
(`runOpsStore.mixedResidency.test.ts`) changes only `//` comments.
- **`internal-packages/run-engine`: 12 files, 69 tests pass** — every
file that constructs the router, exercised at runtime.
- **The `@ts-expect-error onLegacyRead` lock still fires.**
`tsconfig.build.json` excludes `*.test.ts`, so a green typecheck does
not cover it. A scratch probe confirmed `tsc` still reports `TS2353` for
`onLegacyRead` and no error for the three real options.
- **All 48 construction sites outside the package compile unchanged.**
`tsconfig.check.json` also excludes `*.test.ts`, so the 25 webapp test
files were checked with the test exclusion dropped and compared against
the same check on the base commit: 614 errors before, 614 after, zero
present in one and not the other. Those 614 are pre-existing in
never-typechecked test files.
- `typecheck` passes for `run-store`, `run-engine` and `webapp`. `knip`
reports nothing in `run-store`.
## Also
Refreshes the sixteen stale `runOpsStore.ts` line references in
`runOpsStore.mixedResidency.test.ts`, each verified against the symbol
it names.
## Notes for the reviewer
- The riskiest possible mistake in this diff is a fan-out passing the
wrong order — the compiler cannot catch it, because both orders are
`readonly ShardKey[]`. The five `#precedence` sites are `#findRunsOpen`,
`findRunsByIdempotencyKeys`, `#collectManyWaitpoints`,
`findManyTaskRunWaitpoints` and `findManyWaitpointTags`. Those are the
lines worth the closest read.
- Four sites previously derived "the other store" by object identity
(`home === this.#new ? ...`). They now compare keys. The two are
equivalent: in single-database mode both keys map to the same store
object, and when the stores are distinct, identity and key comparison
agree.
- No changeset and no `.server-changes` note: the package is internal
and the one behaviour change is unreachable in production, so a release
note would tell a user nothing.
- Two CI checks fail for reasons that predate this branch and reproduce
on the base commit: `lint` (~16 unknown `react/*` rules make
`.oxlintrc.json` fail to parse, which disables oxlint entirely —
including the two `trigger-runops` fences) and `knip` (`unrun`, an
unused devDependency on the default branch). Both want their own fix.
## Summary
Adds a second generation of run-ops id, plus the resolver that reads a
store key straight out of an id. A gen-2 id keeps the existing
26-character layout, but the character at index 24 becomes a routing
shard key instead of a region code, and the version character at index
25 becomes `"2"`. Nothing mints gen-2 ids yet, so this is inert on
merge.
## Design
The version character is a single character, so the gen-1 and gen-2
shape checks can never both match. That is what makes the two
generations provably disjoint rather than disjoint by convention.
```ts
resolveShard(id) // gen-2 body -> its shard key, [a-z0-9]
// gen-1 v1 body -> "new"
// anything else -> "legacy"
```
`resolveShard` is total: it returns a key for any input string,
including an empty or malformed one, and never throws.
`classifyResidency` keeps its signature and its two values, and now
reports gen-2 ids as part of the dedicated family, so existing consumers
of that boolean are unaffected.
The body stays 26 characters rather than 27 deliberately. The older
27-character format is still in the wild and has to keep resolving to
legacy, and a longer gen-2 shape would need probabilistic disambiguation
against it. A rare misroute is not an acceptable property for a routing
key.
The one behavior change is that a 26-character body ending in `"2"` now
routes by its shard key instead of falling back to legacy. Two test
assertions pinned the old result and are updated here. A repository-wide
search confirms they are the only two of their kind.
Verified against the full run-store corpus (68 files, 370 tests) with no
test-file changes there, plus the run-engine residency and waitpoint
suites. No changeset: the new surface has no caller, so a version bump
would tell a user nothing.
Cuts CPU on the `engine/v1/worker-actions/*` routes a managed supervisor
calls, and adds the benchmark harness the numbers come from.
Measured on a local stack: **on-CPU per completed run 9.07ms → 6.59ms
(−27%)**, busy fraction 45.6% → 33.8%, with every worker-action p50 down
23–27%. Load was 5,000 runs / 24 virtual supervisors / 90s window /
30,120 requests / 0 errors.
Query-count work from the same investigation is deliberately **not**
here — it will follow as a separate PR.
## The three changes
**1. Split the event-loop monitor in two (~14% of on-CPU, plus ~5pp of
GC).**
`eventLoopMonitor.server.ts` installs a global `async_hooks` hook:
`init` writes a `Map` entry for *every* async resource the process
creates, `before` calls `process.hrtime()` and `context.active()` on
every one. Enabling any async hook also puts V8 on the slow path for
promise instrumentation process-wide. `EVENT_LOOP_MONITOR_ENABLED`
defaulted to `"1"`, so this was the shipping configuration.
The blocked-loop detector is now opt-in (`EVENT_LOOP_MONITOR_ENABLED`,
default `0`). The event-loop *utilization* gauge — a single interval
timer with no per-request cost — moves to its own flag
(`EVENT_LOOP_UTILIZATION_MONITOR_ENABLED`, default `1`) and stays on, so
the useful half survives without the expensive half.
A/B under identical load:
| | monitor on | monitor off | change |
|---|---|---|---|
| on-CPU per run | 9.08ms | 7.25ms | −20% |
| GC self time | 9.80% | 5.05% | −4.75pp |
| dequeue p50 | 76.6ms | 62.8ms | −18% |
| attempts/start p50 | 56.3ms | 43.5ms | −23% |
**2. Bucket route matching by first static path segment (10.4% → 3.9% of
on-CPU).**
`patches/@remix-run__router@1.23.3.patch` already memoized flattened
branches and compiled path regexes. What remained was the linear scan:
`matchRouteBranch` walked the ranked branch list calling `matchPath` per
branch across 521 route files, so every worker-action request paid a
scan proportional to the whole route table.
Branches are now indexed by their lowercased leading segment, with one
always-considered list for branches whose leading segment is dynamic,
splat or optional (and for root/pathless paths). A request walks only
its own bucket merged with that list. Route-matching self time dropped
64% (3.6s → 1.3s over a 90s window).
Ordering is preserved exactly: both lists hold indexes into the already
rank-sorted branch array and are walked in ascending-index order, so the
first match found is the same branch the full scan would have found.
Bucketing lowercases on both sides, so case-insensitive matching still
resolves and `caseSensitive: true` routes are still rejected by
`matchPath` itself. A pathname whose own leading segment can't be
bucketed falls back to the full scan.
Verified equivalent to the unpatched matcher over 20,050 pathnames
(literal, dynamic, splat, optional, case variants, basenames,
percent-encoded) with zero mismatches.
`apps/webapp/test/routeMatchingPatch.test.ts` pins the matching
semantics rather than the optimisation, so it still passes without the
patch.
**3. Demote per-heartbeat and per-dequeue `info` logs to `debug`.**
These are the two highest-rate engine calls and each wrote a synchronous
structured log line on every request. Synchronous `console` writes can
block the loop when stdout backs up, which costs more than the ~1.3% CPU
share suggests.
## The harness
Two benchmarks, neither in the default suite (they run for minutes,
attach the V8 profiler, and report numbers rather than assert on them).
See `apps/webapp/test/bench/README.md`.
- `apps/webapp/test/bench/engineHttp.bench.test.ts` — spawns a real
webapp against throwaway Postgres/Redis containers, seeds a production
environment with a promoted managed deployment, and drives a closed-loop
supervisor pool through the full lifecycle. Profiling runs over CDP
rather than `--cpu-prof` so it covers only the measured window instead
of being swamped by boot, and `performance.eventLoopUtilization()` is
sampled *inside* the webapp process.
-
`internal-packages/run-engine/src/engine/bench/runEngineLifecycle.bench.test.ts`
— drives `RunEngine` directly, profiling enqueue and lifecycle
separately so engine cost isn't mixed with request-stack overhead.
- `apps/webapp/test/bench/analyzeProfile.ts` — dependency-free
`.cpuprofile` analyzer that symbolicates through the build's source maps
and ranks CPU by package, self time and total time. Percentages are
shares of on-CPU time (V8's `(idle)`/`(program)` excluded).
`startWebapp` gains `overrideEnv`, applied after the worker-disable
defaults, so the HTTP bench can re-enable the run engine worker that
drains the master queue into the worker queues a supervisor dequeues
from.
The local OTel collector gains a traces pipeline. It only defined a
metrics pipeline, so pointing `INTERNAL_OTEL_TRACE_EXPORTER_URL` at it
locally failed and the webapp silently fell back to the console span
logger.
## Configuration
For operators upgrading:
- `EVENT_LOOP_MONITOR_ENABLED` (now defaults to `0`) — the
per-async-resource blocked-loop detector. Set to `1` to restore the
previous behaviour and keep emitting `event-loop-blocked` spans.
- `EVENT_LOOP_UTILIZATION_MONITOR_ENABLED` (new, defaults to `1`) — the
`nodejs.event_loop.utilization` gauge. Unchanged in behaviour; it just
has its own flag now so it survives turning the detector off.
## Notes for review
- `pnpm-lock.yaml` changes only because the router patch content
changed, which changes its patch hash.
- One thing the profile ruled out: with a real OTLP collector receiving
spans, tracing costs ~1.7% of on-CPU at 100% sampling and ~0.8% at the
production rate. Span shipping is not a hidden cost, so nothing here
touches it.
- Caveats on the numbers: a laptop, not production hardware, so DB and
Redis *latency* are unrepresentative (client-side CPU is what's ranked);
single webapp process; throughput varies ~5% run to run, which is why
the claims rest on on-CPU per run rather than req/s.
## Verification
- 20,050-pathname router equivalence check vs the unpatched matcher,
zero mismatches
- `apps/webapp/test/routeMatchingPatch.test.ts` (12 cases) passes
- webapp e2e smoke suite (68 tests) passes through the patched router
- run-engine suites covering the snapshot/attempt paths pass
- `typecheck`, `format`, `lint`, `knip` clean
Two small tweaks to the `Switch` primitive, so every variant and call
site picks them up:
1. **Track is 2px shorter.** `large` 44 → 42px, `medium` 32 → 30px,
`small` 24 → 22px. The checked thumb travel drops by the same 2px so the
thumb stays flush at both ends.
2. **Holding the switch down stretches the thumb into an oval** pointing
the way it's about to travel — rightwards when off, leftwards when on.
Pure CSS via `group-active:`, no new state or handlers.
The thumb's `transition` shorthand doesn't cover `width`, so it's now
`transition-[translate,width,background-color]` (same 150ms
duration/easing as before). `size-N` on the thumb became `h-N w-N` so
the press rule overrides the same `width` utility.
Verified in headless Chrome across all five variants in both states:
correct widths at rest, thumb flush at both ends, stretch grows the
right direction, and no overflow of the track.
<img width="266" height="108" alt="CleanShot 2026-08-21 at 10 16 14"
src="https://github.com/user-attachments/assets/ee95a399-0a40-48c4-a325-a1166b3bd88a"
/>
🤖 Generated with [Claude Code](https://claude.com/claude-code)
<!-- conductor-workspace-link -->
---
[Open workspace in
Conductor](https://app.conductor.build/workspace/c1ce8d0f-9ed2-4fbc-8084-a3989484cc53)
---------
Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>
## Summary
The platform notifications admin page can now save a notification as a
draft without committing to a schedule, then publish it later by
entering start and end dates. Drafts stay hidden from the webapp panel,
the CLI, and the "What's new" changelog until they are published.
## Design
A draft is an `isDraft` flag on `PlatformNotification`, not nullable
dates, so the existing index and every read query stay intact. All three
reader queries filter on the flag, so a draft can never surface
regardless of its placeholder dates. Publishing writes the real start
and end dates and clears the flag; the publish dialog validates the
range and shows inline errors. Editing a draft keeps it a draft, with
the schedule fields hidden until publish.
Also folds in a small tweak: the "Send preview to me" test button now
appears when editing a notification, not just when creating one.
## Summary
The Queue Metrics dashboard UI is gated by a per-org feature flag, so
there was no way to look at it for a real org without turning it on for
every member of that org. An admin impersonating into an org now sees
the metrics UI there regardless of the flag, so it can be checked
against real data before anyone else in the org sees it.
Nothing changes for a normal session: a member of an org whose flag is
off still gets the classic Queues page, and the gated sub-routes still
404.
## Design
The gate had no request and only resolved the org flag. It now takes the
request and resolves impersonation itself, rather than each caller
computing a boolean and passing it in, so the rule lives in one place
and a new call site cannot forget it. Seven call sites gate on this,
which is exactly why.
Two things narrow the bypass:
- It keys on **impersonation**, not `user.admin`. Impersonation is
scoped to one org and is deliberate; keying on admin status would
silently hand every admin the preview in their own day-to-day orgs.
- It yields to the **view-as-user** toggle. That toggle exists so an
impersonating admin can see what the member sees, and unreleased UI
leaking through it would make it lie. Suppressing a read-only view there
stays inside the display-only contract in `hasAdminDisplayAccess` (added
in #4421).
The bypass also stays behind the gate's existing org-membership lookup.
Since the acting user id is the impersonation target, that lookup is
what keeps the preview confined to the org actually being impersonated
into.
Verified end-to-end against a running instance across the matrix: member
with the flag off gets the classic view and 404s; the same org under
impersonation gets the metrics view and a 200; flipping view-as-user
returns it to the member's exact experience and back; and the flag-on
path is unchanged. An admin who is merely a member, not impersonating,
still gets the classic view.
One thing worth flagging: a few route comments say that with the flag
off no metrics reads fire. That remains true for every member session
and for the org as a whole, but an admin actively previewing does
exercise that org's real Redis and ClickHouse reads. That is inherent to
previewing, and bounded to one admin session.
## Summary
Follow-up to #4738. Splits the dashboard agent's base URL into two: the
instance that hosts the agent project (used for sessions), and the
instance the agent acts against as the user (used by its read-tools).
#4738 only needed the first, but moved the second along with it, which
breaks the tools when the agent runs on a different instance than the
webapp.
## Root cause
The agent's read-tools call the API as the logged-in user via a
delegated user-actor token. The webapp signs that token with its own
`SESSION_SECRET`, scoped to its own `userId` and `environmentId`, so it
can only be verified by, and only resolves the user's data on, that same
instance. #4738 routed the injected `apiOrigin` those tools use to the
agent's host instance, so the token no longer verifies and the data
isn't there.
## Fix
`dashboardAgentApiOrigin()` stays the agent's host instance (sessions,
task triggers, realtime, the `in` forward). A new
`dashboardAgentUserApiOrigin()` returns the webapp's own origin
(`API_ORIGIN ?? APP_ORIGIN`) and is injected into the run metadata the
tools use. Same-instance deployments resolve both to the same host, so
behavior is unchanged there.
## Summary
Lets the dashboard agent point at a specific Trigger instance instead of
assuming it runs on the same instance as the webapp. Adds an optional
`DASHBOARD_AGENT_BASE_URL`; when unset it falls back to the SDK default.
## Root cause
The agent's session start, token mint, head start, in-proxy and the
client transport all built the agent's base URL from the webapp's own
origin (`API_ORIGIN ?? APP_ORIGIN`). That only holds when the agent
project runs on the same instance as the webapp. When it runs elsewhere,
`DASHBOARD_AGENT_SECRET_KEY` belongs to that other instance, so the
webapp's own API rejects it with an "Invalid API key" and the chat can't
start.
## Fix
`dashboardAgentApiOrigin()` now returns `DASHBOARD_AGENT_BASE_URL` or
the SDK default, never the webapp origin. A concrete default (rather
than an unset value) keeps it independent of `TRIGGER_API_URL`, which a
webapp may point at a different host. Every server call site already
routes through that helper; the client transport reads the value from
the root loader via a new `useDashboardAgentBaseUrl` hook.
<!-- ccr-slack-attribution -->
_Requested by **Matt Aitken** · [Slack
thread](https://triggerdotdev.slack.com/archives/C045W9WM3E1/p1786741966214949?thread_ts=1786741966.214949&cid=C045W9WM3E1)_
`idempotencyKeys.reset()` now honours an explicitly passed `scope` even
when the key material happens to be 64 characters long.
**Before:** `resetIdempotencyKey` treated *any* 64-character string as
an already-computed hash and sent it to the API verbatim. That
short-circuit ran before the scope logic, so if your key material is
itself a 64-character digest (a common pattern when you hash your own
dedup identity) the `scope` you passed was silently discarded and the
un-hashed material went on the wire. The server stores the hash, so the
reset matched no run and returned 404 every single time. Key material of
any other length worked fine, which made this look arbitrary.
**After:** a 64-character key with an explicit `scope` is sent verbatim
first and, only when that attempt comes back a definitive not-found,
retried as the derived scope hash. Every call that worked before behaves
identically, and the previously impossible case now resolves on the
fallback.
## How
A 64-character string is forwarded unchanged, exactly as before, when:
- the idempotency key catalog recognises it (it came from
`idempotencyKeys.create()` in this process), or
- no `scope` was passed, so there is nothing to derive a hash from, or
- the scope hash cannot be derived (e.g. `scope: "run"` outside a task
context with no `parentRunId`).
Otherwise the key is ambiguous: it may be raw material the caller wants
hashed with the scope, or it may already be the stored hash. Reset sends
the verbatim value first because that is what every previous version
sent, so anything that resolved before still resolves with the same
single request, the same target run, and the same errors. The derived
hash is the new behaviour, so it only runs once the verbatim attempt has
failed with a 404, a definitive "no run under this key". Any other error
(a 503, a connection error) leaves the verbatim key's state unknown, and
resetting a different key on unknown state would be an untargeted write
the caller never asked for, so those errors surface unchanged. That has
an honest cost: when the endpoint answers 503 for a miss it cannot
confirm, the caller sees the 503 and retries rather than silently
falling through to the derived key. When both attempts miss, the
verbatim attempt's 404 is surfaced, again matching what previous
versions threw.
A side benefit of this order: a key from `idempotencyKeys.create()`
reset with a `scope` from a cold process resolves in a single request,
because the created key is itself the stored value.
`isIdempotencyKey` is deliberately left alone: it applies the same
length rule on the trigger path, but it is self-consistent there, and
changing it would invalidate already-stored keys.
The `attachedOptions?.key` / `attachedOptions?.scope` fallbacks below
the old guard were unreachable (every catalog entry is a 64-character
digest, so it always hit the short-circuit first) and re-deriving from
them produces the identical hash anyway. They are removed rather than
left as dead code.
---
## ✅ Checklist
- [x] I have followed every step in the [contributing
guide](https://github.com/triggerdotdev/trigger.dev/blob/main/CONTRIBUTING.md)
- [x] The PR title follows the convention.
- [x] I ran and tested the code works
---
## Testing
Tests in `packages/core/src/v3/idempotencyKeys.test.ts` drive the real
`resetIdempotencyKey` against a local HTTP server and assert on the
exact values that reach the wire, in order. Nothing is mocked. They
cover:
- 64-character material + explicit `scope` derives the global- and
run-scoped hash once the verbatim key misses (fails without this change)
- the verbatim key wins when runs exist under both the verbatim value
and the derived hash, so the pre-existing target is preserved
- keys from `idempotencyKeys.create()` are forwarded unchanged: catalog
hit, no scope, and scope with a cold catalog (the last now a single
request)
- a transient failure of the verbatim attempt surfaces its error without
ever touching the derived key
- error surfacing: a double miss reports the key the caller passed, and
a non-404 from the fallback is not swallowed
- ordinary short material is still hashed, and underivable run/attempt
scopes still send a 64-character key verbatim while still throwing for
shorter material
```
pnpm run test ./src/v3/idempotencyKeys.test.ts --run # 18 passed
pnpm run build --filter @trigger.dev/core # clean
pnpm run format && pnpm run lint # clean
```
---
## Changelog
`idempotencyKeys.reset()` now works when your idempotency key is itself
64 characters long. Previously any 64-character key was assumed to be
already hashed, so passing one along with a `scope` silently ignored the
scope and the reset never found a matching run.
---
## Follow-ups (not in this PR)
- `docs/idempotency.mdx` describes the `idempotencyKey` parameter of
`reset()` as "the 64-character hash string" in one place while showing
raw material plus `{ scope: "global" }` a few lines later. Worth
reconciling.
- No surface currently exposes the stored hash that the reset endpoint
matches on: `ctx.run.idempotencyKey`, the run page and the
`idempotency_key` query column all show the user-provided key. That is
what leads people to send a value reset cannot match.
---------
Co-authored-by: Claude <noreply@anthropic.com>
Co-authored-by: Matt Aitken <matt@mattaitken.com>
## Summary
Enforces `react/react-compiler` as an error for the webapp now that all
reported compiler diagnostics are fixed or narrowly scoped. Removes the
unused lazy-ref helper made obsolete by the ref initialization cleanup.
## Summary
Scopes React Compiler diagnostics to route statements where refs
intentionally coordinate virtualized views, live reload state, transport
lifecycles, and deferred callbacks. Other compiler diagnostics remain
active in those routes.
## Summary
Scopes React Compiler diagnostics to component and hook statements where
refs intentionally coordinate editors, animations, polling, deferred
callbacks, and other imperative integrations. Other compiler diagnostics
remain active in those components.
## Summary
Replaces render-time ref initialization with lazy state for frozen form
defaults, the tooltip's virtual positioning element, and the side menu's
first-paint visuals. Editable alert fields now update immutable state
snapshots.
## Summary
Scopes React Compiler diagnostics for component and hook effects that
intentionally synchronize with navigation, submissions, browser APIs,
streams, timers, or authoritative server values. Each suppression stays
on the reported synchronization call rather than disabling analysis for
the component.
## Summary
Derives controlled tab, tag, and checkbox values directly during render
instead of copying them through effects. Modal drafts now reset from
their open event, and the route-backed alert dialog renders open
immediately without a mount-time state update.
## Summary
Scopes state synchronization that intentionally resets editable drafts
from authoritative server values, deployment state, or programmatic
filter changes. These values cannot be derived during render without
removing user control between resets.
## Summary
Removes manual memoization where derived values are already rebuilt each
render, narrows the dashboard watch callback to a stable chat
identifier, and scopes two intentional memoization patterns that protect
local edits and serialized synchronization.
## Summary
Makes stable dashboard history refs explicit memo inputs and scopes the
remaining compiler diagnostics to callbacks whose local handlers or
lifetime-stable values cannot be represented accurately in dependency
arrays.
## Summary
Captures chat-history age when the menu opens so rerenders cannot change
labels mid-view. The waitpoint deadline form also reuses one intentional
wall-clock snapshot for all calculations in a render.
## Summary
Records when live metric responses arrive and uses that timestamp to
evaluate gauge freshness and waiting duration. Cached or failed
responses remain untrusted until revalidated, while rendered values stay
stable between polling updates.
## Summary
Derives session and API key expiry states from a timestamp captured by
each route loader. Every status on a page now uses one consistent point
in time instead of changing according to when an individual component
rerenders.
## Summary
Uses explicit bucket timestamps when rendering usage charts instead of
anchoring missing timestamps to the current render time. Tooltips now
remain stable across rerenders, and examples use a deterministic
timestamp.
## Summary
Keeps render inputs and shared regular expressions immutable. Grouped
selects now compute each section's shortcut offset directly from
preceding sections, which also makes numeric shortcuts follow the
displayed item order reliably.
## Summary
Calls dashboard hooks directly instead of passing them as ordinary
callback values, and subscribes to optional Ariakit stores through an
unconditional hook. This keeps hook ordering stable while preserving the
existing behavior when a provider is absent.
## Summary
Adds targeted lint suppressions for components built around libraries
that React Compiler intentionally declines to memoize, plus one
unsupported function-reference pattern. Each suppression is scoped to
the affected component so other compiler diagnostics remain actionable.
## Summary
Enables exhaustive React Hook dependency checking and resolves the
existing violations across the dashboard and React hooks package.
Effects and callbacks now track current values without introducing
request, subscription, or render loops.
## Design
Dependencies are included directly when the hook lifecycle should follow
them. Timers, Remix fetchers, and realtime subscriptions use stable
callbacks or latest-value refs where restarting work would change
behavior.
Unnecessary memoization was removed where ordinary derivation is
clearer. Full lint and typechecks for the webapp and React hooks package
pass.
## Summary
`goose up` against `internal-packages/clickhouse/schema` panics on
`main` today, so ClickHouse migrations cannot be applied from a fresh
checkout. Renumbering the external deployment id migration from 040 to
041 clears it.
## Root cause
Two migrations claim version 40.
[#4615](https://github.com/triggerdotdev/trigger.dev/pull/4615) added
`040_create_task_events_search_v2.sql`, and
[#4661](https://github.com/triggerdotdev/trigger.dev/pull/4661) added
`040_add_task_runs_v2_external_deployment_id.sql` a day later. #4661 was
opened before #4615 merged, so 040 was genuinely free at branch time,
and because the two files have different names there is no textual
conflict for git or a rebase to surface. Both merged green, and no
workflow in this repo runs `goose`, so the collision only shows up the
first time someone actually migrates.
goose parses the numeric filename prefix as the version and refuses
duplicates:
```
panic: goose: duplicate version 40 detected:
.../040_create_task_events_search_v2.sql
.../040_add_task_runs_v2_external_deployment_id.sql
```
It aborts while collecting the directory, before executing any SQL, so
nothing was half applied and there is no migration state to repair.
This migration gets renumbered rather than the `task_events_search_v2`
one because goose keys on the version number and not the filename:
version 40 is already recorded wherever 040 has been applied, so
renaming that file would re-run an applied migration.
Verified with a full `goose up` against ClickHouse 26.2.19.43 (the image
pinned in `internal-packages/testcontainers`): migrations apply cleanly
through version 41, and `task_runs_v2.external_deployment_id` lands as
`String DEFAULT ''`.
## Summary
Speeds up webapp test jobs by balancing measured work across runners,
reducing repeated container setup, and ensuring test workers release
shutdown resources promptly. Unit tests run across 24 duration-aware
shards, while E2E tests run across two balanced shards.
## Design
`RunEngine` shutdown now closes processing resources before support
resources, continues cleanup if one close fails, and reuses one shutdown
promise for concurrent callers. Redis workers clear completed shutdown
deadlines so finished tests no longer wait on idle timers.
Container-heavy suites are split only where it improves parallelism, and
repeated replication and engine fixtures are consolidated where one
end-to-end case provides coverage. Timing weights are refreshed for all
affected files.
Dependency installation overlaps container pulls, and both workflows use
WarpBuild's Node setup action.
<!-- ccr-slack-attribution -->
_Requested by **Iss** · [Slack
thread](https://triggerdotdev.slack.com/archives/C045W9WM3E1/p1787161814493949)_
**Before:** archiving a branch dropped the query string on the way back
to the branches list, so the list reset to page 1. Working down a long
list meant re-navigating to the page you were on after every archive.
**After:** you land back on the exact page you archived from, with
`page`, `search` and `showArchived` intact.
The archive action now redirects to the page the request came from
instead of rebuilding a bare branches path.
## How
The archive dialog already submits the page it was opened from as a
hidden `redirectPath` field (`${location.pathname}${location.search}`),
and the failure path already redirected to it — only the success path
ignored it and rebuilt the path with `branchesPath`/`branchesDevPath`,
which have no query string. Both paths now redirect to the submitted
path, run through the existing `sanitizeRedirectPath` helper to keep the
redirect same-origin (the same idiom used by
`resources.batches.$batchId.check-completion`).
## ✅ Checklist
- [x] I have followed every step in the [contributing
guide](https://github.com/triggerdotdev/trigger.dev/blob/main/CONTRIBUTING.md)
- [x] The PR title follows the convention.
- [x] I ran and tested the code works
---
## Testing
Three files change:
- `apps/webapp/app/routes/resources.branches.archive.tsx` — the fix.
- `apps/webapp/test/archiveBranchRedirect.test.ts` — new test that
drives the archive action and asserts the redirect `Location`: the query
string survives on both success and failure, and an off-origin
`redirectPath` falls back to `/`. Reverting the fix makes two of the
three cases fail, so the test covers the regression.
- `.server-changes/archive-branch-keeps-list-page.md` — release-note
entry, since this is a user-facing server-only change.
Also ran `pnpm run typecheck` and `oxlint` for `apps/webapp` — both
clean.
---
## Changelog
Archiving a branch now returns you to the same page of the branches list
instead of resetting it to page 1.
---
## Screenshots
_None — no visual change._
---------
Co-authored-by: Claude <noreply@anthropic.com>
## Problem
Every merge to main touching the agent queued a gated `staging`+`prod`
deploy that sat `pending` on a reviewer approval nobody grants
routinely. Because the gated runs never completed, they never drained
the concurrency queue and cancelled each other, so the Actions tab
filled with never-completing runs and the agent only ever actually
deployed via a manual dispatch + approval.
The reviewer gate bought nothing here: the agent deploys with
`--skip-promotion`, so a deploy lands **dormant** and nothing goes live
until the consuming webapp flips `DASHBOARD_AGENT_VERSION`. Promotion is
already a deliberate act (the env-var flip); gating the dormant deploy
on top of that just created the pile-up.
## Change
- **Remove the reviewer gate** by dropping the required-reviewers rule
on the `dashboard-agent-*` environments (repo-settings change, done).
The `environment:` key **stays** so the per-environment scoped deploy
token still resolves — no secret migration.
- **`workflow_dispatch` `ref` input** — deploy a specific commit SHA,
branch, or tag; defaults to the ref the run launches from. Checkout uses
`github.event.inputs.ref || github.sha`.
- **Require the ref to be an ancestor of `main`.** Constrains which
commit gets deployed to merged code only. A push is always main's tip
(passes trivially); a dispatched unmerged ref is rejected before the
deploy step. Because an explicit `ref:` checkout doesn't create
remote-tracking branches, `origin/main` is fetched explicitly before
`git merge-base --is-ancestor`.
- **`cancel-in-progress: false`** (kept). Cancelling the runner wouldn't
stop the remote build (it finishes server-side), and a superseding
concurrent deploy would race the same project's indexer. With the gate
gone, deploys are short, so a brief queue can't pile up.
- `max-parallel: 1` stays (parallel deploys of the same project race at
the indexer).
## Owner actions (repo settings — not in the diff)
1. **Remove required-reviewers** on `dashboard-agent-staging` and
`dashboard-agent-prod` — done.
2. **Add a deployment branch policy** on both environments restricting
deployments to `main`. This is the authoritative token guard:
`workflow_dispatch` runs the workflow file from the selected ref, so the
in-file ancestor check alone can't protect `TRIGGER_ACCESS_TOKEN` (a
branch could edit the check out). GitHub enforces the branch policy
server-side against `GITHUB_REF` regardless of file contents. With it in
place, the workflow only runs (and the token is only exposed) when
dispatched from `main`, and the in-file check then constrains the
independent `ref` input to merged commits.
## Pile-up root cause
The stacking was caused by the **reviewer gate** (runs waited forever,
so the queue never drained), not by `cancel-in-progress`. Removing the
gate is what fixes it; `cancel-in-progress` stays `false`.
Two defects that surface when a run parked on an external deployment id
gets pushed by a debounce key. Both were reproduced against a local
instance before being fixed.
## 1. The run is expired before it is due
```
now | status | statusReason | delayUntil | expiredAt
13:57:06 | EXPIRED | EXTERNAL_DEPLOYMENT_NOT_FOUND | 14:01:37 | 13:57:02
```
Killed 4m35s before its own scheduled start, blaming a missing
deployment.
**Why.** The park deadline is armed **once**, when the run is first
parked, from `max(now, delayUntil) + deadline`. Debounce pushes
`delayUntil` out afterwards and nothing re-arms it:
- `rescheduleDelayedRun` reschedules `enqueueDelayedRun:<id>`, not
`expireParkedExternalDeploymentRun:<id>`
- the redis-worker reschedule is an update-only `ZADD … XX`, and a
parked run has no `enqueueDelayedRun` job, so that call is a silent
no-op
Repeat triggers on one key walk `delayUntil` away from a deadline that
no longer moves. Once it crosses, the run dies while parked and not yet
due.
**Fix.** The expiry job already loads `delayUntil`, so it re-arms from
the current value and returns instead of expiring a run that is not due.
The guard lives in the expiry job rather than the debounce path
deliberately: it covers **every** caller that moves `delayUntil`, so a
future call site can't reintroduce this by forgetting to re-arm. It
stays bounded by the debounce max-duration contract, so a hot key can't
postpone expiry indefinitely.
## 2. The run reports itself as delayed while it is parked
```
RUN_CREATED | PENDING_VERSION | Run is waiting for a deployment of 'debounce-test-2'
DELAYED | DELAYED | Delayed run was rescheduled to a future date ← after one debounce push
```
The row stays `PENDING_VERSION`; the latest snapshot claims `DELAYED`,
so the run page describes a parked run as delayed. Happens on the
*first* push.
**Fix.** `rescheduleRun` hardcoded `DELAYED`/`DELAYED`. The snapshot
statuses are now supplied by the caller and **default to `DELAYED`**, so
the ordinary delayed path is byte-identical, and `rescheduleDelayedRun`
passes the parked statuses through when the run is parked.
## Reproducing
Repeated triggers on one debounce key against an id that hasn't landed:
```bash
curl … -d '{"options":{"externalDeploymentId":"x","debounce":{"key":"k","delay":"5m"}}}'
```
Three triggers correctly fold into one parked run; the defects show up
on the pushes.
## Testing
Two tests, each verified red before green and failing alone:
- a run whose delay was pushed past the deadline stays `PENDING_VERSION`
instead of expiring
- a debounce push on a parked run leaves a
`RUN_CREATED`/`PENDING_VERSION` snapshot, not `DELAYED`
`56 passed` across parking, pendingVersion, delayedRunSystem and
debounce; `43 passed` in `PostgresRunStore`. Typecheck, lint, format
clean.
## Notes
- Stacks on #4665, so it lands after the whole external-deployment-id
series.
- No changeset: this fixes unreleased behaviour introduced by the stack
below it, so no user has seen it.
- Both found by Devin's review on #4664, and both confirmed end to end
on a local instance before fixing.
Deployments page: an always-visible External ID column after Deployed
by, and an External ID row in the deployment inspector under Worker
type, both showing an en dash when a deploy carried no id. The Vercel
Linked column now renders before Git, still only when a Vercel
integration is connected. Also corrects the blank-row colSpan, which was
already off by one before this column existed.
Run inspector: an External deployment ID row between Version and SDK
version, read from the run annotations, so an operator can see which id
a run was pinned to - including a run that expired before its deployment
ever arrived, where the locked version is empty but the id is the whole
story. Buffered runs read the id from the same annotations rather than
reporting none.
Long ids are head-truncated with the full value behind the copy button:
a commit SHA is meaningful in its prefix, and the inspector panel can be
narrowed to 250px, where an unbroken 40-character SHA would otherwise
scroll the properties list sideways and push the copy button off-panel
(TRI-12923, TRI-13000).
The SDK discovers an external deployment id at runtime (explicit
TRIGGER_EXTERNAL_DEPLOYMENT_ID always; platform commit-SHA variables and
generic fallbacks when TRIGGER_AUTOMATIC_SKEW_VERSION_PROTECTION=1) and
sends it alongside lockToVersion; the server resolves precedence
(version > external id > current). An id held by a deployed deployment
pins the run to that worker; an in-flight or unknown id parks the run in
PENDING_VERSION with the id in TaskRun.annotations, wakes it pinned when
a deployment carrying the id finalizes (ClickHouse candidates, Postgres
authoritative), and expires it after a deadline that re-checks Postgres
before acting. Parking outranks delaying and preserves delayUntil. The
id is projected to ClickHouse task_runs_v2.external_deployment_id during
replication. Redis cache for id-to-worker resolution, guarded
version-aware writes.
Ids are not unique. Several deployments can hold one id - a --force
rebuild is the ordinary way to get there - so resolution always picks
the highest version among the candidates, never the newest by timestamp.
The rule is applied identically on both paths that can bind a run to a
worker: resolveExternalDeployment at trigger time, and
PendingVersionSystem when a landing deployment wakes a parked run.
Version comparison is numeric on the counter half, so 20260807.10
outranks 20260807.9.
A run whose id never lands expires at the deadline with
EXTERNAL_DEPLOYMENT_NOT_FOUND and an error naming the id it waited for,
which is what a failed build or a typo looks like from the caller.
Default deadline is one hour (EXTERNAL_DEPLOYMENT_PARK_DEADLINE_MS).
Debounce registration happens in both the parked and the delayed branch
through one helper, so a debounced run that parks still binds its
debounce key; without it every later trigger for the same key created
another parked run, and all of them executed when the deployment landed.
The two DELAYED-only status checks in DebounceSystem also accept
PENDING_VERSION, without which the lock-contention fallback would
rethrow a 5xx the SDK retries and amplifies, and the fast path would
push every trigger on a parked key through the redlock.
Resolution is skipped in development. A dev environment cannot hold a
WorkerDeployment - trigger dev registers a BackgroundWorker with nothing
behind it, and deploy --env refuses dev - so an external deployment id
there could only ever park, and the parked run then expired against the
dev TTL while a connected dev worker sat idle. The id is still annotated
so the dashboard shows what the app sent (TRI-13000).
A deploy can carry an opaque external id (commit SHA, CI run id, release
tag). Repeating an id that already deployed returns the existing version
as a no-op instead of rebuilding; an id with a build in flight is
rejected with 409 naming that version; a failed id rebuilds freely.
--force is non-destructive to deployments that already succeeded - both
persist and the higher version wins - but cancels a build still in
flight, so one id never has two live builds racing to define it.
Cancelling writes a terminal status and appends a finalized event, which
aborts a build the platform drives; a build it does not drive keeps
running but can never land, and the CLI says so. Ids are deliberately
not unique - reuse is resolved in application code by highest version,
never timestamps. The no-op path mints no build credentials and no event
stream (TRI-12923).
What that means for callers: a --force rebuild leaves two deployments
holding one id, and runs triggered with it go to the higher version once
the rebuild lands, so the takeover needs no separate promotion. Until a
successful build exists for an id, runs triggered with it park and then
expire rather than falling back to current - a failed build is therefore
visible to the caller as expired runs, not as runs on the wrong release.
An external deployment id is an opaque, caller-chosen name for a release
- a commit SHA, a CI run id, a release tag. This adds the shared
contract that both halves of the feature read, and nothing else: no
deploy writes one yet and no trigger sends one.
ExternalDeploymentId is defined once and reused by
InitializeDeploymentRequestBody.externalId and
TriggerTaskRequestBody.options.externalDeploymentId, so a value accepted
by one half can never be rejected by the other. A value that is blank
once trimmed is treated as absent rather than rejected, so an unset CI
variable expanding to an empty string is not a 400. The 128 character
limit fits a SHA-256 commit hash with room for composite ids, and
EXTERNAL_DEPLOYMENT_ID_MAX_LENGTH is the single source of truth that the
request schemas and the CLI both read.
RunAnnotations.externalDeploymentId records the request, not the
outcome: lockedToVersionId and taskVersion are overwritten when a run
locks, whereas this stays true forever, and it can carry the pin for a
run parked before its deployment exists.
Also lands the runtime discovery helpers as pure functions over an
environment reader: the explicit TRIGGER_EXTERNAL_DEPLOYMENT_ID
variable, the platform and CI commit-SHA table, and the
TRIGGER_AUTOMATIC_SKEW_VERSION_PROTECTION gate. Nothing calls them yet.
refs TRI-13000
Migrations only, no code reads them yet. Postgres: nullable non-unique
externalId on WorkerDeployment plus a CONCURRENTLY-built (environmentId,
externalId) index in its own migration file. ClickHouse:
external_deployment_id String DEFAULT '' on task_runs_v2 (plain String,
not LowCardinality - commit SHAs are high-cardinality). Part of task run
version skew protection (TRI-12998).
## Summary
Move tree selection onto semantic tree items and use native expansion
buttons.
Dashboard and story tree rows now share mouse and keyboard selection
through `getNodeProps`. Expand and collapse affordances are named
buttons instead of clickable layout elements.
Base: [#4700](https://github.com/triggerdotdev/trigger.dev/pull/4700)
nicktrn