ensureHandle was doing a post-hoc `if (h.draft === undefined) h.draft = baseline`,
which relies on the saveInitialValue=false skip to swallow that seeding
write. Hand the baseline to `UserDraft.use({ defaultValue })` instead —
useLocalStorageValue uses it as the initial $state value when localStorage
is empty, so lastSerialized is correct out of the gate and no setter call
is needed.
Earlier commits in this PR wired the resource and variable editors to a
single multi-workspace bundle stored under the user's session workspace
key — which mixed workspaces in one localStorage entry and required a
custom multi-key fix-up pass to persist edits for other workspaces.
Reset both editors to their pre-PR shape and apply the minimal change:
the per-workspace `Record<string, ResourceState>` (resp. `VariableState`)
becomes `Record<string, UserDraftHandle<…>>`, with one handle per
workspace created via `UserDraft.use(…, { workspace: ws })`. The handle
keys its own localStorage entry under that workspace, so cross-workspace
edits stay cleanly separated and reactivity flows through the handle's
`draft` accessor — `bind:` on form fields just works.
Adds `manualRelease: true` + `handle.release()` to `UserDraft.use` so
the editors can register handles lazily inside an effect (Svelte 5
forbids `onDestroy` outside component init). The editors register a
single top-level `onDestroy` that releases every collected handle.
After a successful save, the per-workspace autosave is cleared via
`UserDraft.remove(itemKind, path, { workspace })`.
Apps and flows used to autosave under un-scoped keys (`flow`/`flow-{path}`,
`app`/`app-{path}`, `rawapp`/`rawapp-{path}`) with a base64-encoded
state envelope. This adds a one-off migration that rewrites surviving
legacy entries under the workspace-scoped `userdraft/w/{ws}/{kind}/{path}`
keys with the new `{ value }` wrapper, transforms the payload where the
shape differs (drops the flow view-state envelope, defaults the new
raw-app `summary` field), and drops the source key.
The migration lives in its own file (`userDraftLegacyMigration.ts`)
so the new UserDraft service stays free of legacy decoders. Idempotent
via a `userdraft/legacy_migrated_v1` sentinel; runs from the logged-in
root layout once a workspace is known. Defensive shape checks avoid
clobbering co-resident apps that happen to use the same key prefixes.
Migrates the flow, app, and raw_app editor routes to the same
`LocalDraftStaleModal` flow already used by scripts: compare the
recorded meta against the current `version` / `versions[last]` and
`draft_created_at`; on mismatch, surface the choice in a modal.
Adds `UserDraft.saveMeta` for routes that don't hold a live handle
(the app editor reads via `UserDraft.get` and the handle lives in
the child `AppEditor` component). It writes meta directly to
localStorage and tolerates the no-entry case.
Replace the script editor's toast-based "Discard / Show diff" pattern
with a dedicated modal that surfaces *why* the local autosave is out
of date: a new DB draft on the server, or a new deployed version.
Adds `checkStaleness` (UserDraftMeta vs current backend revs, draft-rev
priority) and a `setMeta({ force: true })` mode so the "Keep current
draft" acknowledgement persists even when it happens to be the
entry's first state mutation — under `saveInitialValue: false` an
ack-only setMeta would otherwise be skipped and the modal would
re-fire on next mount.
The modal lives at LocalDraftStaleModal.svelte; the script editor
wires it as a template for the remaining editors. Other editors
(flows, apps, raw_apps, resources, variables, triggers) still use
the previous toast pattern and will be migrated in follow-up
commits.
Extends StoredDraft<V> with two optional rev fields used by the
forthcoming staleness modal:
- remoteRev — the deployed version's id/hash/timestamp at the moment
the local draft was created. Compared against the latest deployed
rev on reload.
- remoteDraftRev — the DB-draft created_at at the moment the local
draft was created. Only meaningful for kinds that have a DB draft
(script, flow, app, raw_app). Checked first so a teammate's draft
push is detected before the "deployed version moved" case.
API additions on the handle returned by UserDraft.use():
- handle.meta — read the rev metadata currently stored.
- handle.setDraftAndMeta(value, meta) — atomic write of value + meta in
a single state.val assignment. Editor routes use this on load so the
baseline rev rides along with the value without consuming the
saveInitialValue=false dedup slot twice.
- handle.setMeta(meta) — update just the rev metadata after the user
picks "Keep current draft" in the staleness modal.
- handle.draft = X — unchanged surface; now preserves existing rev
metadata across user edits.
Plus UserDraft.getMeta() and UserDraft.save() preserves any persisted
rev metadata when called without a live handle.
7 new tests cover the metadata surface; all 35 pass.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Variable
- Add `edited_at TIMESTAMPTZ NOT NULL DEFAULT now()` + `edited_by VARCHAR(50)` to the `variable` table (parity with `resource`); set them on INSERT and on every UPDATE.
- Surface them on `ListableVariable` so `getVariable` / `listVariable` return them.
DB drafts (script, flow, app/raw_app)
- The `*WithDraft` endpoints now also return `draft.created_at` as `draft_created_at`. The draft value alone wasn't enough to tell whether a teammate (or another tab) had pushed a fresh draft while local autosave was in flight; the new field is the staleness signal.
- Wired in `get_script_by_path_w_draft` (`ScriptWDraft.draft_created_at`, including the `prefetch_cached` forwarding), `get_flow_by_path_w_draft` (`FlowWDraft.draft_created_at`), and `get_app_w_draft` (`AppWithLastVersionAndDraft.draft_created_at`). OpenAPI updated to match.
The frontend will read these in a follow-up to implement the local-draft staleness check; this commit only widens the API surface.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Conflict resolution kept the UserDraft persistence on this branch and
incorporated the picker-navigation improvements from main:
- editPathFor / invalidate calls for workspace picker support
- loadFlowToken / loadAppToken / loadScriptToken stale-load guards
- onNavigate prop wiring
Dropped the legacy localStorage `flow-{path}` / `rawapp-{path}` /
`app-{path}` autosave paths that main still carried — they are
superseded by UserDraft (key `userdraft/w/{ws}/{kind}/{path}`).
Dropped main's `decodeState`/window.location.hash autosave for the
script editor for the same reason. The replaceStateFn prop that main
re-added to ScriptBuilder is also dropped (no longer exists on the
component).
Pre-existing typecheck errors in InstanceSetting.svelte and
useJobsLoader.svelte.ts come from main and reflect a stale
$lib/gen client — unrelated to the merge.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
localStorage entries now look like {"value": <draft>} instead of just
<draft>. The wrapping is invisible at the API boundary — UserDraft.use,
.save, .get, .remove all still operate on the unwrapped draft value —
but it leaves room to add metadata (timestamps, originating user,
schema version, ...) later without breaking existing entries.
Internals:
- StoredDraft<V> = { value: V } is what we serialise to localStorage
and what useLocalStorageValue's $state holds.
- wrap()/unwrap() helpers gate the boundary; the handle returned by
use() unwraps on get and wraps on set.
- readPersisted() defensively drops entries whose payload isn't a
{ value: ... } object, so pre-migration drafts written by earlier
commits on this branch are simply ignored (has() returns false,
get() returns undefined) rather than confusingly surfacing as
undefined-shaped drafts.
Test data switched from { value: X } (which collides confusingly with
the wrapper shape) to plain primitives / objects, plus a regression
test for the pre-migration ignore behaviour. 28 tests pass.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
The schedule_ prefix grouped all the trigger editors under what looked
like a "scheduler" namespace; trigger_ is what these actually are
(triggers — including the cron-style schedule). Mechanical rename
across UserDraftItemKind, every trigger editor's UserDraft.save/get/
remove calls, and the one test that asserted on the localStorage key.
Behaviour-only impact: existing localStorage keys under
userdraft/w/{ws}/schedule_{kind}/{path} from older builds will be
ignored on next open (no schema migration). Users will lose any
unsaved trigger drafts persisted before this change.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Same pattern, keyed on schedule_schedule. ScheduleEditor doesn't track
an originalConfig (its saveDisabled doesn't compare against a baseline)
so ordering is simpler — initialConfig snapshotted from backend, local
autosave overlaid after.
This completes UserDraft wiring across all 11 trigger editors.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Same pattern: persist getSaveCfg() on edits, overlay local autosave
in openEdit (with initialConfig/originalConfig snapshotted from
backend first), drop the entry on successful update.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Same pattern as the Kafka trigger: persist getSaveCfg() on every edit,
overlay any local autosave on top of the backend value when openEdit
loads the trigger (with initialConfig/originalConfig snapshotted from
backend first so hasChanged correctly reports the overlay as unsaved),
drop the entry on successful update.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Same pattern as the Sqs trigger: persist getSaveCfg() on every edit,
overlay any local autosave on top of the backend value when openEdit
loads the trigger, drop the entry on successful update.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Persist the trigger's getSaveCfg() output to
userdraft/w/{ws}/schedule_sqs/{path} on every edit, overlay any
existing local autosave on top of the backend value when openEdit
loads the trigger, and clear the entry on successful update.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
loadScript() assigned the backend value to scriptHandle.draft and then
deep-mutated parent_hash on the next line. Under
useLocalStorageValue's saveInitialValue=false contract only the very
first write is the baseline — the parent_hash mutation right after
counted as a second write and was persisted to localStorage, so
opening an existing script would silently write a draft entry even
though the user hadn't touched anything.
Combine `parent_hash` (and the topHash override) into a single
bakedBaseline so each branch of loadScript performs exactly one
assignment to scriptHandle.draft. Mirrored across the local-autosave
branch's discard callbacks too.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* feat: add EditableInput component
* feat: add EditorHeader for flow editor with file picker entry point
* feat: WorkspaceItemPicker for editor header navigation
* feat: clickable breadcrumb in EditorHeader, scoped picker keyboard nav
* fix: reload flow on URL change and reset highlight in search mode
* feat: editor header layout polish and trigger removal
* feat: apply EditorHeader to script, app, and raw app editors
* fix: show generated initial path for new apps in EditorHeader
* fix: align EditorHeader new-app path with draft drawer's Path component
* fix: read page.params.path in loadApp to reload on URL change
* fix: remount AppEditor when navigating between apps
* fix: clear app/files on URL change so editor remounts with fresh data
* fix: route picker selections to /apps/edit or /apps_raw/edit based on raw_app
* fix: drop Save button from path popover; defer rename to deploy
* refactor: drop pathPopoverContent snippet, bind newEditedPath directly
* refactor: dedupe editor header plumbing (editPathFor, userPathPrefix, breadcrumb snippet)
* fix: freeze breadcrumb during path edit so popover doesn't drift
* fix: drop spinner from path dependency-check; render nothing when no usages
* fix: swallow 404 in checkFlowOnBehalfOf so renaming a flow doesn't toast
* Revert "fix: swallow 404 in checkFlowOnBehalfOf so renaming a flow doesn't toast"
This reverts commit 82dec462ae.
* refactor: drop moveRenameManager dep from EditorHeader; pass onBehalfOfEmail as prop
* refactor: replace breadcrumb-snapshot effect with open/close setter
* refactor: drop unused dirtyPath state from EditorHeader
* fix: surface Path validation error in pen popover
* fix: decouple Path validation error from hideFullPath toggle
* refactor: use InputError for path validation message (slide transition)
* fix: re-derive Path meta from external path changes (sibling sync)
* docs: note Path's meta could be replaced with function-form bindings
* fix: 'Exit & see details' uses deployed path, not live store
* fix: undo/redo shortcut uses shiftKey instead of fragile case-match
* fix: type errors and keep edit pen visible while popover is open
* chore: remove unused meltComponents/Accordion wrapper
* fix: focus search input on picker open via popover openFocus selector
* fix: refocus picker search input on every popover open
* fix: pre-seed picker loaded state from cache so accordion opens at the right place
* fix: breadcrumb and picker track savedPath, not draft-renamed live path
* fix: inject current draft item into picker so breadcrumb scope isn't empty
* docs: add component-level and prop-level docs to EditableInput
* feat: warn that a deploy is needed when path is edited on a saved item
* feat: show same path-change-needs-deploy message in flow/script settings
* fix(flows): persist draft-renamed path through reload and dirty check
* fix: include path in unsaved-changes diff so renames trigger the modal
* feat: nested folders in picker tree and breadcrumb
* fix: per-segment popover state so switching breadcrumbs closes the previous one
* refactor: replace accordion picker with drill-through picker
* refactor(picker): review fixes, drill polish, and breadcrumb collapse
* fix(picker): review fixes — banned bindable, drop sibling-sync, load races, mouse highlight
* fix(picker): smooth-scroll highlighted row into view on open
* fix(picker): second-pass review fixes — load races, cache invalidation, breadcrumb/picker a11y, raw_app routing
* fix(RowIcon): apply size prop to resource_type and fallback divs
* feat(picker): add cross-kind 'All' root; deeper breadcrumb segments open there
* fix(picker,editor): third-pass review — search loading state, pen autofocus, allow empty summary, a11y
* fix(picker,editor): fourth-pass review — own check, flowbuilder arrows, editableinput double-save, customui.path gate, invalidate races
* fix(editor,picker): codex review — fresh URL state per load, granular whitelabel topBar gates
* fix(flows): clear localStorage in auto-reload to break URL-state loop
ResourceEditor and VariableEditor can stage edits for several target
workspaces in a single drawer session (see deployTo / states[ws] map).
The previous UserDraft wiring only persisted states[$workspaceStore] —
the user's session workspace — so any edit made under a different
target workspace tab disappeared on refresh.
Persist the entire `states: Record<wsId, State>` bundle as the draft
value instead. On lazy-fetch we pick the local state for that ws if
present and divergent from the backend; on bootstrap for new
resources/variables we restore states for every workspace the user
had staged. The localStorage key still lives under the user's session
workspace via UserDraft, but its contents now cover all target
workspaces from that session.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
The first time a value flows into a UserDraft.use() handle — typically
the editor route loading the backend value via flowHandle.draft =
backendFlow — is the baseline, not a user edit. Persisting it on the
spot puts a copy of the backend into localStorage on every page open
and produces spurious "local autosave" toasts on next visit when the
serialization round-trips differently.
useLocalStorageValue now takes options.saveInitialValue (default true,
backward compatible). When false, the first time the serialised form
of the state changes — via the setter or via a deep mutation — the
lastSerialized cache is updated but localStorage is not touched. Every
write after that persists normally. UserDraft.use() passes false.
Tests updated to reflect the new contract (first write is the
baseline) and a regression test added for the second-write-persists
behaviour.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
The /flows/add and /flows/edit routes drive FlowBuilder from a flowStore
whose getter reads flowHandle.draft directly. Calling UserDraft.remove
synchronously before goto() therefore wiped the in-memory entry, made
flowStore.val collapse to emptyFlow(), and tripped
UnsavedConfirmationModal against the just-saved value — even though the
deploy/save-draft itself succeeded.
Drop those explicit removes in onSaveInitial, /add onDeploy, and
/edit onDeploy. The empty-path entry self-cleans on unmount via
onDestroy ref counting; for the non-empty edit path the next visit's
load-time diff will silently overwrite localStorage when the local
autosave matches the deployed value. Restore-draft/restore-deployed
keep their explicit remove because they navigate to the same route
(no modal) and loadFlow immediately rehydrates the handle.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Setting a typed instance setting (e.g. `object_store_cache_config`) to
`null` via `PUT /api/settings/instance_config` was silently ignored:
`Json<InstanceConfig>` deserialized null to `None`, and `to_settings_map`
then dropped it via `skip_serializing_if = "Option::is_none"`, so
`diff_global_settings` (Merge mode) never saw the deletion.
Add `#[serde(remote = "Self")]` to `GlobalSettings` plus a manual
`Deserialize` impl that captures top-level explicit nulls into `extra`,
which already round-trips correctly via flatten. The handler stays
fully typed.
Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* fix(frontend): branch download UI on shouldDownloadViaClient instead of intercepting in onclick
When OpenAPI.TOKEN is set, several download links rendered an `<a href>`
to the API and relied on an `onclick` handler to call `e.preventDefault()`
and route the request through `downloadViaClient`. This is fragile in
embedded contexts (e.g. the whitelabel React SDK) where Svelte's hydrated
event listener may not intercept the click in time, so the browser
follows the unauthenticated `href` straight to the API.
Mirror the drawer pattern already used in `LogViewer` and
`FlowStatusViewerInner`: render a `<button>` calling `downloadViaClient`
when `shouldDownloadViaClient()` is true, and fall back to the plain
`<a href download>` otherwise. Affects the LogViewer top bar, the large-
result download in DisplayResult, the inline S3 link in ObjectViewer,
the CSV link in ParqetCsvTableRenderer, and FileDownload.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* fix(frontend): treat custom HEADERS / basic auth as token equivalents in shouldDownloadViaClient
Customers wiring the SDK with cookie-bypass auth via custom request
headers (e.g. `OpenAPI.HEADERS = getAuthHeaders()` returning a Bearer
header) had `shouldDownloadViaClient()` return false because it only
checked `OpenAPI.TOKEN`. The plain `<a href download>` branch then
followed the link without those headers, so authenticated downloads
silently degraded to the cookie path (or failed when there is no cookie).
Widen the check to any non-cookie auth: TOKEN, HEADERS, or USERNAME
(basic auth). Route `downloadViaClient` through the generated client's
`getHeaders` so all configured auth schemes are applied consistently
instead of hand-building an Authorization header for TOKEN only.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* docs: add global ai mode plan
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
* feat: add global ai draft mode
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
* fix: scope global ai mode to scripts and flows
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
* refactor: simplify global ai workspace item shape
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
* refactor: split global ai write tool into per-type tools
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
* feat: add global ai schedule and trigger workspace item tools
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
* feat: add dev-only /global_drafts route to inspect ai draft store
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
* feat: add edit_script and patch_flow_json global ai tools
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
* feat: add deploy_workspace_item global ai tool with confirmation
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
* feat: emit open-resource action card after deploy_workspace_item
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
* feat: add delete_workspace_item global ai tool with confirmation
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
* chore(system_prompts): emit RESOURCES_BASE and resource/variable zod schemas
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
* feat: add global ai resource and variable workspace item tools
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
* fix: search_resource_types uses listResourceType to avoid embedding feature dep
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
* Revert "fix: search_resource_types uses listResourceType to avoid embedding feature dep"
This reverts commit 6d1d19514a.
* feat: emit open-resource action card for variable and resource deploys
* feat: add global ai raw app workspace item tools
* feat: split raw-app prompt into chat-only authoring and cli prefix
* feat: add init_app global ai tool to scaffold raw apps from templates
* fix: pass write_flow value as JSON string for gemini compat
* refactor: hoist countExactMatches and applyExactReplace to chat/shared
* refactor: extract editableFlowJson module shared with global mode
* fix(global): preserve flow schema and groups across draft and deploy
* feat: extract inline scripts from flow reads and patches in global mode
* refactor: add findAndReplace helper for match-validated text patches
* refactor: extract getInlineRunnableContent helper for app file tools
* refactor: extract assertNotGeneratedAppFile guard for /wmill.d.ts
* feat: gate global ai mode behind localStorage flag for dev rollout
* chore: bump svelte to ^5.55.5 in raw app template (sync with main)
* fix: isolate global ai draft rollout
* fix: preserve global ai deploy metadata
* fix: harden global ai draft tools
* chore: remove global ai plan doc
* fix: align raw app prompt guidance
---------
Co-authored-by: Claude Opus 4.5 <noreply@anthropic.com>
json_to_document cast any i64 to u64 via `as` and wrapped in PosInt.
Negative integers in a tool's JSON schema (e.g. "default": -1,
"minimum": -1) became huge unsigned values like 18446744073709551615,
which Bedrock rejected with `tools.N.custom.input_schema: int too big
to convert`.
Try as_u64 first to preserve full u64 range, fall back to as_i64 for
negatives via NegInt, then as_f64.
Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* [ee] fix(license): bump ee-repo-ref to fix CE build
#9089 introduced a call from `windmill-api-settings/src/ee.rs::validate_license_key`
(license-gated) to `windmill_common::ee::compute_instance_hash` (enterprise-gated).
CE builds compile the caller (private + license) but not the callee, breaking the
docker image build on main with E0425 since #9089 landed.
Companion EE PR drops the enterprise gate from `compute_instance_hash`. The
function only depends on common state (BASE_URL, get_instance_uid,
canonical_base_url) and the parent `ee` module is already gated by `private`.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* chore: update ee-repo-ref to f9494c6320bb5fd07c1e9e09734b7fd5fbe7aa38
This commit updates the EE repository reference after PR #569 was merged in windmill-ee-private.
Previous ee-repo-ref: 0f495da1ef984cad22342c90691c0dc8b0a20faa
New ee-repo-ref: f9494c6320bb5fd07c1e9e09734b7fd5fbe7aa38
Automated by sync-ee-ref workflow.
---------
Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Co-authored-by: windmill-internal-app[bot] <windmill-internal-app[bot]@users.noreply.github.com>
* [ee] feat(license): offline (URL-bound) license keys
Offline keys are a 4-segment variant for air-gapped customers — no
phone-home, embedded seat/CU caps, locked to the instance's base_url.
Existing 3-segment online keys are unchanged.
Companion PRs:
- windmill-labs/windmill-ee-private (full design + EE impl)
- windmill-labs/windmill-customer-service (issuance + portal)
- windmill-labs/windmill-cf-worker-keygen (signing)
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* [ee] refactor(license): bind offline keys via instance hash; simpler CU enforcement
- /settings/license_status now surfaces an `instance_hash` superadmins share
with support when requesting an offline key
- OfflineMetadata: `hash` replaces `base_url`; OfflineCapStatus reports
`current_cu` (last 2min) and drops the grace-period fields
- verify_license_key now takes a db so EE can recheck the hash
- InstanceSetting.svelte: hash copy-block + simpler status panel
- Bump ee-repo-ref
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* [ee] chore(license): bump ee-repo-ref
Pulls in the current_cu clamp + prod public key restoration.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* [ee] refactor(license): split instance_hash endpoint; minimal cap UI; restore workers expiry toast
- `instance_hash` is no longer part of /settings/license_status responses; it
lives at GET /settings/instance_hash (super-admin only) so it isn't re-emitted
on every status poll. The UI doesn't show it — admins fetch it explicitly when
requesting a key from support.
- InstanceSetting offline cap UI is now two compact green/red status lines
(Seats X.X/Y and CUs X.X/Y) placed above the action buttons, matching the
existing "Latest key renewal" badge style. The block-panel is gone.
- "Latest key renewal" line and the "Renew key" button are now hidden when an
offline key is loaded (renewal is server-disabled for offline keys).
- Restore parseLicenseKey + checkLicenseExpiration toast on /workers
(works for both 3- and 4-segment keys).
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* [ee] chore(license): bump ee-repo-ref
Pulls in the plain-SHA256 instance hash + stats_ee revert.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* [ee] chore(license): bump ee-repo-ref
Picks up the alert wording change.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* [ee] chore(license): bump ee-repo-ref
Picks up the instance_uid cache so the periodic verify_license_key cycle
no longer hits global_settings.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* [ee] refactor(license): rename /settings/license_status → /offline_license_status
The endpoint was only used by the offline-license UI; the other fields it
returned (license_key_id, license_key_valid, kind, offline metadata) were
unused. Rename to clarify scope and flatten the response — it now returns
just the OfflineCapStatus (or null when no offline license is loaded).
Frontend uses `offlineCapStatus != null` as the "is offline" check.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* [ee] fix(ci): regenerate sqlx cache for the inline worker_ping query
After reverting unused stats_ee helpers (fetch_worker_pings*), the
inline `sqlx::query_as!(WorkerPingRecord, ...)` in get_stats_payload
lost its cache entry — CI's check_ee_full + cargo_test were failing
under SQLX_OFFLINE=true with E0282 type-inference errors.
Re-running update_sqlx.sh regenerates the cache file under its
current hash and prunes a couple of stale entries.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* [ee] fix(license): address cubic-bot review
- get_offline_license_status: propagate enforce_offline_caps errors as 500
instead of swallowing into a "no offline license" (Option::None) response
- canonical_base_url: rewrite the doc to match the actual fallback behavior
(lowercase + trailing-slash strip on URL parse failure); the original
cross-service contract is gone since the customer-service no longer
canonicalizes (treats the instance hash as opaque)
- check_seat_cap_for_new_user: take an email and short-circuit when the
email is already in `usr ∪ workspace_invite` so net-zero invite upserts
and invite→user transitions aren't spuriously blocked at cap. Mirrors
the dedup rule the count itself uses.
- Bump ee-repo-ref to pull in the EE-side change
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* [ee] chore(license): bump ee-repo-ref
Picks up the exact-delta seat-cap check (replaces the simple existence
short-circuit). Regenerates the new sqlx cache for the bool_and query.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* [ee] fix(license): propagate get_instance_hash errors; bump ee-repo-ref
- get_instance_hash: replace `.ok().flatten()` with map_err+? so DB errors
during instance_uid lookup surface as 500 instead of silently returning
`{"instance_hash": null}` (same pattern get_offline_license_status already uses)
- Bump ee-repo-ref to pull in the enforce_offline_caps cached-state preservation
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* chore: update ee-repo-ref to c6cd1afe2d9e04809b30751cd1687b28a65e62b1
This commit updates the EE repository reference after PR #566 was merged in windmill-ee-private.
Previous ee-repo-ref: a6d91016ae0d43c46604313aecae3aa9c778c8e0
New ee-repo-ref: c6cd1afe2d9e04809b30751cd1687b28a65e62b1
Automated by sync-ee-ref workflow.
---------
Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Co-authored-by: Ruben Fiszel <ruben@windmill.dev>
Co-authored-by: windmill-internal-app[bot] <windmill-internal-app[bot]@users.noreply.github.com>
* chore: narrow secret-file Read deny rule to dotfiles/extensions
* feat(vault): configurable JWT auth mount path and fix setup docs
* chore: bump ee-repo-ref for vault jwt mount path
* chore: bump ee-repo-ref after rebase onto EE main
* chore: update ee-repo-ref to a1cd60b54e8595b4e5ce6b654e675e4bbe2253b2
This commit updates the EE repository reference after PR #567 was merged in windmill-ee-private.
Previous ee-repo-ref: c274f233a0ebb54afa296c3db15ff330e1baebcf
New ee-repo-ref: a1cd60b54e8595b4e5ce6b654e675e4bbe2253b2
Automated by sync-ee-ref workflow.
---------
Co-authored-by: windmill-internal-app[bot] <windmill-internal-app[bot]@users.noreply.github.com>
* deps: pin tokio-postgres to MaterializeInc fork to fix query_typed_raw deadlock
`pg_executor`'s `Client::query_typed_raw` (and `Client::prepare` on the
streaming path) deadlock when the result schema contains a column whose
Oid the client doesn't know yet — citext, custom enums, custom domains,
postgis types. Easy to reproduce against any partitioned table with a
citext column: ~100+ rows is enough on localhost, less on slower links.
`psql` works fine for the same query because the simple-query protocol
doesn't trigger the typeinfo lookup path.
## Root cause (unchanged tokio-postgres bug for years)
`query::query_typed` calls `get_type(client, oid).await` synchronously
while still holding the original query's `Responses` stream. The
original query's `DataRow`s back up in the per-request
`mpsc::channel(1)`, `Connection::poll_read` stops draining the wire,
and the typeinfo sub-query response (queued on the same socket behind
those DataRows) never arrives. Classic head-of-line blocking.
## Fix
Pin `tokio-postgres` / `postgres-types` / `postgres-protocol` (via
[patch.crates-io]) and the workspace `rust-postgres` /
`rust-postgres-native-tls` aliases to the
[MaterializeInc rust-postgres fork at `78c1222577`](https://github.com/MaterializeInc/rust-postgres/tree/master).
MI's [PR #33 "bigger-channels"](https://github.com/MaterializeInc/rust-postgres/pull/33)
(merged 2025-12-11) resized the per-request response channel from
`mpsc::channel(1)` → `mpsc::channel(1024)`. That gives the connection
task 1024 batches of headroom while a streaming consumer is paused
mid-stream — orders of magnitude more than realistic typeinfo
deferral needs (≈3 batches).
## Why MaterializeInc and not a windmill-labs fork
`windmill-trigger-postgres` already depended on the imor fork for the
`postgres-replication` crate (logical replication: `CopyBothDuplex`,
`LogicalReplicationStream`, `TupleData` decoding including binary
tuples). That crate has never been on upstream rust-postgres — petrosagg's
[PR #752](https://github.com/rust-postgres/rust-postgres/pull/752) was
closed in 2021 in favour of a smaller split,
[PR #778](https://github.com/rust-postgres/rust-postgres/pull/778) is
still open today after five years. petrosagg keeps the replication work
alive on the MaterializeInc fork.
MaterializeInc is a strict superset of what we previously got from imor:
- imor's binary-tuple commit (sha `20265ef38e`) was merged into MI master.
- petrosagg has added perf + correctness fixes on top (allocation reuse,
proper decoding fixes).
- The deadlock mitigation (`channel(1024)`) was added three weeks before
this issue surfaced.
MI tracks upstream rust-postgres with a periodic catch-up merge (12-18 mo
cadence; last on 2025-12-03, ~100 commits picked up). Not an abandoned
fork.
## Why this works now (didn't on earlier attempt)
A previous attempt at this PR (`248ccb5a97`) hit CI failure because the
MI fork's `postgres-types 0.2.11` requires `serde_core ^1.0.221`, but
Windmill's workspace pinned `serde = "=1.0.220"` for swc_common 0.37.5's
`pub use serde::__private as serde;` hack. Bumping serde above 1.0.220
broke the swc_ecma_ast `Deserialize` derive under the
`enterprise,deno_core,…` feature set.
The earlier blocker is now resolved by #9111 which bumped the deno +
swc pin set to a "goldilocks" combination where `swc_common 14.0.4`
drops the `__private` hack, freeing the workspace serde pin to `^1`.
serde now resolves to 1.0.228, which satisfies MI's `serde_core ^1.0.221`
requirement transitively — no extra workspace pin needed.
## Diff shape
Two files only:
- `backend/Cargo.toml` (+34/-1): three new `[patch.crates-io]` entries
(`tokio-postgres`, `postgres-types`, `postgres-protocol` → MI fork)
plus comment block, plus the two workspace deps (`rust-postgres` /
`rust-postgres-native-tls`) repointed from imor's fork to MI's.
- `backend/Cargo.lock` — auto-regenerated. Replaces all `imor/rust-postgres`
references with `MaterializeInc/rust-postgres`, bumps the affected
crate versions to MI's set (tokio-postgres 0.7.11 → 0.7.15,
postgres-types 0.2.7 → 0.2.11, postgres-protocol 0.6.7 → 0.6.9,
postgres-native-tls 0.5.0 → 0.5.2). No source code changes.
## Verification
- `cargo check --features quickjs` → clean.
- `cargo check -p windmill-worker --features quickjs` → clean (pg_executor
builds).
- Repro tested earlier in the thread that produced this PR: the
partitioned-citext-table query on Neon goes from "hangs indefinitely"
(server idle on `wait_event=ClientRead` while client awaits typeinfo
behind undrained DataRows) to "completes in ~1.0s, 100 rows" with the
MI fork's `bounded(1024)` response channel.
## Caveats
- **`bounded(1024)` is a mitigation, not a closure.** Theoretical failure
mode remains at >~64 MB single-query results with a custom-Oid column
(typeinfo defers for >1024 batches of ~64 KB each). The strict-correct
fix is `mpsc::unbounded()` — proposed as a follow-up PR to MI. For
realistic Windmill workloads, 1024 batches of headroom is well past
the ~3-batch typeinfo deferral that's actually needed.
- **`postgres-replication` is now upstream-of-fork's only home.** No
realistic path to upstream rust-postgres merging it. The MI pin is
intended to stay in place until either upstream changes course
(unlikely) or MI publishes to crates.io (also unlikely — they don't
publish releases of the fork).
🤖 Generated with [Claude Code](https://claude.com/claude-code)
* chore: update ee-repo-ref to 8fe0d290fb0b71c24184eb5ad99bbdc7c813697c
This commit updates the EE repository reference after PR #568 was merged in windmill-ee-private.
Previous ee-repo-ref: 4d01d171228196f28ddabc1150242bfab623d5cf
New ee-repo-ref: 8fe0d290fb0b71c24184eb5ad99bbdc7c813697c
Automated by sync-ee-ref workflow.
---------
Co-authored-by: windmill-internal-app[bot] <windmill-internal-app[bot]@users.noreply.github.com>
VariableEditor persists the current workspace's edit state via
UserDraft.save on every mutation, keyed on editPath ('' for new
variables → in-memory only). Backend fetches now overlay a matching
local autosave when one exists, and initNew() rehydrates from the
in-memory empty-path entry so opening a fresh "Add variable" drawer
keeps any unsaved work from the previous open. After a successful
save we drop the corresponding entry.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
ResourceEditor registers a UserDraft.use<ResourceState> handle keyed
on the initialPath (empty for new resources, in-memory only). A
$effect deep-tracks the current workspace's edit state and forwards
mutations to the handle; on bootstrap and lazy backend-fetch the
local autosave wins over the backend value when they diverge. After
a successful save() we call UserDraft.remove so the local autosave
doesn't outlive the deploy. Cross-workspace deploys always start from
the live backend value rather than the local draft.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Diego Imbert
Guilhem
centdix
Ruben Fiszelandrubenfiszel
Samuel Wilk