Both automated reviews flagged the hoisted block's comment as narration
(AGENTS.md: record constraints, once, ≤4 lines). Drop the "(validate before
opening the tx)" restatement and compress the deadlock note to the single
load-bearing constraint; the fork comment is unchanged.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
create_schedule opened the RLS transaction (user_db.begin) first, then ran
reads that deliberately use the non-RLS `db` pool — fork-ness and
permissioned_as/email resolution — while holding it. Acquiring a second pooled
connection while the tx holds one self-deadlocks on a single-connection pool
(embedded Postgres, PgBouncer statement mode, any max_connections=1 setup): the
read blocks on the sqlx acquire timeout, then errors.
Move those reads (and the ScheduleType::from_str validation) above
user_db.begin(). They don't depend on the tx and bypass RLS by design, so the
result is semantically identical; the RLS transaction is simply opened later and
held for less time. Same class of fix as #9970 (migration bootstrap on the
migrator's held connection).
Note: sibling paths keep the same latent pattern on branches this change does
not touch (push_scheduled_job reads the pool under the tx for flow schedules;
edit_schedule/set_enabled for cross-user permissioned_as) — a possible follow-up.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
* fix: bound postgres result collection so it cannot OOM the worker
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
* fix: render the sql result limit exactly so the error can be set verbatim
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
* docs: point the fraction rationale at the renderer that still emits them
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
* perf: stop re-parsing every collected row to rebuild it as a RawValue
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
* style: drop a dangling doc line and an unrelated rustfmt reflow
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>
* fix: bound duckdb result collection so an oversized result cannot OOM the worker
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
* fix: keep the duckdb cap a worker-survival limit rather than a cloud product one
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
* fix: refuse an oversized blob before it expands to one json value per byte
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
* fix: share one expansion budget across a row's values, nested ones included
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
* fix: bound the row's own serialization so escaping cannot outgrow the budget
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
* fix: charge a json column before parsing it into a value tree
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
* docs: trim the json budget rationale and name what the budget does not cover
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
* fix: gate the sql result size limit on the duckdb feature
Its only consumer is the duckdb executor, so the minimal build compiled it
as dead code and failed under -D warnings.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>
* feat: bound how much disk a single duckdb job can spill
* fix: name the env var and correct duckdb's unreachable spill-cap advice
* fix: do not blame an unset env var for duckdb's default spill cap
* style: keep the duckdb spill-cap invariant comments within four lines
* docs: size the duckdb spill cap against the disk cloud pods actually use
* fix: tell MCP clients which tool parameters may be omitted
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
* refactor: make the mcp property-key rename testable and shorten the hint
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
* fix: keep the mcp omission hint from calling flow inputs optional
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
* fix: skip the mcp omission hint on a parameterless tool
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>
* docs: design for where the npm proxy keeps cached registry content
* feat(npm-proxy): keep package files on disk and in the object store
* fix(npm-proxy): degrade when the cache is unwritable, stream and bound it
* fix(npm-proxy): keep the happy path off the heap and isolate pull scratch
* fix(npm-proxy): bound the upload, verify pulled trees, keep oversized manifests
* fix(npm-proxy): protect live scratch, bound uploads by parts, refuse traversals
* fix: let the blocking unpack own the scratch it writes into
* fix: replace a cache directory that is not a package instead of deferring to it
* fix: evict by moving a package off the live path, not by deleting it in place
* fix: leave a package the sweep cannot move rather than deleting it in place
* fix: take one registry snapshot through a cache miss
* fix: stamp a pulled package as used so the sweep does not evict it first
* fix: explain duckdb failures caused by job isolation
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
* fix: apply the isolation policy to the schema-sync pre-pass
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
* chore: bump ee ref for the out-of-memory hint wording
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
* fix: bump the bundled DuckDB engine to 1.5.5
The 1.5.5 duckdb crate no longer hands back a 96-bit `rust_decimal`, so a
DECIMAL wider than that renders instead of panicking inside an `extern "C"`
frame — which, being unable to unwind, aborted the whole worker process and
left the job running as a zombie. `SELECT
'1234567890123456789012345678.9012345678'::DECIMAL(38, 10)` was enough.
Adapting to the crate's API: `Value` is now `#[non_exhaustive]` and gained
`UHugeInt` and `Geometry`, and `rust_decimal` became an optional feature that
the `decimal`/`numeric` argument path still needs.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
* fix: address review findings on the duckdb bump
Run the FFI crate's own tests in CI: it is excluded from the workspace, so the
`cargo test --all` in backend-test never reached them and the new guard against
the worker-aborting DECIMAL would not have run. build_dev.sh now honors a
caller-pinned CARGO_TARGET_DIR so the test build reuses that compile instead of
building the bundled engine a second time.
Also pin UHUGEINT rendering, and correct the rust_decimal rationale —
`Decimal::new` is public without the feature, so the reason is that the feature
reproduces the exact binding the crate used to derive, not that nothing else can.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
* fix: address review nits on the duckdb bump
Name the unsupported DuckDB type rather than dumping the value, which may be
arbitrarily large or hold data that does not belong in an error message, and
say which column it came from.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
* chore: pin the ee ref to the narrowed duckdb extension allowlist
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
* feat: keep duckdb spilling behind the local-filesystem fence
* chore: repin the duckdb fork after adding the reset-test exclusion
* docs: stop claiming the duckdb patch has been filed upstream
* docs: point the backend duckdb bullet at the fork's rationale
* fix: place lock_temp_directory so no existing struct member moves
* fix: skip the extension-load guard when the repo is unreachable
* refactor: trim the fork comments and fail the extension guard in CI
* chore: repin the duckdb fork onto upstream duckdb-rs main
* fix: keep the engine patch applying on a CRLF checkout
* docs: link the upstream issue tracking the underlying problem
* chore: repin the duckdb fork onto the patch as filed upstream
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
* chore: update ee-repo-ref to 88568d11162ffa11723e7955e613224bab4f0568
This commit updates the EE repository reference after PR #720 was merged in windmill-ee-private.
Previous ee-repo-ref: 22f075c1164d9dd5a3ba92d682905aabd071d273
New ee-repo-ref: 88568d11162ffa11723e7955e613224bab4f0568
Automated by sync-ee-ref workflow.
* chore: repin the duckdb fork onto the cmake/fmt build fix
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
* test: pin the immutability half of lock_temp_directory
The spill test proves the exemption works; nothing proved the lock that makes
it sound. A rebase could drop the refusals and leave every other tripwire green.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>
Co-authored-by: windmill-internal-app[bot] <windmill-internal-app[bot]@users.noreply.github.com>
* feat(ata): prefer the npm proxy when the instance configures a registry
* fix(npm-proxy): cap tarball extraction and stop pinning a failed config probe
* fix(npm-proxy): keep large packages cacheable by using a single shard
* fix(npm-proxy): cache the archive so a large package is served, not refused
* fix(npm-proxy): read archives off the runtime, keeping only what types need
* fix(npm-proxy): charge a retained entry for what it allocates, not its bytes
* fix(npm-proxy): size retention for real packages and read the manifest back
* fix(npm-proxy): charge path bytes and pin the manifest read-back
* fix(npm-proxy): stop retaining past the budget instead of refusing the package
* fix(parser-py): keep first param when def main( line has trailing comment
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
* chore: bump windmill-parser-wasm-py to 1.782.0
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
---------
Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
* fix: carry the validated token endpoint with MCP OAuth credentials
get_or_refresh_mcp_client already resolved and checked the token endpoint on
both its cached and freshly-registered paths, then dropped the result. Keeping
it on McpClientCredentials lets the callers that post the client_secret there
connect to the address that was checked, and removes a second lookup they were
each doing on their own.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
* refactor: hand out the token URL with the client pinned to it
Makes the pin unrepresentable-if-wrong rather than documented: the validated
target is private and reachable only through token_request, which returns the
URL together with the client pinned to the address it was checked against, so
a caller cannot pin one host and post to another.
Adds the test that was missing under the whole guard: that the pinned client
really does connect to the pinned address instead of resolving the host. The
accept loop is bounded, so a pin that stops working fails in seconds.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
* fix: keep the token endpoint private behind token_request
Leaving the URL public still allowed posting the client_secret to it on an
unpinned client, so the invariant was only documented. Both the URL and its
validated target are now private and reachable together, and the pinning test
resets the accepted socket to blocking so it does not read empty on macOS.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
* test: drop the non-blocking reset from the pinning test
Linux hands back a blocking socket from accept regardless of the listener's
flag, and no runner here builds this crate for a platform that does otherwise.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
* chore: update ee-repo-ref to f8d523195e40fd1d740595dcab6ce5cdc1bdbf09
This commit updates the EE repository reference after PR #718 was merged in windmill-ee-private.
Previous ee-repo-ref: 729df45314c6f2168b44eddb6edea401b0495d6d
New ee-repo-ref: f8d523195e40fd1d740595dcab6ce5cdc1bdbf09
Automated by sync-ee-ref workflow.
---------
Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>
Co-authored-by: windmill-internal-app[bot] <windmill-internal-app[bot]@users.noreply.github.com>
Co-authored-by: Ruben Fiszel <ruben@windmill.dev>
* feat(raw-apps): route in-browser npm installs through the npm proxy
* fix(npm-proxy): follow npm range semantics and cache packuments
* fix(npm-proxy): bound the packument cache by bytes and stream tarballs
* fix(npm-proxy): keep a v-prefixed pin exact and read the tarball once
* chore(raw-apps): bump the ui_builder pin to the npm-proxy installer
* feat(triggers): nested any_of / all_of filter groups
A trigger filter entry can now be a group — `{"any_of": [...]}` or
`{"all_of": [...]}` — nesting further entries, so criteria like
`A AND B AND (C OR D)` are expressible. Existing flat `{key, value}` lists keep
their meaning, combined by the trigger's `filter_logic` as before.
Filters are compiled once per connection: the set of top-level keys the whole
tree references is collected up front, so a message is parsed in a single
streaming pass that captures only those keys, instead of one full pass per leaf
filter as before. Filters that fail to parse are now logged rather than dropped
silently, since a nested group is easier to mistype than a flat entry.
The editor gains "Add group", rendering groups recursively with their own
AND/OR selector; Kafka and WebSocket triggers share it.
Fixes WIN-2345
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
* fix(triggers): drop empty filter groups instead of evaluating them
A group with no criterion cannot evaluate to a constant: true makes an `or`
filter list accept every message, false mutes an `and` list. Two clicks in the
editor ("Add group", save) produced one. Drop it when compiling so its siblings
stay in force, and reject at save time the filters the listener would otherwise
drop silently.
Also restore the item shape of `$ref`-typed arrays in the generated agent
schemas: the extractor only resolved refs at the property level, so moving
`filters.items` to a shared schema flattened it to a bare object. Resolving them
inside `items` too also recovers the shapes `initial_messages` and the MQTT
`topics` had already lost.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
* fix(triggers): name the offending entry when a nested filter is invalid
Serde's untagged error only reports that the outermost entry matched no
variant, whatever depth is actually wrong, which defeats the point of
validating a group at save time. Walk the tree instead and report the path.
Normalize the WebSocket editor's filters to [] on load, as the Kafka editor
does, so the list component can rely on an array.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
* fix(triggers): key filter rows by node so deletion keeps values aligned
The value editor seeds itself from `code` once, so an index-keyed row reused
for a different filter kept showing the deleted row's value.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
* chore: bump ee-repo-ref after merging main
The merge pulled OSS code that needs EE symbols newer than the companion
branch's base, so the companion was merged with EE main too.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
* perf(triggers): keep filter short-circuiting from materializing unread fields
The single-pass scan deserialized every referenced key before the boolean tree
ran, so an AND whose first leaf rejects the message still allocated the large
objects the later leaves name — the shape this feature exists for. Borrow the
wanted keys as raw slices during the scan and parse a field only when
evaluation actually reaches it.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
* feat(triggers): none_of filter group
Negation of a nested group, so a trigger can exclude what it must not react to
without inverting every other criterion. A key the message does not carry
satisfies it: there is nothing there to match.
Only groups can negate — the root's operator is the trigger's filter_logic
column, which has no value for it.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
* feat(triggers): address a nested field with a dotted path
`{path: "a.b.c", value: v}` alongside the existing `{key, value}`, so the common
case reads the way people write it instead of nesting the shape into the value.
A separate field rather than dots in `key`, which already means the top-level
field spelled that way — overloading it would resettle what existing triggers
over flattened payloads match.
Paths address objects only for now: a path through an array does not match
rather than guessing an element, and array containment stays on the value side.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
* docs(triggers): mention none_of in the filter_logic description
Plus a test for the empty-path-segment rejection, which had none.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
* chore: update ee-repo-ref to 78859aab0c6e78283ec8d2b37e8c410963afdc83
This commit updates the EE repository reference after PR #722 was merged in windmill-ee-private.
Previous ee-repo-ref: 0e42ba72ccc38a6b0a380f58afe0db36d284f4c9
New ee-repo-ref: 78859aab0c6e78283ec8d2b37e8c410963afdc83
Automated by sync-ee-ref workflow.
* fix(triggers): reject a criterion naming both key and path
The untagged enum takes such an entry as a `key` criterion and drops the
`path`, which is the silent-ignore the save-time validation exists to prevent.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
* refactor(triggers): drop the label next to the key/path toggle
The toggle already shows which one is selected.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
* fix(triggers): reject an entry that combines a criterion with a group
Generalizes the key+path fix: the untagged enum settles a half-and-half entry
on the first variant that fits and ignores the rest, so a criterion carrying a
group key lost the whole subtree without a word.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>
Co-authored-by: windmill-internal-app[bot] <windmill-internal-app[bot]@users.noreply.github.com>
* fix: bump the bundled DuckDB engine to 1.5.5
The 1.5.5 duckdb crate no longer hands back a 96-bit `rust_decimal`, so a
DECIMAL wider than that renders instead of panicking inside an `extern "C"`
frame — which, being unable to unwind, aborted the whole worker process and
left the job running as a zombie. `SELECT
'1234567890123456789012345678.9012345678'::DECIMAL(38, 10)` was enough.
Adapting to the crate's API: `Value` is now `#[non_exhaustive]` and gained
`UHugeInt` and `Geometry`, and `rust_decimal` became an optional feature that
the `decimal`/`numeric` argument path still needs.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
* fix: address review findings on the duckdb bump
Run the FFI crate's own tests in CI: it is excluded from the workspace, so the
`cargo test --all` in backend-test never reached them and the new guard against
the worker-aborting DECIMAL would not have run. build_dev.sh now honors a
caller-pinned CARGO_TARGET_DIR so the test build reuses that compile instead of
building the bundled engine a second time.
Also pin UHUGEINT rendering, and correct the rust_decimal rationale —
`Decimal::new` is public without the feature, so the reason is that the feature
reproduces the exact binding the crate used to derive, not that nothing else can.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
* fix: address review nits on the duckdb bump
Name the unsupported DuckDB type rather than dumping the value, which may be
arbitrarily large or hold data that does not belong in an error message, and
say which column it came from.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
* chore: pin the ee ref to the narrowed duckdb extension allowlist
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
* chore: pin the ee ref to the verified duckdb extension allowlist
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
* chore: pin the ee ref to the allowlist regression test
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
* chore: update ee-repo-ref to 04dd9c5c352f04995cd0470400a877261f956561
This commit updates the EE repository reference after PR #716 was merged in windmill-ee-private.
Previous ee-repo-ref: fe7eb440a5bbae37774d3a96b69ab5c46c0b8936
New ee-repo-ref: 04dd9c5c352f04995cd0470400a877261f956561
Automated by sync-ee-ref workflow.
---------
Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>
Co-authored-by: windmill-internal-app[bot] <windmill-internal-app[bot]@users.noreply.github.com>
* fix: scope a fork's cloned app policy and custom path to its creator
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
* fix: gate a cloned anonymous app on the parent's own deployment rule
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
* docs: state why a cloned anonymous app is gated more strictly than create_app
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
* style: wrap an over-long comment line in clone_apps
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
* fix: clone an app's execution_mode unchanged
Forcing `publisher` on a cloned app was a speed bump rather than a boundary:
protection rules are workspace-scoped and are not cloned, so the fork's creator
can publish an anonymous app there with no rule in the way. It was also the one
policy field a deploy back to the parent carries verbatim, since `update_app`
recomputes the identity but writes the policy wholesale, so a fork's copy could
silently close the parent's public endpoint.
The identity rewrite is what closes the hole this addresses: the fork's endpoint
no longer runs as whoever the parent published it as.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
* fix: ignore an app's run-as identity when comparing workspaces
`compare_two_apps` hashed the whole policy, so a fork whose apps were re-pointed
at their creator reported every one of them as changed. Nothing could clear those
entries: the deploy offers the target's current identity, the deployer's, or a
typed-in one, never the source's, so the difference survives however many times
the item is deployed. `script` and `flow` already compare no identity.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>
* chore: bump ee ref and refresh query cache
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
* chore: bump ee ref
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
* chore: bump ee ref
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
* chore: bump ee ref
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
* chore: bump ee ref
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
* chore: update ee-repo-ref to 236115e11f074d86675aa5acdf5061dd3e64f43c
This commit updates the EE repository reference after PR #719 was merged in windmill-ee-private.
Previous ee-repo-ref: 62bc50118d09374b8a45756504520cfb6e5f0210
New ee-repo-ref: 236115e11f074d86675aa5acdf5061dd3e64f43c
Automated by sync-ee-ref workflow.
---------
Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>
Co-authored-by: windmill-internal-app[bot] <windmill-internal-app[bot]@users.noreply.github.com>
Co-authored-by: Ruben Fiszel <ruben@windmill.dev>
* feat: version resource values with history, diff and restore
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
* fix: record resource versions in a trigger so direct writes are covered
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
* fix: show the selected version's value and tighten history write access
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
* perf: gate resource version recording in trigger WHEN clauses
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
* feat: clear a resource's past versions, and address review nits
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
* fix: restore the displayed version and keep author attribution on pooled writes
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
* fix: scope history to the selected workspace and gate clearing on ownership
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
* fix: gate restore on write access and clearing on the signed-in workspace
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
* refactor(frontend): share the version-history row between script and resource drawers
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
* perf: trim resource version history in the monitor sweep, not on write
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
* fix(frontend): match the script versions drawer shell for resource history
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
* perf(frontend): highlight version values instead of mounting monaco
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
* fix(frontend): match the script drawer's code preview presentation
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
* fix: rank version trim in one windowed pass instead of a correlated delete
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
* refactor(frontend): treat the newest version as current by position
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
* perf: gate the resource version trim to an hourly sweep
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
* fix: unnest the version row action and correct the trim cadence docs
* perf: cap the history listing and use sets for reference lookup
* feat: warn when a resource is written more than 60 times a minute
* fix: lower the resource write advisory to 20 per minute
* fix: discard stale history loads and never diff against an unread value
* fix: correct the write advisory boundary and document the eviction lock
* fix: read history and the live value from one snapshot
* refactor: read the drawer's diff baseline from versions, not the live resource
* fix: open the history drawer with no version selected
* fix: disarm the clear confirmation and clear the pane when the selection moves
* fix: explain the missing diff and drop a guard that can no longer fire
---------
Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>
* fix(smtp): explain why a test email failed instead of 'deadline has elapsed'
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
* fix(smtp): keep non-SMTP error codes and retire a stale test alert
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
* chore: update ee-repo-ref to f0df8b82c4c089d384423ed64b8504506084820d
This commit updates the EE repository reference after PR #721 was merged in windmill-ee-private.
Previous ee-repo-ref: 1ffaf3dea81e007c6c11146c1e12e97e83f5b938
New ee-repo-ref: f0df8b82c4c089d384423ed64b8504506084820d
Automated by sync-ee-ref workflow.
---------
Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>
Co-authored-by: windmill-internal-app[bot] <windmill-internal-app[bot]@users.noreply.github.com>
* fix(duckdb): cast columns in quicksearch so nested types can be previewed
DuckDB's `CONCAT` implicitly casts scalars but rejects nested types:
D SELECT CONCAT(' ', ['a','b']);
Binder Error: Cannot concatenate types VARCHAR and VARCHAR[] - an explicit
cast is required
Quicksearch concatenates every visible column, so one LIST, STRUCT or MAP column
makes a table impossible to preview — both the grid and its row count fail:
Binder Error: Cannot concatenate types VARCHAR, VARCHAR, BIGINT, ...,
VARCHAR[], ... and TIMESTAMP WITH TIME ZONE - an explicit cast is required
LINE 1: ... FROM "raw"."accounts" WHERE ($1 = '' OR CONCAT(' ', "id", ...
Every scalar in that list concatenates fine on its own — VARCHAR, BIGINT,
DOUBLE, BOOLEAN, DATE and TIMESTAMPTZ were each checked individually — so the
array column is the entire cause.
Cast each column in the predicate. The comparison is textual either way, so no
result changes, and the projection is untouched: casting there would change the
types the caller reads back. This follows the shape already used for MSSQL in
`mssql_needs_cast_for_eq`.
Both DuckDB quicksearch sites are covered, SELECT and COUNT. Fixing one leaves
the grid rendering while the row count still errors.
Tests include the live path: the Database Manager sends a
`-- WM_INTERNAL_DB_SELECT {...}` marker and the backend expands it, so the new
test drives that expansion with the real 27-column definition captured from a
failing job, `sync_id VARCHAR[]` included. It fails without the fix and passes
with it.
* fix(frontend): cast columns in the DuckDB quicksearch
Same defect as the Rust query builders, in the implementation that actually
runs. `make_select_query` / `make_count_query` in windmill-common have no callers
anywhere in the repo; the query the browser sends is built here.
DuckDB's CONCAT implicitly casts scalars but rejects nested types, and
quicksearch concatenates every visible column, so one LIST column makes a table
impossible to preview — both the page and its row count fail with
Binder Error: Cannot concatenate types VARCHAR, ..., VARCHAR[], ... and
TIMESTAMP WITH TIME ZONE - an explicit cast is required
The helper lives in select.ts and is imported by count.ts so the two cannot
drift, and both call sites are fixed: fixing only SELECT leaves the grid
rendering while the row count still errors.
* fix(duckdb): cast only list columns in quicksearch, leaving other SQL byte-identical
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
* test(frontend): pin the DuckDB quicksearch column list byte-for-byte
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: Ruben Fiszel <ruben@windmill.dev>
Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>
`list_users` and `list_pending_invites` had no ORDER BY, so Postgres returned
rows in heap order. An UPDATE rewrites the row at the end of the heap, which
sent the member whose role was just toggled to the bottom of the list the
settings page refetches right after.
Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>
* feat(flow-editor): measure the redesigned step panels
Instruments the flow editor's step, loop and branch panels on the existing
anonymous `feature_usage` channel, so the redesign can be judged on how the
panels are actually used rather than on nothing.
Eight event kinds under a new `flow_editor` feature: panel opens and their
dwell (bucketed, per placement), placement-preference overrides, which
settings get configured or cleared, settings that read as invalid, the
prop-picker connect lifecycle, AI input suggestions, and the step header
menu that "Save to workspace" now lives behind.
Settings changes are diffed off `describeStepSettings`, the same view the
graph badges render, so the telemetry vocabulary cannot drift from the one
on screen. Only `panel_open` and `setting` carry an entity id — one opaque
id per editor mount — since a per-entity row is only worth its cost where
the spread per editing session is the question.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
* fix(flow-editor): keep the panel telemetry honest
Review follow-ups on the instrumentation:
- The top dwell bucket was `120s+`, and `+` is outside the charset
`is_identifier_shaped` accepts, so `log_feature_usage` skipped those
events and still answered 204 — the longest visits vanished with no
error on either side. Renamed to `120s_plus` and pinned every emittable
key against the backend's charset in a test, since the producer is
TypeScript and the validator is Rust.
- Dropped the per-session entity id from `setting`: it would pay a row per
session per day across twenty-four keys, for a distribution its plain
counter already largely answers.
- An armed connect that went away with its component never reported, so
`open` did not balance against `insert` + `abandon`.
- Session preview tabs keep hidden editors mounted, which billed panel
time nobody spent. `FlowEditorView` now publishes the visibility it
already knows about.
- Re-picking the active placement row logged a move, which also made
`auto:from_docked` mean two different things.
- The last dwell of a session was lost on tab close, since Svelte tears
components down on navigation but not on `pagehide`.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
* refactor(flow-editor): narrow the telemetry to panel placement
The eight-kind instrumentation measured more than could be read. With nothing
recorded before the redesign there is no baseline to compare panel opens, dwell
times, settings usage or connect funnels against, so those counters answered
questions nobody could act on while costing a row per key per day in an
instance-wide table.
What remains are the three numbers the modal panel is actually judged on: how
often the 1280px breakpoint puts the panel in a modal, and how often people
override that in each direction.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
* fix(flow-editor): stop counting placement in session preview tabs
Preview tabs keep every flow editor mounted and laid out at panel width
whether or not it is the visible one, and that panel is narrower than the
breakpoint by construction. Each flow tab opened in a session therefore
emitted a `breakpoint_modal` on mount, and one drag of the session panel
across 1280px emitted one per mounted tab — with no host dimension in the
key to separate that from the crossings the counter exists to measure.
Also corrects the comment on the no-op placement guard, which justified
itself with a key vocabulary that no longer exists.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
* fix(flow-editor): make the three placement counters comparable
Sessions were excluded from the breakpoint counter but not from the two
override counters, so a pin made in a session landed in the same bucket
used to judge the breakpoint, with no crossing in the denominator to read
it against. All three are now gated together.
An override is also only counted when it moves the panel. Choosing
"Detached" on an editor the width had already put in a modal states a
preference without changing anything, and the aggregate carries no width
to separate that from the wide-screen override that is the actual signal.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
* docs(flow-editor): describe the two override keys by what emits them
They documented themselves as overriding `auto`, which is no longer the
rule: pinning Attached on a wide editor overrides `auto` and emits
nothing, while going from an Attached pin to Detached below the
breakpoint emits `force_detach` even though `auto` would have produced a
modal there too. This file is what someone reads when interpreting the
numbers, and "override of auto" is the misreading the emission rule
exists to prevent.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
* fix(flow-editor): count the panel moving, not the breakpoint being armed
The tracker held "the breakpoint is responsible for this modal" rather
than "the panel is modal", so on a narrow editor pinning Detached and
releasing it back to Auto emitted a second breakpoint_modal for a panel
that never moved. It also died with the editor, which FlowBuilder rebuilds
through a `{#key}` on every reload — each rebuild re-armed it and counted
the same narrow editor again.
Both inflate the denominator that the two override counters are read
against, and both bias it the same way: toward concluding that nobody
overrides the breakpoint.
The tracker now follows the panel's placement across preference changes,
and FlowBuilder owns it from above the `{#key}`, which also puts the
session exclusion in one place instead of at each call site. The
moves-only rule moves into `forcedPlacementEvent` so both halves of it sit
in the module the tests can reach.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
* fix(flow-editor): ignore placements measured before the editor is laid out
A reload rebuilds the editor through `{#key renderCount}`, and the panel
controller is rebuilt with it: its width restarts at zero, which resolves to
`docked` because that is what is safe to render rather than because the editor
is wide. The breakpoint tracker read that transient as the panel having docked
and counted the real width landing as a fresh crossing, inflating the
denominator both override ratios are read against.
`useFlowPanelMode` now exposes `measured`, and the tracker skips anything
unmeasured instead of recording it as a placement.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
* docs(flow-editor): state the placement invariants once each
The width-zero rule had accumulated at four sites, two of which forward it
without being able to break it. Keep it beside the guards that enforce it.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>
* fix: scope cloned app policy and custom path to the fork's creator
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
* refactor: share the app custom-path scoping rule across its call sites
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
* docs: tighten the cloned-app-policy comments
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
* docs: correct the execution_mode and custom-path scoping rationale
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>
* feat: allow custom dev workspace environment labels
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
* fix: reject dev labels that shadow a tracked branch's namespace
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
* fix: guard dev labels against a repo's assumed default branch
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
* chore: state the badge-cap rationale once and drop unenforceable openapi constraints
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
* refactor: offer a fixed list of environment labels instead of free text
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
* fix: match the accepted label set to the openapi enum exactly
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
* docs: stop describing the label set as dev/staging only
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>
* feat: add public sharing option for job pages
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
* fix: gate public run sharing and address review findings
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
* fix: address review nits on public run sharing
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
* fix: key public run view on workspace, job and token
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>
* [ee] feat: improve duckdb isolation
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
* chore: update ee-repo-ref to c4e6cbc1a1efeca5b71920c7903db2347d0eeda0
This commit updates the EE repository reference after PR #713 was merged in windmill-ee-private.
Previous ee-repo-ref: f630f7e73cb863e312430738d81d802a3971f7cd
New ee-repo-ref: c4e6cbc1a1efeca5b71920c7903db2347d0eeda0
Automated by sync-ee-ref workflow.
---------
Co-authored-by: Claude Opus 4.8 <noreply@anthropic.com>
Co-authored-by: windmill-internal-app[bot] <windmill-internal-app[bot]@users.noreply.github.com>
* fix: keep the same_worker pin when a suspend ends without approval
A disapproved or timed-out approval gate hands the flow back through the
UpdateFlow channel with unrecoverable = true. That flag means "the previous
step's worker died", and it is read by six sites. Five of them happen to want
what it does here, but continue_on_same_worker and continue_with_runners do
not: the worker that ran the approval step is alive, so unpinning the error
handler and routing it by tag breaks the ./shared contract of a same_worker
flow and can land it on a worker group that cannot run it — the same defect
#10551 fixed for the three producers that hand back a live flow.
Replace the boolean with StepFailureKind so the suspend producer can say
"worker alive, but this failure is not the module's to handle" instead of
overstating a worker death. The failed module's error policy is deliberately
still bypassed: the failure is recorded against the step the gate was holding
back, which never ran, so its retry would re-open the gate and its
continue_on_error would skip it outright (verified: the gated step is marked
Failure with a nil job id and the flow jumps past it). suspend.
continue_on_disapprove_timeout remains the way to continue past a gate.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
* feat(flow-editor): flag that continue on error does not cover the approval gate
A resolved approval is recorded against the step the gate holds back, not
the step carrying the suspend, so continue_on_error never sees it: the flow
still stops on a disapproval or timeout. Point users at
suspend.continue_on_disapprove_timeout, which is what actually continues past
a gate, whenever both settings are on and that one is not.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>
* feat(debugger): install debug session deps from the instance registry settings
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
* fix(debugger): keep install-time registry credentials out of the session-visible tree
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
* docs: drop em dashes from the debugger registry docs and comments
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
* fix(debugger): stop installing for a session that went away during the settings fetch
Also serves nativets sessions the npm settings their installer reads.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>
When a module completes without spawning a job — an empty branch, an
empty for-loop, or a module already marked Success — the flow hands
itself back through the UpdateFlow channel, and the result processor
resumed it with unrecoverable = true regardless of what sent it. That
flag means "the previous step's worker died", which holds for none of
the three producers except a suspend that ended without approval.
The stale argument was inert until continue_on_same_worker and
continue_with_runners started reading it, since when the step after such
a module is pushed as an ordinary queued job. It is then routed by tag
and can land on any worker in the pool, breaking both the ./shared
directory contract and the guarantee that a same_worker flow stays on a
worker able to run it — a step whose tag resolves to a worker group that
cannot execute its language fails instantly, taking the flow with it.
Carry the flag on the UpdateFlow message so each producer states its own
case, rather than having the shared receiver assume the worst. The three
that hand back a live flow forward whatever their caller reported, so a
genuinely unrecoverable failure still crosses the hop unchanged.
Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>
* feat: allow a dev workspace to have its own dev workspace
Fixes WIN-2324
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
* fix: keep every dev workspace in a chain on a distinct deploy branch
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
* fix: count a dev workspace the caller has no seat in as holding its label
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
* refactor: keep the attach form standing when a candidate takes the last label
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
* fix: keep the label toggle visible when a candidate's dev workspace clashes
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
* docs: describe the cycle guard by what holds, not by what changed
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
* fix: refuse to archive a fork-backed dev workspace that owns a nested dev
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
* fix: put the deploy target and item filters under the pairing they configure
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
* fix: refuse to archive any dev workspace that owns a nested dev
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
* docs: fix the fixture family count
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
* fix: put the deploy target with the pairing line it restates, above protections
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
* fix: name the same family head in the workspace menu and the scope picker
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
* fix: stop offering to delete a dev workspace from the sidebar settings menu
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
* docs: state the visibility boundary the lineage root actually resolves to
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
* fix: serialize dev-pairing creation against teardown of the same workspace
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
* fix: lock both sides of an attach so adjacent pairings cannot share a label
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
* fix: serialize dev pairings on one key, the invariant being chain-wide
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
* fix: scope the pairing lock to the chains an operation reads
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
* fix: hold the pairing lock across renames and re-check the cycle under it
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
* fix: hide the fork-delete action until the workspace entry has loaded
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
* fix: lock archive before it reads the pairing state it acts on
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
* docs: describe the archive lock test by what it pins, and drop an unused fixture row
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>
* fix: forward proxy and TLS settings to debugger subprocesses
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
* fix: reach uv and the bun debugger with the forwarded network settings
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
* fix: map every CA variable spelling onto the one uv reads
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
* fix: keep package-index credentials out of debugged user code
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
* fix: install debugger dependencies outside the interpreter running user code
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
* fix: sandbox and bound the debugger dependency installer
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
* docs: correct the installer timeout rationale
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
* docs: scope the uv --cert note to the commands prepare-deps runs
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
* fix: build the debug venv against the interpreter that runs the script
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
* fix: do not start the debuggee for a session that already went away
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
* fix: remove the debug script when the session is gone before it starts
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>
* feat(mcp): serve the 2026-07-28 spec alongside the legacy protocol
* fix(mcp): keep oauth discovery strict and preserve request limits
* fix(mcp): allow the protocol's own headers through CORS
* fix(mcp): expose the auth challenge to browser clients
* chore: update ee-repo-ref to c1665a881b61616f96ffe7702b44840905304660
This commit updates the EE repository reference after PR #711 was merged in windmill-ee-private.
Previous ee-repo-ref: bc1c001e3e386342415dfb8ac31c6b97f6629320
New ee-repo-ref: c1665a881b61616f96ffe7702b44840905304660
Automated by sync-ee-ref workflow.
---------
Co-authored-by: windmill-internal-app[bot] <windmill-internal-app[bot]@users.noreply.github.com>
* fix: honor python index settings in prepare-deps and report install failures
* fix: forward python registry env to the debugger's prepare-deps
* fix: scope registry credentials to the prepare-deps subprocess
* fix: install python debug dependencies from the service, not the session
* fix: bound the debugger dependency install and keep the proxy bypass default
* docs: name the nsjail config that isolates debug sessions
hugocasa
Ruben Fiszel
Vladislav Kuzmin
Guilhem
AlexRV12