* fix: reuse existing transaction in push instead of acquiring new connection
In push_inner, fetch_authed_from_permissioned_as was acquiring a new
connection from the pool to fetch job permissions, even though a
transaction was already open. Use fetch_authed_from_permissioned_as_conn
with the existing transaction instead, reducing pool pressure when many
jobs are pushed concurrently.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
* improve contention
* improve contention
---------
Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>
* refactor: extract windmill-api-scripts and windmill-api-users subcrates
Split the monolithic windmill-api crate by extracting scripts.rs, flows.rs,
users.rs, and users_oss.rs into dedicated subcrates. This reduces incremental
rebuild times when editing these modules.
Changes:
- Create windmill-api-scripts crate (scripts.rs + flows.rs, ~4.3K lines)
- Create windmill-api-users crate (users.rs + users_oss.rs, ~2.4K lines)
- Move clear_schedule to windmill-queue (shared by scripts, flows, workspaces)
- Move username utilities (VALID_USERNAME, INVALID_USERNAME_CHARS,
generate_instance_wide_unique_username) to windmill-common/src/usernames.rs
- Move COOKIE_DOMAIN, IS_SECURE, WithStarredInfoQuery, BulkDeleteRequest,
WebhookShared to windmill-common for cross-crate access
- Original files in windmill-api become thin stubs with pub use re-exports
- EE-dependent route handlers remain in windmill-api (create_user, rename_user,
set_password, reset_password, etc.)
- Feature forwarding for enterprise, private, parquet, no_auth
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
* refactor: extract windmill-api-workspaces subcrate (Step 3)
Move workspaces.rs, workspaces_extra.rs, workspaces_oss.rs, and
workspaces_ee.rs into a new windmill-api-workspaces crate (~7K lines).
Routes that depend on windmill-api internals (AI copilot, teams,
tarball export, critical alerts, stripe) remain in the windmill-api
stub. The subcrate handles all other workspace management routes.
Also moved send_email_if_possible to windmill-common/email_oss.rs
to make it available across subcrates without circular deps.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
* all
* refactor: extract windmill-api-groups subcrate (groups.rs + folders.rs)
Extract groups.rs (1,093 lines) and folders.rs (833 lines) into a new
windmill-api-groups subcrate. Both modules had clean dependencies on
already-extracted crates (windmill-api-auth, windmill-common,
windmill-api-workspaces). Also removes unused re-exports of
get_instance_username_or_create_pending and INVALID_USERNAME_CHARS
from windmill-api/src/utils.rs.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
* refactor: add granular_acls.rs and folder_history.rs to windmill-api-groups
Extract granular_acls.rs (395 lines) and folder_history.rs (68 lines) into
the windmill-api-groups subcrate. Both modules only depend on already-extracted
crates and belong to the same access-control domain as groups and folders.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
* fix: remove unused imports and dead code from subcrate extraction
- Remove unused BASE_URL import from lib.rs
- Remove workspaces_extra.rs and workspaces_oss.rs re-export stubs (no consumers in windmill-api)
- Remove dead send_email_if_possible OSS stub (callers moved to windmill-api-users)
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
* all
* chore: bust CI cargo cache for subcrate split
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
* fix: re-export BASE_URL for EE files that use crate::BASE_URL
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
* fix: forward no_auth feature to windmill-api-users, remove dead code
- Add "windmill-api-users/no_auth" to windmill-api's no_auth feature
so the login bypass in users.rs:1600 activates correctly
- Remove dead send_email_if_possible from windmill-api-users/users_oss.rs
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
* chore: re-enable cargo cache for backend tests
Cache was disabled to bust stale entries from before subcrate split.
Now that a clean build has run, re-enable for faster CI.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
* chore: install mold+clang in CI workflows
The .cargo/config.toml uses mold linker for x86_64-linux.
Build scripts require linking even during cargo check.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
* chore: increase cargo test timeout to 30 min
Exit code 143 (SIGTERM) means the 20-min timeout was hit during
compilation without cache. Bump to 30 min as safety net.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
* fix: limit cargo build jobs to 4 to prevent OOM in CI
Exit code 143 (SIGTERM) after 8 min = OOM kill during compilation.
8 parallel LLVM codegen jobs exhaust memory on ubicloud-standard-8.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>
* fix: prevent schedule pool connection exhaustion
Schedules were disabling themselves after upgrading from v1.605.0 to
v1.614.0 due to pool connection deadlock. The root cause was
fetch_authed_from_permissioned_as acquiring a pool connection inside
push() while a transaction already held one, exhausting the pool under
pressure.
Fix: pre-compute Authed before db.begin() for the normal path, and
reuse the transaction connection via fetch_authed_from_permissioned_as_conn
for the on_behalf_of_email path. Peak pool usage drops from 2 to 1 for
all schedule push paths.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
* fix: replace pool exhaustion tests with comprehensive schedule push tests
Replace the 16 pool-specific tests with 19 tests covering all schedule
push code paths: script/flow scheduling, on_behalf_of_email (script and
flow), retry wrapping, duplicate detection, invalid timezone/cron/args,
script/flow not found, paused schedules, clock shift detection, disabled
schedule, path mismatch, push failure disabling schedule, and trigger
metadata.
Also simplify the obo_authed pattern in push_scheduled_job to use a
single match assignment instead of two bindings with .or() chaining.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>
* feat(backend): retrieve root_job for MiniPulledJob + fix root job for flow jobs
* nit
* nits
* set root_job only if additional information meaning if not equal to innermost or parent
* nit
* feat: add root job env var
* nits
* fix build
* nit
* fix build
* nit
* sqlx
* feat: ai agent steps base
* better backend and graph
* feat: anthropic, log viewer
* nit
* fix(frontend): hide tool nodes from timeline
* move ai agent actions from flow status to flow status module
* nits and workspace/hub scripts support
* tmp ref
* fix merge
* feat: display agent tools status in the graph
* fix reactivity
* fix flow status
* nit
* app compiles with every ee substituted
* Replace all oss files content
* Revert "Replace all oss files content"
This reverts commit ea4017d59f.
* delete all ee
* hide all _ee files under private flag
* hide every oss stuff when private flag set
* pub use *
* gitignore and substitute script
* pub mod for ee needed for ee repo
* small mistakes
* remove oidc_oss impl
* ee ref (temp)
* ee ref
* fix --all-features selecting private in OSS CI
* ee repo ref
* allow unused
* init
* croner with fall back on scheduling failure
* adding packages
* fallback mechanism v1
* removing debug console log
* improving cargo organization
* queue cargo.toml
* use workspace cargo-tz
* test next occurrence when creating new croner schedule
* toggle label
* treat legacy edit requests where no cron_version is provided as v1
* fix edit schedule
* Add concurrency_key table and field to payload
Add a new table custom_concurrency_key for storing concurrency keys on a
per job basis. Add the field to the job payload to populate the table on
creation of a new job.
* Rename to custom_concurrency_key
Be more explicit about the nature of the option variable by calling it
a more appropriate name.
* Delete concurrency_keys on delete_expire_items
* Propagate db errors up
* Add ref to EE changes
* update sqlx
* Add concurrency_key to tests
* Revert renaming of FlowValue field
FlowValue field concurrency_key must be named concurrency_key for
serialization purposes.
* Remove debug print statement
* Get concurrency_key from Job
Whether it is a Completed job or a queued one, add a ways to get the
associated concurrency key
* Add endpoint to get concurrency_key
* Add endpoint and button to get job concurrency_key
Add a line with a link on the FlowMetadata that goes to the
concurrency_groups page. add endpoint to get the concurrency_key of a
job
* Merge concurrency_key tables into one
migration + change all related querrys
* Add concurrency plot to the runs page
* Prepare sqlx
* Add concurrency Key filter on UI
* Prepare sqlx
* Prepare sqlx
* Fix filter and order query
The limit makes more sense if we cut the older rows, so order by
started_at
* Factor interpolation to reuse on concurrency key
Factor the arg interpolation logic into a function and finish the
processing of concurrency key before insertion
* Remove old concurrency key processing logic
* Second transaction with userdb
To send all concurrency intervals but revealing only uuids of accessible
jobs, make a second transaction with the userdb.
* Remove old second endpoint
the intervals endpoint now also gets the concurrencyt key information
for all jobs
* Show external jobs
* Put filters into a dropdown
Create a ToggleButtonMore and put elements into a dropdown
* Add toggle between two graphs
* Add filter functionality to concurrency graph
* Improve concurrency graph front
* Fix concurrency groups page
* Prepare sqlx
* Add ref to ee
* Change migration to create new table instead
Instead of renaming the custom_concurrency_key_ended atble, we create a
new table and we will delete custom_concurrency_key_ended in the future
when it is no longer linked to any jobs
* Do small UI improvements
* Fix range fusion by not filtering past jobs
Instead of filtering jobs in the backend on a startedAfter value, limit
the query to 1000 and query all possible towards the past to get a good
context for the graph in most sane situations
* Improve frontend UI
