Repository navigation
Releases
Windshift v0.9.5
Windshift 0.9.5
Windshift 0.9.5 reworks time tracking around one place to log work. You can open the time-log form from
anywhere in the app, start and stop a timer from the Time & Projects sidebar entry, and run a timer without
a project. When a stop already has the information a worklog needs, the entry is booked in one step;
otherwise it is saved as a recoverable draft that stays out of reports and item totals until you complete or
dismiss it.
Action flows gained a JavaScript node that can call the v2 REST API as the action's executor. The script
never sees the credential it acts with: Windshift issues a short-lived token for the executor, accepts only
relative v2 paths, and applies the same permission checks as a normal API client, so a script can never act
with more permission than its executor has. Action definitions are now versioned, and each run records the
revision it used. Actions can also be renamed and duplicated, and the canvas previews are driven by live
metadata.
Deployment owners can apply an operator-owned policy file that denies capabilities and sets instance-wide
ceilings without forking the application or changing organization preferences. The policy is read at
startup, requires a restart to change, and cannot be edited through the API.
Imports grew on two fronts. A new plane-import command loads a Plane export while preserving item
numbers, authors, and timestamps, and the Jira importer gained Data Center workflow, screen, and Assets
support through a read-only companion plugin. Sign-in added GitHub alongside OIDC, and customer portals can
now restrict access to named organizations and to customer-only identity providers.
The frontend received a performance and consistency pass: collection views stay fast with long lists, the
rich-text editor is no longer loaded at startup, locale catalogs load in parallel, and date pickers were
unified with a configurable week start. A hand-written Dutch locale joins the built-in languages, and the
design system received an accessibility pass.
Highlights
Time tracking from one surface
Time logging no longer depends on being on a particular page. The command palette, item detail, the /time
page, mobile, and keyboard shortcuts all open the same form in place, so logging work never interrupts
navigation. The Time & Projects sidebar entry is now a dropdown with a start/stop timer, project search, a
drafts badge, and navigation (GitHub #496).
A timer can start without a project. On stop, the worklog is booked directly when project, description,
date, times, and duration are all present; when they are not, the incomplete entry is saved as a user-owned
draft. Drafts are recoverable from the sidebar badge, stay out of reports, item logged time, and briefings,
and live until you complete or dismiss them. A retried stop cannot double-book, and the v1 timer contract is
unchanged.
JavaScript action nodes with executor-scoped API access
An execute_script node runs JavaScript and can call the v2 REST API as the action's executor. The script
never sees the credential it acts with: Windshift issues a short-lived token for the executor, accepts only
relative /rest/api/v2 paths, and applies the same permission checks as a normal API client, so a script
can never act with more permission than its executor has. Execution stays bounded by timeouts, budgets, and
cancellation, and every change carries action and execution attribution (WI-2174, WI-2178).
Every save of an action now creates a revision, and each execution records the revision it ran, so you can
see exactly what ran and when. Revision history is browsable in the editor. Changing the executable
configuration of a privileged action requires action.set_actor, so editing a script is treated as
delegated code execution (WI-2173). Actions can also be renamed and duplicated from the manager (WI-2168,
WI-2170), and the flow editor previews are driven by live metadata (WI-2254).
Operator controls for deployment owners
A new operator-owned JSON policy file, loaded at startup with --operator-policy or
WINDSHIFT_OPERATOR_POLICY, lets a deployment owner deny capabilities and set instance-wide ceilings. The
deny list covers instance lifecycle, maintenance, runtime diagnostics, plugin execution, LDAP, server
scripts, imports, generic mail, and AI; limits covers attachment and blob bytes, concurrent requests and
streams, requests per minute, automation steps per day, mail recipients per day, and AI tokens per day.
Managed mail mode, AI model allowlists, and an optional import admission directory are configured in the
same file. Unknown fields, invalid capabilities, and negative ceilings reject startup, and the file cannot
be edited through the API (WI-2175).
Denied capabilities return 403 operator_restricted; storage ceilings return 413 operator_limit_exceeded;
work admission returns 429 operator_limit_exceeded with Retry-After. Usage is counted across restarts,
so a restart does not grant a fresh allowance, and an unavailable required budget store fails closed.
Effective controls are readable by system administrators, while public capability discovery exposes only
denials and mail mode.
Plane and Jira Data Center imports
plane-import loads a JSONL contract produced from a Plane database into a Windshift workspace. It
preserves the source's item numbers, authors, timestamps, states, priorities, and labels, restores comments
and parent links, and keeps a mapping of imported records so a re-run updates its own rows instead of
duplicating them. Rows edited in Windshift since import are left alone, and number collisions the importer
does not own fail loudly. The command runs the migration catalog first unless --skip-migrate is passed,
and prints a JSON report (#379).
For Jira Data Center, a read-only companion plugin exposes workflow graphs and guards, screen schemes, and
Assets over an admin-only REST surface that the public DC API does not provide. The importer prefers the
companion and falls back to the public API when it is absent. Data Center Assets are read through the
legacy Insight API, and Assets import is now opt-in: choose schemas in the import wizard, and selecting none
imports nothing and records jira_assets_not_selected (WI-2059, WI-2069, WI-2070, WI-2078). Saved-filter
discovery is scoped to the target workspace, and Cloud workflow conditions now decode whether the source
sends an array or an object (WI-2079).
GitHub login and customer access policies
GitHub is available as a login provider alongside OIDC, for both the main application and customer portals.
A provider can require membership in a named GitHub organization, and sign-in requires a verified email and
never requests repository scopes (#501).
Customer portals gained customer-only identity providers and resource-scoped access. A customer-only
provider admits portal customers rather than internal users, and a resource such as a board carries an
access policy with mode public or an organization allowlist, so admission is decided independently of
content visibility. Existing self-managed deployments are unaffected until a provider or policy is
configured.
Boards, portals, and sharing
Boards can be published to a portal channel as a placement with its own label and an audience that is
independently narrower than the standalone public link. External publication and standalone availability
are now separate settings, with a sharing history, a revoke reason, and an allow-republish path. Owners and
system administrators get a Public Boards management page listing each publication with its audience,
locations, and sharing history (WI-2240).
The customer portal gained a Compact experience that composes modules, including placed boards, requests,
forms, knowledge base, drafts, approvals, and profile, with a configurable navigation order.
Frontend performance and consistency
Collection views stay fast with long lists and no longer need a virtualization preference (WI-2147). List
and backlog gained the board's multi-select mode and the same bulk action toolbar, so all three views offer
identical selection and bulk actions.
The rich-text editor is no longer loaded when the app starts. It loads only when needed and shows a clear
error with a Reload action if it fails. The startup JavaScript dropped by roughly 516 KB raw and 159 KB gzip
(WI-2144). Locale catalogs now load in parallel, so a non-English interface starts fetching while English is
still pending (WI-2150). SLA badges and panels now refresh only what changed and keep the right scope when
several workspaces update at once (WI-2112).
Date pickers were unified behind one calendar-aware implementation, a date-time picker was added, and
regional settings gained a week-start preference that the picker, roadmap, and weekly calendar follow. The
design system received an accessibility pass covering focus, contrast, typography, and motion (WI-2221), and
modal focus and keyboard ownership were unified so only the topmost dialog responds (WI-2222, WI-2224).
Also in 0.9.5
- Dutch language pack. A Dutch (Nederlands) translation joins the built-in languages with
full key coverage, including Dutch date formats for backend-rendered email, so the interface no longer
falls back to English (#532). - Configurable backlog row fields. Backlog rows can show a chosen set of fields, selected from a picker,
instead of a fixed layout (#490). - Agent run indicator. The comments tab shows a live indicator above the thread: a green dot with "Agent
running" while a runner works, or a grey dot with "Agent queued" while it waits. Clicking opens the Agent
log tab. - Raw Markdown editor plugin. A reference
editor.markdownprovider edits a surface as exact Markdown
source in a plain textarea, for cases where the WYSIWYG editor's source normalization is not wanted. - Roadmap fixes. Dependency arrows render reliably, a...
Windshift v0.9.4
Windshift 0.9.4
Windshift 0.9.4 adds transition screens, so a workflow can require input before a status change instead of only guarding
it. A transition can now carry a screen of required and optional fields; a move is refused until the required fields are
satisfied, the values are written with the status change in one transaction, and the workflow designer gained a
transition inspector to bind, create, and edit those screens.
On the Jira side, imports now finish instead of stopping at an arbitrary two-hour mark, download attachments and
complete per-issue subresources concurrently, and no longer fetch per-issue watchers by default. Alongside those, portal
and organization custom field values are keyed by field id, the filter builder addresses custom fields by their unique
cfid_<id>, and several customer, board, filter, and notification bugs are fixed.
This release also makes a large pass at right-to-left languages and localization. Arabic was reviewed and corrected,
Persian is new, user text now renders with its own direction, and email follows the recipient's language. Arabic and
Persian are here to be tested and verified, not declared complete; see the right-to-left section for the current state.
Important upgrade notice: the v1 API (/rest/api/v1) is now marked deprecated and frozen. New integrations, clients,
and tooling must use v2 (/rest/api/v2), and a lint guard blocks new first-party v1 usage. If you run Jira imports,
review the new JIRA_IMPORT_* variables in the upgrade notes. It is also advised to update the ws CLI alongside this
release; the CLI can now be installed with Homebrew (see the Homebrew install section below).
Highlights
Transition screens
A transition can carry a screen, a set of fields collected before the move. The move is refused with screen_required
and a structured missing_fields payload until every required field is satisfied; optional fields and already-satisfied
required fields never block. The submitted values are overlaid onto the condition context so validators see them, and
the fields and the status change are applied in one transaction, so a rejected transition writes nothing (WI-2011,
WI-2012).
Every user-initiated move opens a shared transition modal when the target transition has a screen: item detail, the
board context menu and drag, mobile, personal tasks, and the weekly calendar. The modal renders the shared field form
and surfaces missing required fields inline. Bulk moves cannot collect screen input, so a batch that contains an item
whose target transition needs a screen is refused whole with 409 screen_required and the blocked items listed, instead
of partially applying and reporting the rest per item; the board and support queue offer a retry without them (WI-2013).
The workflow designer gained a right-hand transition inspector. Selecting an edge binds, unbinds, or creates and binds a
screen, saving immediately through PUT /transitions/{id}/screen, and a screen-bearing edge carries a badge. A screen
can be laid out in the inspector itself through the shared field editor, and a screen in use cannot be deleted. The
binding lives on workflow_transitions.screen_id, is returned by the transition governance endpoint, and is
deliberately kept out of the bulk graph-replace payload so a graph save never clobbers it (WI-2014).
Jira imports run to completion
The importer no longer runs under a hard-coded two-hour deadline. A larger project used to stop mid-import, fail its
remaining bulk fetches with a deadline error, count unprocessed issues as failures, and report a generic
completed_with_errors with no explanation. Per-request HTTP timeouts and the worker lease already bound the work, so
the cap is gone and a large project is allowed to finish. An operator can still set JIRA_IMPORT_MAX_DURATION; when a
configured limit is hit, the import stops before the next batch without counting unprocessed issues as failures and
reports an explicit timed_out status naming the limit and the imported/total counts (WI-1981).
The importer also stopped issuing every Jira request sequentially, which bound throughput to round-trip latency (about
1.8 requests/second observed) rather than the client's 10 requests/second limiter. A bounded worker pool (default 8,
JIRA_IMPORT_CONCURRENCY to override) now drives the two phases that dominated large imports: attachment downloads and
the per-issue comment/worklog/Jira Service Management completion calls. Reuse checks, record writes, and the shared
progress counters stay on the main goroutine, so the persisted state is unchanged (WI-1982).
Two more changes cut import time and noise. Per-issue watcher import is now off by default, because it made one
sequential Jira request per issue for a low-value mapping; set JIRA_IMPORT_WATCHERS to opt back in, and the watcher
count the bulk fetch already returned is still preserved (WI-1983). Attachment downloads now go through the same retry
helper as JSON requests, so a 429 or 5xx is retried with an honored Retry-After and jittered backoff instead of
dropping the file silently, and a per-client throttle pauses every worker for the Retry-After window rather than
stampeding the limit (WI-2010).
Custom fields are addressed by id
Portal customer and customer-organization custom field values were keyed by the human-readable field name, so duplicate
names collided and a rename orphaned the stored values. They are now stored by numeric field id, like items and assets,
with the legacy name key kept as a rollback mirror for a few releases. Existing rows are migrated, writes are validated,
and the delete guard and cleanup scheduler cover both JSON columns (WI-2025).
The filter builder had the same class of bug in a different place: custom filter fields took their id from the
cf_<name> alias, so two custom fields sharing a name produced identical option keys and the field picker crashed with
each_key_duplicate. Filter fields now use the backend's collision-free cfid_<id> name for custom fields; the legacy
alias is still accepted for saved queries (GH #454).
Right-to-left text, Arabic, and Persian
Arabic and Persian are the focus of a large localization pass in this release, and it is explicitly a review pass. Both
locales are here to be used, tested, and verified, not declared finished. Expect rough edges in layout and wording, and
treat them as under review.
User text follows its own direction. User-generated text inherited the interface direction, so an Arabic title was
left-aligned in a right-to-left interface and mixed-language phrases reordered around it. Item titles, descriptions,
comments, custom-field values, portal request titles, dashboard rows, and list, board, roadmap, milestone, and iteration
titles now render with dir="auto" and align to their resolved direction. The shared text primitives (Badge, Chip,
Lozenge, Text, TruncatedFieldValue) follow their content direction, item keys stay left-to-right as code, and Input and
Textarea default to dir="auto" so typed text follows what is typed. A .bidi-isolate utility handles inline
mixed-direction runs, and a follow-up sweep covered data tables, the command palette, the rich-text editor root, and
one-off name sites in selects and chips (GH #461, WI-2038).
Arabic was reviewed and corrected. The Arabic locale carried literal machine-translation artifacts: a runner became
"racer", Notion became "the idea", Mermaid became the mythical creature, a permissions principal became "school
principal", and an approver pool became "swimming pool". Those, along with garbled sentences, untranslated English
blocks, and terminology drift, were corrected, and the remaining English fallbacks were filled so the locale covers
every key. Terminology was settled on milestone = معلم, labels = تسميات, backlog = قائمة المهام المعلقة, slug = المعرّف
المختصر, and organization = مؤسسة (WI-2035).
Persian is new and still in review. Persian (fa) ships as a right-to-left locale, mirrored key for key from English,
listed in the language picker, and negotiated from fa-* browsers. A terminology review corrected machine-translation
artifacts and filled empty help strings. The date picker is calendar-aware: it renders the Persian calendar with the
locale's week start and numerals, while the stored value stays a Gregorian YYYY-MM-DD string, so storage and APIs are
unchanged. The locale gate requires full key coverage so Persian cannot silently fall back to English, but its layout
and copy are still being verified (WI-2033).
Email follows the recipient. Transactional and notification emails are no longer English-only: built-in copy, digest
content, and dates render in the recipient's language and timezone, portal customers can choose their language, and
admins can override email templates per locale (WI-2034).
The Simplified Chinese (zh-CN) locale also received a mistranslation and terminology review (WI-2037).
Also in 0.9.4
- Board configuration merges instead of replacing. The navigation page toggled a board's view settings by
round-tripping the whole configuration through the full-replacePUT, which the strict v2 decoder rejected because
the fetched columns carry server-managed fields, so every toggle failed with "Request body is not valid JSON", and the
reset path zeroed columns, card fields, list columns, backlog statuses, and more. New
PATCH /collections/{id}/board-configurationandPATCH /workspaces/{id}/board-configuration
endpoints apply merge patches, writing only the fields present and reusing the existing write path
for authorization, validation, view-settings merging, auditing, and create-on-missing
(GH #456). - Attachments can be hidden from portal customers. A new
PATCH /attachments/{id}setsis_internal, so an editor
can hide or reveal an item attachment after upload instead of only at upload time (WI-2032). - Organization avatars upload again....
Windshift v0.9.3
Windshift 0.9.3
Windshift 0.9.3 turns milestones into real planning hubs and makes the hierarchy views
usable at scale. A milestone now carries a Markdown description, a comment thread, linked
pages, and an Activity feed, and the global milestones page lists everything you can see
with filters and an Updated column. The tree, map, and roadmap share one lazy loader, so
you can work at a chosen hierarchy level and expand only what you need. On the plugin side,
host functions and plugin routes are now default-deny, and a new editor extension point lets
a plugin replace the Markdown editor. Alongside those, the release preserves Jira issue keys
through import, stops imports from triggering live side effects, bounds a long list of read
paths, and fixes a batch of AI, email, and time-tracking bugs.
Important upgrade notice: if you run plugins, read the upgrade notes. Plugins must now declare
the host functions they use, and plugin routes require a session unless they opt into public
access. It is also advised to update the ws CLI alongside this release.
Highlights
Milestone planning hubs
A milestone is no longer just a target date. It has a Markdown description, a comment thread,
linked pages, and an Activity tab that merges milestone changes, comments, and the status and
comment activity of the items in it. The global milestones page lists every milestone the
viewer can see — global plus each accessible workspace — with name, workspace, and status
filters, a Created and an Updated column, and completed or cancelled milestones hidden by
default. The list opens in manual order and can sort by the latest activity.
Activity is cheap to sort because each milestone stores last_activity_at, set in the same
transaction as any activity write (description, status, target date, page links, comments,
and item membership or status changes). Moving an item to another workspace now keeps its
global milestones and drops only the old workspace's; the move preview says which are kept.
The API adds milestone comment, page-link, and activity endpoints, and the CLI gains
ws milestone comment list|add|edit|delete and ws milestone page list|add|remove.
Milestone comments and page links are audited, and comment permissions follow the item
comment rules: local milestones use their workspace's item.comment and comment.edit_others,
while global milestones let any viewer comment and only authors edit or delete their own.
Thanks to pienter-tech for this feature
(#382).
Hierarchy views work at scale
The tree, map, and roadmap now share one lazy hierarchy loader. A level picker chooses the
hierarchy level to show — with a filtered count per level and a sensible default — and the
base page loads at that level while subtrees load on expand. Each root pages with its own
offset, so a large subtree offers a per-row "Load more" instead of a single truncated page.
When a filter hides an intermediate parent, the matching child appears under its nearest
visible ancestor with a parent reference chip (GH #271) rather than a synthetic row. The
roadmap tree panel also gains expand and collapse all (GH #397). On the backend this adds
include_has_children on the item list, GET /items/level-counts, and POST /items/hierarchy.
Plugins are default-deny
Plugin host functions were previously registered for every plugin, so a plugin that asked for
little still received process execution, outbound HTTP and SMTP, KV storage, comments, and SCM
access. A plugin manifest now declares the permissions it needs from a closed catalog —
process.exec, network.egress, mail.send, storage.kv, comments.write, scm.write —
and anything it does not declare is denied. log stays available to every plugin, an unknown
permission name fails the load instead of silently disabling a host function, and when license
enforcement is configured the effective set is the declared permissions and a valid
license. The admin plugin list reports requested versus effective permissions, and Module
settings shows a badge per permission (#421).
Plugin routes had the mirror problem: they mounted on the raw mux with no authentication,
authorization, CSRF, rate limit, or body cap, so every enabled route was public. A route in the
manifest now declares its auth — session (the default), admin, or none — and the router
applies the matching chain, including a body cap and an IP-keyed limiter for public routes.
Public routes are additionally gated behind PLUGIN_PUBLIC_ROUTES, which defaults to off.
Plugin handlers receive an authoritative principal, and the admin list shows each route's
auth level.
Bring your own Markdown editor
A new editor.markdown extension point lets an installed plugin provide its own editor for an
item description or a knowledge page. A provider declares whether it toggles beside the native
editor or replaces it (the manifest also accepts comments as an entity type, though no surface
is wired to it yet). The host stays the only writer: it hands the plugin the current
Markdown, persists saves through the existing API, confirms before discarding unsaved edits, and
refuses a save over 256 KB. The native Milkdown editor is untouched when no provider is installed,
and the opt-in toggle only appears when an enabled plugin offers one. The plugin itself
ships from a separate repository
(#444, #445).
Also in 0.9.3
- Italian joins the language picker. The frontend now ships a complete Italian (
it) catalog —
all 8045 keys, with placeholders and plural forms matching the English reference — and
negotiatesit-*browsers automatically. Thanks to @m0nt0
(#449). - Jira issue keys survive import. An imported Jira issue keeps the numeric suffix of its
source key, so existing commit and PR references toPROJECT-123keep resolving after a
project import (GH #416). If the number is already taken or retired in the target workspace,
the importer falls back to the normal allocator and records a fidelity finding. - Deleted item keys are retired. Deleting the highest-numbered item in a workspace no longer
lets the next create reissue a key that commits and PRs already reference; the number is
reserved in the same transaction as the delete (WI-1958). - Loop transitions are real transitions. Jira teams that use same-status loop transitions no
longer lose them on import (WI-1954), and a newstatus_loopaction trigger fires when a
declared self-transition runs, with optional status and item-type filters. Undeclared
same-status updates stay plain no-ops (WI-1960). - Email intake and replies are easier to clear. A stuck intake receipt can be discarded by
its outbox row from the Processing Log (GH #371), and a channel's customer-reply tab offers
"Discard all pending (N)" to cancel a burst in one action while leaving in-flight deliveries
alone (WI-1967). - Imports never drive live side effects. Imported comments, items, and assets are stamped
with their origin, and the durable email, notification, item/asset action, and logbook
consumers skip them, so a Jira Service Management import can no longer queue one customer
email per imported agent comment (GH #439). - Completed iterations leave the backlog. Terminal (completed and cancelled) iteration
sections are hidden by default, with a per-view "Show completed iterations" toggle, and closed
sections reject drag-and-drop (GH #448). - Workspace lookups stop fanning out.
GET /workspacesaccepts anidsfilter, so a channel
with many allowed workspaces resolves its picker in one authorized read instead of one request
per id (#431). - Create-form pickers follow the workspace. The assignee and user custom-field pickers reload
when the workspace scope changes, so Editors and Viewers no longer see a denied, empty roster
on a reopened create form (WI-1961). - Time tracking survives customer changes. A worklog's customer is derived from its project
instead of stored on the row, so deleting a customer organisation no longer deletes time logs
for an unlinked project (GH #403), and the time customers view surfaces its delete guard error
instead of swallowing it (GH #402). - The tree reads correctly. One guide line is drawn per ancestor level and the parent chip
sits after the type icon, so nesting no longer appears to run backwards (WI-1955). - A bound sort without a direction is ignored. A table given
{ key: 'name' }no longer
shows a descending arrow while leaving rows unsorted; the next click starts ascending. - Settings pages share one scaffold. The catalog, status, role, group, and notification
managers moved onto a commonEntityManagerPage, with shared inline editor and detail-shell
scaffolds, so the settings surfaces behave and look consistent (WI-1940;
#411,
#413,
#414,
#415,
#417,
#418). - Shared frontend building blocks. The work-item list row and pagination footer, edge-drop
and rank recovery, progress detail, the set detail shell, mobile draft handling, the attachment
lightbox, and the inline editor scaffold were extracted into shared pieces, and the remaining
Label+Input sites moved toTextField(#401,
#404,
[#405](https://github.c...
Windshift v0.9.2
Windshift 0.9.2
Windshift 0.9.2 makes the dashboard extendable and the board quicker to navigate. You can describe a
custom widget in plain language and Windshift generates a validated, permission-aware program for
it, backed by a component library of cards, metrics, tables, charts, and graphs. Boards gain
multi-select bulk actions, keyboard navigation, saved filters, milestone swimlanes, and per-user
views that follow you across devices. Alongside those, the release tightens portal, on-call, and
cross-workspace authorization, makes email intake and outbound mail sturdier, adds optional
S3-compatible attachment storage, and moves more work off the collection paint path.
Important upgrade notice: It is advised to also update your ws cli if you are using it with the latest
Windshift version to ensure api compatibility. Read below for the fixes.
Highlights
Custom dashboard widgets
The dashboard is no longer limited to the built-in catalog. A new "Create custom widget" entry lets
you describe what you want in natural language; the model produces an OpenUI program, you preview
it at the real tile width, refine it, adjust the CQL queries, and save it.
The component library ships Card, Section, Title, Text, Divider, Badge, EmptyState, Metric, MetricRow,
ItemsTable, ItemList, BarChart, LineChart, PieChart, ProgressList, ActivityFeed, and LinkButton,
plus canvas-rendered network and matrix graphs.
Data access stays yours. Item queries resolve through the list service with the viewer's
accessible-workspace list, asset queries resolve against a builder-selected asset set.
Custom widgets are model-generated, so the catalog offers them only when a model is configured and
the feature is enabled. The AI pipeline meters generations against a per-user rate limit (20 generations / hour),
validates the result with the real CQL parser and a dry run, and repairs positional errors before giving up.
Boards get even better
Boards get an explicit multi-select mode. Selection checkboxes appear only while the mode is active,
selected cards stay opaque while the rest dim, and a sticky bulk toolbar carries status, assignee,
priority, and label actions with a progress and error summary. You can also Shift-Click on a card to
enter selection mode.
Per-user collection view state — show-completed, group-by, collapsed columns, sort mode, iteration
filter, tree and map preferences, and personal saved filters — is now stored in user preferences
rather than browser-local storage, so it follows you across devices. Existing local values remain
readable as a fallback, and every new change is written to the server.
Milestone swimlanes
Boards can group by milestone, with a configuration dialog to pin global milestones and choose
whether completed milestones are shown. Cards on hidden completed milestones collect into a single
completed-milestones lane instead of disappearing, and a leading rescue lane surfaces open items
whose primary milestone is completed or canceled so the mis-assigned work can be moved to the right
milestone. The milestone picker is searchable, and empty global lanes stay hidden unless pinned.
Multi-value milestone fields
A new multi_milestone custom field type stores an array of milestone IDs, with validation, editing
and rendering in the item sidebar, filtering, and CQL support. CQL also resolves bare custom-field
names such as affectedVersion = 1 in addition to the cf_, custom., and cfid_ forms.
This field serves primarily to migrate existing Jira issues more faithfully.
Attachments can live in S3
Attachment blobs can now be stored in any S3-compatible bucket instead of local disk. Set
STORAGE_BACKEND=s3 together with S3_BUCKET and S3_REGION, and Windshift routes uploads,
downloads, previews, and diagrams through the new blob store. Static credentials are optional, so
AWS deployments can use the standard credential chain (environment, web identity/IRSA, instance
profile); self-hosted stores such as MinIO, RustFS, and Wasabi set S3_ENDPOINT and usually
S3_FORCE_PATH_STYLE=true. Downloads stay authorized and are proxied through Windshift with ranged
reads, and /readyz reports bucket reachability from a cached probe. Local disk remains the
default and existing deployments need no change. Thanks to Marco
(@valentimarco) for this feature
(#378).
Also in 0.9.2
- Sidebar order follows you. Workspaces in the sidebar are ordered by your most recent visits
instead of a fixed order, and the recent order is preloaded before the menu opens so the list does
not reshuffle as you navigate. - Start and end dates as list columns. Collections can show the start and end date system fields
as list columns alongside the existing date columns (GH #391, thanks to
@Jayden-SysData). - Comment sort order sticks. The newest-first toggle on comments is stored in your user
preferences and loads with the rest of your settings instead of resetting on every visit. - One request for global catalogs. A new
GET /api/v2/catalogs/global-bootstrapreturns item
types, statuses, categories, priorities, and optional custom fields in a single response, and the
global workspace scope hydrates from it instead of five separate calls. Hydrated user-display
records are preserved across refreshes and only reset on a scope or auth change. - Item type icons in key badges. Homepage activity feeds carry item type metadata, and the
shared dashboard row renders the colored type icon next to the key. - Jira connections are recoverable. Saved connections can be re-verified with a replaced token,
removal soft-deletes the connection so import jobs and ID mappings keep their provenance, and
destructive import actions move behind a row menu with a soft "Remove from list" alongside the
confirmed data deletion. - Saved queues are editable again. Editing a saved queue sent the wrong content type and failed
with a 415; the merge-patch request and the workspace default-view selector now both work
(GH #386, thanks to @jetedv). - Source-control credentials stay in sync. A refreshed OAuth grant is mirrored between the user
and workspace copies, so a periodic repository sync no longer leaves the user copy dead and
milestone releases failing (GH #387, thanks to @Jayden-SysData). wsproject-config safety. A[server]section in a git-trackedws.tomlis ignored with an
explanation, so a repository cannot redirect a token to its own host, while a user's own
untracked file is still honored.ws config init --non-interactivenow readsWS_URLand
WS_TOKENfrom the environment as documented.- Due-date reminders are opt-in. Assignees can turn on due-date reminders in their profile and
choose how many days ahead (1–30) to be notified; reminders are never sent unless enabled. A
background sweep writes one notification per item, due date, and assignee, so repeated ticks and
restarts do not duplicate a reminder while a changed due date re-reminds. The sweep interval and
batch size are configurable withWINDSHIFT_REMINDER_INTERVALand
WINDSHIFT_REMINDER_BATCH_SIZE(GH #392, thanks to
@Jayden-SysData). - Kubernetes deployment example.
deploy/kubernetes.mddocuments a copy-and-adapt set of
manifests underdeploy/kubernetes/, including the single-replica constraint, an SSE-safe
ingress, and the shape of the required secret.
Performance
- Collection pages appear faster. The unfiltered item count is no longer computed while the page
renders; it is deferred and filled in shortly after. The workspace directory is warmed in the
background, card details load only when a card scrolls into view, field availability comes from
the shared effective-config cache, and assignee names for a whole page load in a single authorized
request. - The event engine's idle ticks are gated with a single statement per tick, and consumer cursors are
pruned against the active set every tick so inactive consumers do not leave stale entries
(@Faks). - The global workspace scope no longer clears hydrated users on refresh, and reference catalogs load
in one request instead of five. - Free-text item search is indexed. Title and description no longer fall back to a
LIKE '%term%'
scan; token-prefix matches use an FTS5 index on SQLite and a GINtsvectorindex on PostgreSQL, so
search stays fast as the item table grows. - Permission checks stop serializing. The permission snapshot cache stores typed snapshots in a
sharded, count-bounded cache with a lazy TTL and background sweep instead of marshalling and
decoding the whole snapshot on every check, and the Everyone fallback is stored once per process
rather than duplicated per user. - Notification fan-out decodes once per recipient. Recipient filtering resolves each recipient's
permission snapshot once and batches theis_activeprobe, instead of one decode and one query
per recipient per event. - The mobile pages list and both mobile search surfaces resolve workspaces in one request per
50-workspace batch instead of one request per workspace, and the portal channel settings page
reads connected mailboxes from a single channel-list request. - Applying labels to a multi-item board selection sends one request instead of one per selected
item. - Rate limiting no longer serializes every request on a process-wide write lock: the unified limiter
uses a sharded, count-bounded token-bucket map with an atomic last-seen timestamp.
Security
- The audit pipeline...
Windshift v0.9.1
Windshift 0.9.1 - Hotfix
Windshift 0.9.1 is a hotfix for 0.9.0. It fixes a UI freeze introduced in 0.9.0, where opening a
work item from a collection could lock the entire page. The hotfix also includes the community work
and fixes that landed after 0.9.0 shipped: emoji reactions on comments, watchers and a "Created by
Me" dashboard section, plus fixes for Microsoft 365 email intake, helpdesk first response, and
setup passwords.
Highlights
React to comments with emoji
A comment can now be acknowledged without writing a reply. Six reactions are available: thumbs up,
check, eyes, heart, laugh, and hooray. They show up as chips under each comment, with your own
reaction highlighted, a picker to change it, and a tooltip that names who reacted. The add and
remove endpoints are idempotent and gated the same way as commenting, grouped reactions are
included in comment list, get, update, and batch responses, and a reaction publishes an item change
so open views update live. Internal notes and approval rows never carry reactions
(@dpb97).
Watch a work item
Item detail now shows how many people watch it and lets you add or remove watchers right there. The
API has list, add, and remove watcher endpoints, and the section works on desktop and mobile
(@danielraab).
A "Created by Me" dashboard section
The personal dashboard can show a section for items you created, next to personal tasks and
assigned-to-me. It is part of the default work layout and can also be added from the widget catalog
(@danielraab).
Bug fixes
- Opening a work item from a collection no longer freezes the UI. Items opened from a collection
board, list, or tree, or from the support queue, could leave the page stuck on its loading state
with aneffect_update_depth_exceedederror in the browser console. The shared editor-option
cache was rewritten on every render, so Svelte's update loop never settled. Priming the cache is
now idempotent and runs outside the reactive dependency, so the view settles after the first pass.
The freeze affected items in workspace and global collections, including items created after
0.9.0; personal task cards were unaffected (GH #365). - Passwords longer than 72 bytes are rejected with a clear message. Completing setup with a
passphrase over bcrypt's 72-byte limit returned an internal 500, and the setup page showed a
generic failure. The length is now checked before hashing, so setup returns a message that
explains the problem (@Faks). - Microsoft 365 inbound email connects again. The OAuth flow asked for mailbox-only scopes and
then tried to read the mailbox identity from Microsoft Graph, which rejected the token, so every
connection failed with "Couldn't read the connected mailbox identity" (GH #363). The monitored
mailbox is now entered as configuration before the flow starts, the callbacks no longer look up an
identity, and only the scopes the flow uses are requested. Stored credentials are cleared when the
mailbox or OAuth client changes, and a connection verified for one address cannot be enabled for
another. - First response counts the first public agent reply. The helpdesk pack stopped its
first-response clock on assignment, so a ticket that was answered but never assigned never met the
metric. It now stops on the first public agent comment, and internal notes no longer satisfy it. - Go 1.27.3. The bundled toolchain moves to 1.27.3 for the
crypto/x509fix in GO-2026-6809.
Performance
- Route changes no longer flash the full-page loading spinner. The loading state waits 175 ms, and
route chunks are prefetched on hover, focus, and idle. - Comment feeds read reactions only when they render them, and grouped reactions are served from a
short-TTL cache that is invalidated by reaction writes and comment deletes.
Upgrade notes
- Drop-in replacement for 0.9.0. No database migration, configuration change, or re-login is
required. - Email OAuth channels. The monitored mailbox address is now a field in the channel
configuration. Existing connections keep working. If you reconnect, enter the mailbox address
before starting the OAuth flow.
Thank you
Thank you to Daniel (@danielraab) for watchers and the "Created by
Me" dashboard section, to Dominik (@dpb97) for emoji reactions on
comments, and to Oskars (@Faks) for the setup-password fix. Thanks also
to everyone who reported the collection freeze, the Microsoft 365 intake failure, and the other
issues fixed in this hotfix.
Windshift 0.9.0 - "Sea Trials"
Windshift 0.9.0 makes the helpdesk a first-class part of the product. Service levels arrive with
metrics, goals, and working calendars. The support queue becomes a configurable workspace view with
live counts and bulk actions. Email intake is rebuilt around separate mailbox connections and
routing. Tickets can be merged and split, and portal customers can share requests across their
organisation.
The release also adds customer data export and retention-aware erasure, turns incidents into
work-item state, marks AI chat changes in history, moves workspace bundles and packs between
instances, makes navigation configurable, and speeds up large installations. Five rounds of release
bug hunting also closed a number of security issues before they shipped.
Highlights
Service levels
Service levels are here. Every item is evaluated as it changes, so SLA cycles start, pause, and stop
immediately, and breaches and warnings are delivered through Windshift's event system.
Workspace admins configure metrics, conditions, and goals under Service levels, using Windshift's
query language with validation at save time. Goals can be grouped by priority, and working calendars
define timezone, weekly hours, and holidays. Calendar changes preview their impact, and a running
cycle keeps the calendar it started with, so recalculation never rewrites history. Teams bind to
workspaces, and coverage gaps surface when a team's hours do not line up. Warning thresholds,
pause/stop/start rules, and on-demand recalculation complete the admin surface.
Items show an SLA panel, lists and boards carry SLA indicators, and a compliance report summarizes
completed, breached, and average cycles. SLA triggers and events work in automations and
notifications, and the configuration is available through API v2 and the ws CLI. Jira SLA
configuration and cycles import directly.
The support desk comes together
Support queues become a real workspace view. Queues are scope-aware and configurable: build them
with a query-language filter builder, route by collection, choose which columns show, and see live
counts for unassigned, team-owned, and waiting tickets. Bulk actions are supported, and the queue is
a toggleable navigation entry rather than a fixed tab.
Agents get canned responses to insert into replies, helpdesk triggers and customer actions in the
automation editor, and a customer context sidebar on external requests. Request participants appear
only on external requests, and comments and attachments can be marked internal so they never reach a
customer.
Support metrics arrive as dashboard and workspace widgets: first response, resolution time, reopen
rate, volume, backlog trend, and SLA compliance. First response counts the first public agent
comment, and resolution time is recorded once, on first close. A built-in helpdesk pack provisions
the whole configuration in one apply. Zammad users also get a support overview with ticket history
and status counts (@Optic00,
#306).
Email intake, redesigned
Email channels are split into two clear concepts. A mailbox is a connection: inbound address and
IMAP or OAuth credentials, with no routing or target. An intake is a feed from a folder on a
mailbox that turns new mail into items in one workspace (or a portal), with its own item type,
request type, rate limit, and processing rules. Credentials belong to the connection, and
permissions are checked against both the mailbox and the target. The SMTP AUTH mechanism is
selectable per channel, so servers that only offer LOGIN or CRAM-MD5 work without a global
override (@m0nt0).
The email and portal conversation loop is closed. A portal submission can email the customer, the
customer's reply threads back onto the same ticket, and a new email from someone with open tickets
no longer creates a duplicate. The outbound reply queue is visible and operable: retry a stuck
reply, discard one that should not send, and watch pending and exhausted counts in the metrics. A
per-sender hourly cap holds floods instead of dropping them, and operators can requeue limited mail
after raising the limit. An intake can send an optional submission receipt and link the resulting
request into the portal.
Reply routing, threading, and reply authorization were reworked, and OAuth callbacks are bound to
the account that started the flow, so a forwarded authorization URL can no longer attach someone
else's provider credentials to your identity (WI-1641).
Ticket merge and split
Tickets can be merged into a canonical one and split into a subticket. Merging moves comments,
attachments, links, watchers, email threading, and pending replies, refuses to orphan subtickets,
and marks comments private when requesters differ so one customer never sees another's content. The
duplicate remains as a read-only redirect to the canonical ticket. Splitting creates a child ticket
from selected comments and attachments with an explicit assignee and r...
Windshift v0.9.0
Windshift 0.9.0 - "Sea Trials"
Windshift 0.9.0 makes the helpdesk a first-class part of the product. Service levels arrive with
metrics, goals, and working calendars. The support queue becomes a configurable workspace view with
live counts and bulk actions. Email intake is rebuilt around separate mailbox connections and
routing. Tickets can be merged and split, and portal customers can share requests across their
organisation.
The release also adds customer data export and retention-aware erasure, turns incidents into
work-item state, marks AI chat changes in history, moves workspace bundles and packs between
instances, makes navigation configurable, and speeds up large installations. Five rounds of release
bug hunting also closed a number of security issues before they shipped.
Highlights
Service levels
Service levels are here. Every item is evaluated as it changes, so SLA cycles start, pause, and stop
immediately, and breaches and warnings are delivered through Windshift's event system.
Workspace admins configure metrics, conditions, and goals under Service levels, using Windshift's
query language with validation at save time. Goals can be grouped by priority, and working calendars
define timezone, weekly hours, and holidays. Calendar changes preview their impact, and a running
cycle keeps the calendar it started with, so recalculation never rewrites history. Teams bind to
workspaces, and coverage gaps surface when a team's hours do not line up. Warning thresholds,
pause/stop/start rules, and on-demand recalculation complete the admin surface.
Items show an SLA panel, lists and boards carry SLA indicators, and a compliance report summarizes
completed, breached, and average cycles. SLA triggers and events work in automations and
notifications, and the configuration is available through API v2 and the ws CLI. Jira SLA
configuration and cycles import directly.
The support desk comes together
Support queues become a real workspace view. Queues are scope-aware and configurable: build them
with a query-language filter builder, route by collection, choose which columns show, and see live
counts for unassigned, team-owned, and waiting tickets. Bulk actions are supported, and the queue is
a toggleable navigation entry rather than a fixed tab.
Agents get canned responses to insert into replies, helpdesk triggers and customer actions in the
automation editor, and a customer context sidebar on external requests. Request participants appear
only on external requests, and comments and attachments can be marked internal so they never reach a
customer.
Support metrics arrive as dashboard and workspace widgets: first response, resolution time, reopen
rate, volume, backlog trend, and SLA compliance. First response counts the first public agent
comment, and resolution time is recorded once, on first close. A built-in helpdesk pack provisions
the whole configuration in one apply. Zammad users also get a support overview with ticket history
and status counts (@Optic00,
#306).
Email intake, redesigned
Email channels are split into two clear concepts. A mailbox is a connection: inbound address and
IMAP or OAuth credentials, with no routing or target. An intake is a feed from a folder on a
mailbox that turns new mail into items in one workspace (or a portal), with its own item type,
request type, rate limit, and processing rules. Credentials belong to the connection, and
permissions are checked against both the mailbox and the target. The SMTP AUTH mechanism is
selectable per channel, so servers that only offer LOGIN or CRAM-MD5 work without a global
override (@m0nt0).
The email and portal conversation loop is closed. A portal submission can email the customer, the
customer's reply threads back onto the same ticket, and a new email from someone with open tickets
no longer creates a duplicate. The outbound reply queue is visible and operable: retry a stuck
reply, discard one that should not send, and watch pending and exhausted counts in the metrics. A
per-sender hourly cap holds floods instead of dropping them, and operators can requeue limited mail
after raising the limit. An intake can send an optional submission receipt and link the resulting
request into the portal.
Reply routing, threading, and reply authorization were reworked, and OAuth callbacks are bound to
the account that started the flow, so a forwarded authorization URL can no longer attach someone
else's provider credentials to your identity (WI-1641).
Ticket merge and split
Tickets can be merged into a canonical one and split into a subticket. Merging moves comments,
attachments, links, watchers, email threading, and pending replies, refuses to orphan subtickets,
and marks comments private when requesters differ so one customer never sees another's content. The
duplicate remains as a read-only redirect to the canonical ticket. Splitting creates a child ticket
from selected comments and attachments with an explicit assignee and requester, with no silent
inheritance. Both operations appear in the audit trail like any other change.
Ticket imports improved too. CSV ticket import is available with asset-set portal access, and the
Jira import can map onto existing entities interactively instead of only creating new ones.
Imported customers record their provenance, and search uses Jira's enhanced JQL API (GH #301).
A portal customers can share
Portal customers belong to organisations, and 0.9.0 gives those organisations a real visibility
model. Global admins define contact levels, each organisation chooses who may see shared requests
and who decides sharing, and requests can be shared individually, shared automatically, or not
shared. The creator and explicit participants always keep access, and shared viewers can
participate. Cross-organisation isolation applies to both lists and individual requests, so guessing
an ID fails the same way as leaving it out. Customer management appears in navigation as soon as a
portal exists.
The customer data lifecycle is much more complete. Auto-created customers record their origin and
can be cleaned up in bulk, and a customer can be deactivated for security. Article 15/20 data export
ships for portal customers, erasure respects retention rules and records DSAR evidence, and
background jobs clean up expired portal credentials and knowledge-base events. Requests support
attachments and reply notifications, and asset reports can prefill a request type from row actions.
Incidents and on-call
Incidents are now work items with a dedicated team field, so they live in the same views,
permissions, and history as everything else. Acknowledge, unacknowledge, and resolve handle
concurrent updates safely, and unacknowledging restarts escalation from the current step.
AI you can audit
When the AI chat changes an item, the history entry records it, including the model and cost behind
the change (@valentimarco). Chat turns are metered
like other provider calls, "catch me up" now includes parent and sub-item context, and the item AI
actions menu has a new icon. Agent readiness errors now render properly and point at the
permission grants that fix them. OpenCode Go joins the native LLM providers.
Also in 0.9.0
- Workspace bundles, config sets, and packs. Whole workspaces can be exported as bundles and
imported into another instance, including their configuration sets, with a conformance check that
reports drift and can repair it. Configuration-set templates carry link types, imported config
sets are adopted rather than duplicated, and a failed pack apply rolls back instead of leaving a
half-provisioned workspace. Framework packs get a manifest and apply flow, and the built-in
helpdesk pack ships with the release. - Navigation you control. A new navigation configuration view lets workspaces and global
collections choose what appears in their sidebar and in what order, with per-scope visibility
toggles. The support queue is a toggleable entry, workspace-only tools stay hidden where they do
not apply, and saved settings carry over cleanly as the toggleable set grows. - Delegated administration and restricted workspaces. Management surfaces for non-system-admins,
such as channel management and asset sets, move behind a delegated admin area that appears only
when the user holds a delegated capability. Workspaces can be created with visibility restricted
to the creator, and the admin directory gains a Visibility column so the setting is discoverable. - Collections are shared by default. Collections are now visible to the people who can see their
items: workspace members for workspace-linked collections, every authenticated user for global
collections, with an owner-only Private opt-out. Marking a collection private also unpublishes its
public board. - A portal knowledge base. A channel manager can publish a workspace's Pages tree, either the
whole workspace or a sub-page subtree, into a portal's knowledge base. Published articles get
deep-linkable routes and a browse listing, agents can insert knowledge-page links into replies,
and the portal resolves those links only for pages it publishes. Usage is recorded for reporting. - Helpdesk operations. New Prometheus metrics expose outbox depth, the age of the oldest queued
reply, exhausted replies, rate-limited intake, and poll errors, with a runbook covering setup,
mailbox and outbox recovery, and SLA operations. The audit trail records merge, split, assignment,
bulk-update, agent-reply, automation-run, and knowledge-base publish decisions, and automation
comments no longer re-trigger the rule that created them. - @mention emails. Mentioning someone in a comment sends a dedicat...
Windshift v0.8.9
Windshift 0.8.9
Windshift 0.8.9 adds GitLab as a supported SCM provider, ships a Zammad ticket integration, introduces BDD (Gherkin)
test cases to the test management suite, turns workspace pages into a real knowledge base for the customer portal, and
makes the entire web interface available in Korean. The public API v2 keeps expanding to cover workspace configuration,
custom fields become editable in list views, workspace admins gain credential and theme tooling, the mobile app gets a
native-style rework of its create and edit flows, and new GDPR tooling handles Article 17 erasure requests with
auditable evidence. A release-long consistency pass rebuilds forms on shared design-system primitives, makes the
interface respect the permission model in every corner, keeps large installations fast, and fixes dozens of smaller
rough edges across actions, imports, custom fields, and the portal.
GitLab SCM support
Windshift now connects to GitLab as a source-control provider alongside GitHub
(@oshuej198). Configure a GitLab instance with a personal access token
— including self-hosted installations via an optional base URL — and browse available repositories, link them to
projects, and create pull requests from branches. Large instances get server-side repository search, so the picker can
find projects on a busy GitLab without paging through the unfiltered list, and the picker debounces input so keystrokes
do not hammer the SCM API.
Zammad ticket integration
Zammad tickets can now be linked to Windshift items (@Optic00). The
integration synchronizes ticket state and owners, keeps link invariants intact, and adds return links so a ticket points
back to the item it belongs to. Tickets are reconciled every minute in batches, administrative settings cover the
default group selection, and owner discovery is hardened against edge cases such as missing or ambiguous agents.
The portal knowledge base
Workspace pages become the customer portal's native knowledge base. Articles move out of the hero-search modal onto real
routes, with a browsable listing in stable tree order and search results that land on durable URLs. Agents answering a
portal request can insert knowledge-page links from a permission-filtered picker; a link becomes an in-portal article
link when the page is published through that portal and degrades to plain text otherwise, so customers never see a dead
link or a hint at hidden pages. The portal records view, search, miss, and deflection signals for the knowledge base,
articles render through the same read-only editor as internal pages, the channel's start page is picked with a
searchable page picker, and publications get a full-bleed banner.
BDD test cases
Test management learns behavior-driven development. Test cases can now be authored as Gherkin scenarios — Feature,
Background, Scenario, and Scenario Outline with Examples — and the case detail view renders those scenarios alongside
the regular steps. When a test run executes a scenario outline, each examples row runs as its own case, so data-driven
behavior is tested and reported per example. BDD test cases are exposed through API v2 as well.
The underlying Gherkin parser accepts the supported English-keyword subset with data tables, doc strings, tags, and
comments, and rejects unsupported constructs with precise line and column errors instead of silently dropping them.
GDPR erasure and user offboarding
Administrators get a dedicated, irreversible erasure flow for GDPR Article 17 requests: erasing a user first offboards
them, marks the account as erased, and persists a DSAR completion record with intake evidence, approving admin, and
policy version. Erasure is distinct from administrative offboarding, which is itself now strictly irreversible and
complete, and no activation path can revive an erased account. Audit logs intentionally keep their pseudonymized
identifiers, and work history is retained under a deleted-user-N pseudonym until the containing record is deleted.
API v2 keeps expanding
The public API v2 now exposes the configuration surface: configuration sets, workspace roles, screens and screen-field
configuration, custom-field provisioning, and hierarchy levels, with localized screen responses. The admin frontend
delegates its custom-field and screen mutations to the same shared services the API uses.
Separately, the AI chat and daily briefing endpoints are now reachable with REST v1 bearer tokens through the ai:chat
and ai:read scopes, and the generated OpenAPI contracts were regenerated to match.
Item updates treat custom_field_values as a per-field patch: keys in the payload set their field, a null value clears
it, and keys left out are preserved server-side — a partial payload no longer wipes the fields it does not name, and
history records exactly the fields a patch touched.
Mobile, reworked
The mobile PWA gets a native-style editing pass. Status and assignee edits use bottom sheets with drag-to-dismiss,
creating an item is a dedicated page instead of a modal, and the item detail is editable for the first time, with an
unsaved-changes guard. Pages, search, and the command palette are available in the mobile shell, every mobile input is
floored at 16px so iOS no longer zooms on focus, and desktop deep links redirect to the mobile surface on phone
viewports. A later fix breaks an infinite template-effect loop in the create page.
Korean localization
The web interface is now fully available in Korean (@tttaliesin). A new
ko locale translates every surface — navigation, dashboard, channels, admin, analytics, workflows, workspaces, teams,
and more — and 한국어 appears as a first-class option in the language picker.
Story points rollup
The dashboard gains a story-points-by-assignee widget backed by a new aggregate endpoint (GH #255), and item summaries
roll story points up recursively from children, with a hint on the item detail sidebar showing how the total is composed
(GH #256).
Custom fields in list views
List views now render custom fields through the same renderer as the item detail, and scalar fields — text, numbers,
dates, selects — are editable right in the cell. Pickers resolve labels for selected values that are not in the lazily
loaded option list, so triggers and chips never show raw IDs, and row-level editing no longer leaks option caches across
mixed-workspace collections.
Automation and configuration
Automations on parent transitions can now condition on the count of open children, whether an item sits inside a
parent's subtree, and on milestone fields, and type-filtered item.updated triggers fire on any field change instead of
only when filtered fields change. Effective configuration — workflows, status lists, and similar workspace settings —
resolves through a canonical path with a shared workspace cache and falls back to the default config set, which also
fixes status lists and workflows disappearing in workspaces without an explicit config-set assignment.
Workspace administration
Workspace admins get a credentials module in workspace settings: action credentials can be created, rotated, and deleted
per workspace, while inherited all-workspace or shared rows render read-only. Workspace admins can also assign
configuration sets to their own workspace, and themes gain an optional company logo upload so the application can carry
your branding.
Permissions improvements
The interface now respects the permission model everywhere it renders an affordance. Without item.edit, the item title
renders read-only, board cards do not drag, and the detail page hides its write controls; quick-add and the create modal
require item.create; moderating other people's comments requires comment.edit_others; test-management affordances
follow test.manage and test.execute; page management appears only with effective page permissions; and
board-configuration palette commands and role assignment require workspace.admin. The server draws the same lines.
Performance and scale
A release-long pass bounds the queries that used to grow with data: Todoist sync loads a bounded number of links per
run, recurrence deduplication looks only at the generation window, notification email batching reads bounded chunks,
pull-request link refresh scans page by keyset, and the auth policy affected-users listing paginates. The metrics
collector caches its domain summary between scrapes, event deliveries reconcile from per-consumer cursors, the global
rank migration derives its progress incrementally, and workspace deletion cleans up more thoroughly.
Around the app
- The TRMNL summary plugin gains a next-up milestone, a full-height layout, and a closed-today count on the completed
board. - The personal task history filter for completed tasks gains a None option.
Consistent design system
Forms across the app are rebuilt on shared primitives: TextField replaces ad-hoc Label+Input pairs, dialogs use
TabStrip and ModalBackdrop, loading and error states render through StateDisplay, and list views adopt Badge and
Modal with stable test ids. Hardcoded Tailwind colors are replaced with tokens, dark-mode neutral hover colors are
corrected, decorative transitions and animations are removed, and hover styles move from JavaScript mutations to CSS.
The portal routes its approval confirmations through the shared confirm composable — with the global dialog properly
mounted — and the workspace admin navigation appears immediately after creating a workspace (GH #258).
Reliability and fixes
- Creating an action honors the
is_enabledfield on the create request, so automations can be c...
Windshift v0.8.8
Windshift 0.8.8
Windshift 0.8.8 makes API v2 the primary public interface for integrations,
adds a supported backup and restore workflow for Docker deployments, ships
French and Russian, and hides completed work by default so your views focus
on what is still open. Notification and email delivery were rebuilt to
survive restarts, and a release-long hardening pass fixes dozens of smaller
rough edges across boards, imports, mobile, and the terminal UI.
Upgrading
- If you are running 0.8.5 or newer, you can upgrade directly. This release
requires the schema checkpoint that 0.8.5 records the first time it
starts, so installations on 0.8.4 or older must first upgrade to 0.8.5 or
0.8.5-2, let it start once, and then move to 0.8.8. As always, back up
your database before upgrading. - Building Windshift from source requires Go 1.27.0, unchanged from 0.8.7.
API v2 is now the primary public API
API v2 has been growing under the hood for several releases; in 0.8.8 it
graduates. Windshift's own frontend now runs on v2, and GitHub issue
synchronization goes through the same item services the app itself uses,
so the API your integrations use is the same one the product uses every
day. Items can be fetched directly by their key (such as
WI-123), editing endpoints accept merge-patch requests, pagination
behaves consistently across endpoints, and the generated reference now also
documents the operational endpoints.
The in-product API reference opens on v2 and groups operations by domain.
API v1 remains available for existing integrations, but the reference now
marks it deprecated.
The command line meets v2 halfway: a new ws api command sends
authenticated requests straight to API v2 for anything the CLI does not
cover yet, and ws task ls gained pagination (--page, --limit, --all)
plus status filters that accept workspace status names such as
-s "In Progress"
(@Optic00). The Windshift
TRMNL plugins now talk to API v2 as well.
Back up and restore Docker deployments
Docker Compose deployments get a documented backup and restore workflow,
contributed by @Optic00.
A single script creates timestamped backups of the entire data directory
and, for the supported PostgreSQL setup, a logical database dump, with
checksums and a manifest. Restores run preflight checks and take a rollback
snapshot before touching your data. See deploy/backup-restore.md for the
supported configurations and the exact commands.
French, Russian, and localizable object names
Windshift speaks two more languages: French
(@Sinago87) and Russian
(@oshuej198). The
dashboard (@Optic00) and
the Pomodoro timer settings
(@Optic00) are translated
as well.
Administrators can additionally give built-in objects translated names per
language: statuses, item types, link types, and similar names accept
localized values while the canonical names remain as fallback.
Views that focus on open work
Completed work items are now hidden by default in list, tree, map, roadmap,
and backlog views. A toggle brings them back, and each view remembers your
choice per workspace or collection. Boards are unchanged and keep using
their own configuration for completed columns.
The tree view also keeps every hierarchy whole: when the completed toggle or
a page boundary would otherwise show items detached from their parents, the
missing ancestor chains are loaded so children always render nested under
their root instead of as stray top-level rows.
Every page now carries persistent workspace breadcrumbs: a workspace
picker, a link back to the current workspace, and a highlight of where you
are in the navigation. The main and workspace sidebars scroll, collapse,
resize, and restore avatars more steadily
(@oshuej198 among several
fixes).
Boards got their own pass: the board auto-scrolls while you drag a card,
default column ordering is fixed, background refreshes no longer drop rows
you had loaded with "load more", and switching from a board to another view
no longer leaks the board's completed-status filter into that view.
The test management screens were reorganized end to end, with correct
test-case and step counts and clearer create controls.
A long list of smaller annoyances is also fixed: Markdown list markers
render again (@codekoala)
and the empty-list caret stays aligned
(@codekoala); workspace
creation is reachable from the quick-access menu
(@Optic00); milestone
pickers cancel their requests when you navigate away; collection views
discard responses that arrive after you have moved on; asset configuration
forms can no longer submit overlapping saves or apply responses from a set
you just switched away from; asset queries offer autocomplete; burndown
charts can use story points; category forms, customer management headers,
license page scrolling, the custom-field type picker, and action editor
layouts are restored; new AI chats start with a fresh context instead of
inheriting the previous conversation; date pickers stay open while you edit;
and items without a due date are no longer worded as if they were overdue.
When an item is deleted, everyone viewing it is now notified and the view
navigates away cleanly instead of silently going stale
(@Optic00).
Notification and email delivery, rebuilt
Notifications and their emails are now processed from durable queues:
deliveries survive restarts, older messages no longer starve behind a
stream of newer ones, and a crash mid-delivery can no longer lose a
message. Notification emails render their subjects and plain text without
HTML-escaping artifacts, threaded email replies reliably attach to the
right comment, deactivated users stop receiving notifications and emails
entirely, and the in-app inbox no longer stops at the first page of
results.
For self-hosters, outgoing email channels can now use an unencrypted SMTP
relay for trusted internal relays that offer neither TLS nor
authentication. The option is clearly marked as plaintext in the channel
settings.
Signing in with SSO
SSO providers can now pass through their own email_verified attribute for
OIDC and SAML
(@danielraab). The
accompanying trust setting was reworked into a clearer contract: trusting
provider-managed email addresses skips Windshift's own verification email
when your provider does not report verification status, while addresses a
provider explicitly reports as unverified must still verify. Risky
combinations, such as self-registration with untrusted email, now carry an
explicit warning, and provider configuration errors appear inside the
provider dialog instead of failing silently.
Jira import
Imports that hit problems now finish with a distinct "completed with
errors" outcome that reports how many projects, issue links, and test cases
failed, instead of a deceptively successful summary. Jira custom-field
values that cannot be mapped are preserved raw on the imported item rather
than dropped, story points map correctly, and Xray tests are counted
accurately. Deleting an import is thorough and safe: it removes everything
the import created, including statuses, item types, screens, custom fields,
and attachments, while sparing shared objects that other work still
references, and a partially failed cleanup can simply be retried.
Mobile and the terminal
The mobile PWA installs with richer metadata, recovers to cached content
when the server is unreachable, no longer hangs indefinitely on slow
networks, applies updates more cleanly, and keeps the create dialog above
the navigation bar.
The SSH terminal UI handles disconnects, cancellation, and terminal
lifecycle transitions more reliably, and your theme choice is no longer
overwritten by the server's saved preferences.
Security and privacy
- Revoking access now takes effect everywhere it should. Workspace-derived
content is reauthorized before reuse, item reads enforce collection
visibility, live item streams are rechecked after items move to another
workspace, project worklog listings redact details of items you can no
longer see, and comment editing and deletion enforce current workspace
access. - v2 asset responses no longer expose the creator's email address
(@Optic00). - Dependency updates close known advisories in
golang.org/x/crypto, gRPC,
and the demo application.
Diagnostics and reliability
The diagnostics page gains two new sections for administrators: SCM
connection health, with persisted failure history for scheduled repository
syncs and pull-request refreshes, and domain event consumer health, so
stuck or failing background consumers are visible at a glance. Behind them,
domain events, action executions, and their consumers are recorded
durably, so background work survives restarts without duplicate delivery.
Authorization cache invalidation is fixed, so membership and permission
changes take effect promptly. The server shuts down cleanly without
lingering cache workers
(@Optic00). Worklog
handling is more dependable: times and reports derive from an explicit
timezone, an item's time tab no longer truncates totals at fifty entries,
and editing only a worklog's description no longer disturbs its
timestamps.
Thank you
Thank you to Ben (@Optic00), Nikita
(@oshuej198), Josh
(...
Windshift v0.8.7
Windshift 0.8.7
Windshift 0.8.7 hardens Markdown and HTML rendering, strengthens custom-field
data integrity and presentation, tightens agent-skill and access boundaries,
and makes boards, navigation, forms, dashboard widgets, and integration APIs
more dependable. It also moves official builds and source development to Go
1.27.
Upgrade and compatibility
- Installations older than 0.8.5 must upgrade to 0.8.5 or 0.8.5-2 before
moving to 0.8.7. Back up the database, upgrade to one of those releases,
and allow Windshift to start successfully and record the 0.8.5 schema
checkpoint before installing 0.8.7. - Installations already running 0.8.5, 0.8.5-2, or 0.8.6 can upgrade directly.
This release requires the recorded 0.8.5 checkpoint; direct upgrades from
0.8.4 or earlier are not supported. - Building Windshift from source now requires Go 1.27.0. The module,
official container builders, and release preflight images have all moved to
Go 1.27.0. - External Markdown images are now restricted by default. Existing image
URLs remain stored, but browsers load only local attachments, embedded
images, Unsplash, Atlassian API, and configured Jira hosts. Administrators
who trust their content authors and need images from other HTTP or HTTPS
hosts can enable External Markdown Images under Admin → Security.
The policy change applies when users next load or refresh a page.
Custom fields and forms
- A broad custom-field reliability pass improves validation, cleanup, display,
portal usability, sorting, and query behavior across work items, assets, and
customer-facing surfaces. (WI-1038) - Portal and public forms now treat blank optional select and multiselect
values as unset, require title consistently when it appears on a form, and
avoid offering field types that the portal cannot populate. (WI-1039,
WI-1047, WI-1064) - Custom-field definitions and values are validated more consistently:
in-place type changes are rejected, number and date values are checked,
virtual number behavior is aligned across configuration and submission, and
select fields cannot be saved without options. (WI-1040, WI-1041, WI-1042,
WI-1044) - Board cards, collection lists, customer details, and portal asset reports now
render typed custom-field values more reliably. Multiselect cards keep every
selection, multi-user fields show names, corrupt numbers no longer display
asNaN, empty arrays use the not-set state, and reference fallbacks use
stable IDs. (WI-1043, WI-1048, WI-1049, WI-1050, WI-1060, WI-1061, WI-1062) - Custom-field sorting, filtering, and indexing now handle more edge cases:
ID-backed scalar fields avoid lexicographic ordering, multi-user fields are
not offered as sortable, the visual query builder can round-tripIS NULL
andIS NOT NULL, and SQLite text and date indexes match the expressions
used by queries. (WI-1051, WI-1057, WI-1059) - Asynchronous cleanup and index jobs now cover item, asset, and portal data,
recognize numeric-string option IDs, protect concurrent edits, avoid
duplicate index builds, and retry transient failures while surfacing
permanent ones. Linking-field mirror deletion follows the same in-use,
index-cleanup, and scrub safeguards. (WI-1052, WI-1053, WI-1054, WI-1055,
WI-1056, WI-1058) - Custom-field automations can set boolean and checkbox values and reject
deleted or unknown field IDs instead of persisting orphaned data. (WI-1045,
WI-1046) - Portal customer-list parsing failures are now logged instead of silently
hiding affected rows without diagnostic context. (WI-1063)
Content security and agent safeguards
- Work-item descriptions and comments now preserve their source Markdown for
editing and integrations while browser-facing HTML is produced by one shared
sanitizer across desktop, mobile, public-board, and portal views. Safe
Markdown formatting remains available; raw HTML is displayed as text, and
scripts, event handlers, embedded documents, and unsafe URL schemes are
removed or made inert. - Markdown links and images now follow the same URL policy in the editor,
server renderer, imports, REST responses, and public surfaces. HTTP and HTTPS
links remain supported, along with mail, telephone, page, anchor, relative,
and safe raster-image destinations where appropriate. Script-capable,
protocol-relative, control-character, and backslash-confused destinations
are rejected. - Administrators can explicitly allow images from arbitrary HTTP and HTTPS
hosts with the new External Markdown Images security setting. The default
content-security policy keeps the narrower local, embedded, Unsplash,
Atlassian, and configured Jira exceptions. - Required-SSO authentication now verifies access through the configured SSO
provider, and action user resolution excludes people outside the current
workspace. (WI-1067, WI-1070) - Action destinations are normalized and validated at the execution boundary,
closing URL parsing and hostname cases that could escape the configured
allowlist. (WI-1068) - Agent-skill activation now enforces an aggregate content budget, limits skill
access through review-stable run snapshots, keeps referenced pages ACL-safe,
and prevents skill metadata from changing prompt structure. (WI-718,
WI-719, WI-720, WI-722) - The agent grant form now reflows in narrow settings panels so its controls
remain readable and usable. (WI-1069)
Boards and everyday work
- REST API consumers can now select up to 100 anchor items with CQL or explicit
IDs and batch-load up to 50 direct visible links per item. Per-item
continuation and set-based loading keep dependency graph requests bounded
without an N+1 query pattern. Link expansion is one hop in this release.
(WI-1078) - Label assignments and item-link changes now refresh every affected item, so
incremental/items/changesconsumers stay synchronized and recent-activity
ordering reflects relationship updates. Label-definition edits also refresh
assigned items without treating catalog maintenance as user activity.
(WI-1077) - Collection boards can now include personal tasks, move them between the
personal Open and Done states through endpoint columns, and open the correct
personal-task detail experience. (WI-1073) - Browser back and forward navigation preserves the selected item detail, and
Daily Plan remains usable when no model provider is configured. (WI-1066,
WI-1072) - Form-channel creation uses one consolidated configuration and routing path
while preserving the expected channel result. (WI-1071) - My Tasks, Overdue Items, and Upcoming Deadlines now exclude completed work
using the authoritative completion state, including custom completed
categories such as Closed or Resolved. (WI-808, WI-813)
Maintenance
- Windshift now uses Go 1.27's standard-library UUID package and caps the
number of values accepted in HTTP request headers across the application,
Logbook, and CLI authentication callback servers. CI and lint tooling have
also been updated for Go 1.27 support. - The direct
filepath-securejoindependency is updated from 0.6.1 to 0.7.0;
upstream reports no API change. (WI-1075) - Go code has been updated for Go 1.26 lint behavior while preserving existing
product contracts. (WI-1074) - The README quick-start link now points to the self-hosting guide instead of a
missing documentation route. (WI-803)
Windshift v0.8.6
Windshift 0.8.6
Windshift 0.8.6 is a reliability release focused on consistent date and time
handling, dependable CQL-backed reporting, and safer item creation and listing.
Correct dates, trustworthy analytics, and more resilient work-item operations
Upgrade and compatibility
- Installations older than 0.8.5 must upgrade to 0.8.5 or 0.8.5-2 before
moving to 0.8.6. Back up the database, upgrade to one of those releases,
and allow Windshift to start successfully and record the 0.8.5 schema
checkpoint before installing 0.8.6. - Installations already running 0.8.5 or 0.8.5-2 can upgrade directly. This
release removes the legacy migration path and requires the recorded 0.8.5
checkpoint; direct upgrades from 0.8.4 or earlier are not supported.
Dates, time, and recurring work
- Windshift now applies one temporal contract across authenticated views,
public portal views, worklogs, recurring work, and scheduling. Instants use
the authenticated user's validated timezone, date-only values retain their
calendar date, and public surfaces use an explicit UTC default. Invalid
legacy timezone values safely fall back to UTC. (WI-999, WI-1006, WI-1007,
WI-1008, WI-1012) - REST v1 and MCP worklogs now use the acting user's timezone when none is
supplied. Worklog editing and timesheet grouping preserve the selected date,
and duration-only worklogs have the same start-time semantics across cookie,
REST, and MCP APIs. (WI-1004, WI-1005, WI-1013) - Recurring work validates requested IANA timezones and no longer accepts the
server-localLocalzone. Stored schedule timezones retain a safe fallback
for older data. (WI-1011) - Date filters cover complete civil days, including the final fractional
second returned by CQLendofday(). Additional date displays, worklog
ranges, scheduler boundaries, and server-clock comparisons now use the same
shared rules. (WI-1014, WI-1032)
CQL, analytics, and reporting
- Workspace statistics, milestone progress, and analytics now support CQL and
saved-collection filters that reference joined fields such as status,
priority, item type, iteration, project, completion state, workspace, and
user names. These filters return the correct dataset instead of zeroed
statistics or an HTTP 500 response. (WI-1026, WI-1027) - Workspace statistics now fail explicitly with
INTERNAL_ERRORwhen a core
aggregation cannot be computed, instead of returning a successful but
misleading empty or partial result. (WI-1031) - CQL item-type names are case-insensitive for equality, inequality, and
IN
filters. Thetypealias is retained and follows the same behavior as
itemtypename, including bare and quoted names; numeric item-type IDs remain
numeric comparisons. (WI-1028, WI-1030) - Concurrent asset CQL evaluation no longer mutates shared SQL-generator
state. (WI-1033)
Work-item reliability and performance
- Item creation now retries fractional-rank, workspace item-number, and
PostgreSQL serialization conflicts at the transaction boundary. Standard
creation, copying, recurrence generation, SCM sync, and template cloning use
the same collision-safe behavior. (WI-955) - Creating an item with an inactive or unknown assignee is rejected
consistently across the application API, REST v1, and MCP before anything is
persisted. (WI-1029) - Default-ranked item lists can use the workspace rank index efficiently on
SQLite, avoiding a scan whose cost grew with the total number of items on the
instance. (WI-954) - Item-list totals are invalidated only after committed count-changing
operations and are scoped to the affected workspaces. Concurrent readers no
longer inherit another request's cancellation, and workspace moves update
source and destination totals immediately. (WI-946, WI-956)
APIs and integrations
- Partial collection updates preserve an omitted description. Public
collection updates also validate slug requirements consistently and report
duplicate slugs as conflicts. (WI-1009, WI-1014) - GitHub status synchronization continues after the initiating HTTP request
completes, so updates made through item edit and item-type change paths reach
GitHub reliably. (WI-1010) - Jira readiness and analysis queries now escape project keys, shutdown
requests cannot block on an already-full channel, and remaining low-severity
temporal and scheduling issues identified during the release review have
been corrected. (WI-1014)