Add library job tracking and status monitoring by AlexKempen · Pull Request #37 · frcdesign/FRCDesignApp

AlexKempen · 2026-08-08T14:40:55Z

Summary

Adds a new library jobs system to track background workflow executions (document reloads and group additions) with real-time status monitoring. Users with editor access can now view job history and outcomes via a new /app/library-jobs page, and the UI automatically refreshes when jobs complete.

Key Changes

Database & Schema

Added library_jobs table to track workflow runs with status (running/complete/partial/errored), timestamps, and error messages
Added lastLoadedAt column to group and insertables tables to record when entities were last successfully loaded
Created Drizzle migration for the new schema

Backend Workflow Integration

Modified LoadLibraryWorkflow to accept a libraryJobId parameter and report job completion status
Implemented finishLibraryJob() to mark jobs as finished with their outcome (complete/partial/errored)
Added deriveLoadStatus() to compute overall job status from per-group results
Wrapped workflow execution in try-catch to handle and report errors

New Backend Services

library-jobs.ts: Core job management (create, finish, list, reconcile)
access-level-cache.ts: KV-backed caching of access levels to reduce Onshape API calls during polling
GET /api/library-jobs/library/:libraryId: Endpoint to fetch recent jobs with editor-only access
library-jobs.test.ts & build-status.test.ts: Test coverage for new endpoints

Frontend UI

New /app/library-jobs route displaying a table of recent jobs with status, type, trigger, start time, and duration
useLibraryJobWatcher hook that polls jobs and auto-refreshes library data when jobs finish
Added "Library Jobs" button in navbar (editor-only) with running job indicator
Updated build status cards to show last-loaded timestamps
New format-time.ts utilities for relative time and duration formatting
Integrated job watcher into main app route

Route Generation

Updated routeTree.gen.ts to include the new library-jobs route with proper formatting

Implementation Details

Jobs are created when reload/add-group workflows start and marked with their Cloudflare instance ID for status reconciliation
A safety net reconciles jobs still marked "running" in D1 by querying live Cloudflare instance status
Frontend polling uses adaptive intervals: fast (3s) for fresh jobs, slower (30s) for older ones, to balance responsiveness with cost
Access levels are cached in KV for 1 hour to avoid per-request Onshape API calls during aggressive polling
Job outcomes (complete/partial/errored) trigger appropriate toast notifications and library refresh

https://claude.ai/code/session_01LaqfcR5s9LGTXj6AKof7rh

Surface running workflow state back to admins. Cloudflare Workflows (library reload / add-document) were fire-and-forget with no visibility. Backend: - New `library_jobs` D1 table (+ migration) recording each workflow run: type, status, label, triggeredBy, per-group result, timestamps. - Trigger routes create a `running` row and pass a jobId into the workflow; the workflow records its own completion. - Workflows now propagate per-group error messages instead of swallowing them, and derive a `partial` status when some groups fail but others succeed. - Editor-gated `GET /api/library-jobs/library/:libraryId` lists a library's jobs and reconciles stale `running` rows against the live Cloudflare instance status. - Access level is cached in KV (~1h) so frequent polling doesn't hit Onshape on every request. Frontend: - `useLibraryJobsQuery` polls while a job runs (aggressive then backing off), only for editors. - A watcher auto-refreshes the library and toasts the outcome when a job finishes; a navbar spinner shows while one is running. - New `/app/library-jobs` page lists recent jobs with status, duration, and per-group failure detail, linked from admin settings. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01LaqfcR5s9LGTXj6AKof7rh

Follow-up on review feedback. The first cut rendered per-group load results on the jobs page, which duplicated the existing build-status system (buildIssues + LOAD_FAILED/INSERTABLES_FAILED, shown inline per entity). This moves the per-entity signal onto the entities and slims the run record. - Add lastLoadedAt (time of last successful load) to groups and insertables; stamp it in saveGroup/saveInsertable. Failures leave the last-good time untouched and stay expressed by build checks, so there's no separate per-entity load-status column. - Surface it in the build-status hover card as a dedicated "Loaded 3h ago" header pinned to the top-right (not another row); "Never loaded" when null. - Drop the per-group result column from library_jobs; it's now a slim run-log for the running indicator, auto-refresh, and a compact recent-reloads list (status/label/who/when/duration). The workflow still derives overall complete/partial/errored status. - Regenerate a single clean 0001 migration (creates library_jobs without result, adds last_loaded_at to groups/insertables). Verified additive against origin/cert, which only has 0000. - Extract formatRelativeTime/formatDuration into a shared helper; add load-status and build-status test coverage. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01LaqfcR5s9LGTXj6AKof7rh

Remove the library_jobs table, its routes/helpers/DTO, the jobs page, navbar indicator, watcher, and workflow job-recording — the workflows go back to fire-and-forget. What's kept is the useful part: each group and insertable stores its last successful load time (lastLoadedAt), surfaced in the build-status hover card. To keep the UI from spinning forever, add a small bounded poll: after a reload or add-document is triggered, watch the library's cacheVersion (bumped by the workflow when it finishes) and refresh the library view once it changes, then stop (with a hard time cap). No DB or job records involved. The KV access-level cache stays (it's a general per-request win and keeps the cacheVersion poll cheap). Migration regenerated to a single additive 0001 that only adds last_loaded_at to groups and insertables; verified against origin/cert, which has only 0000. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01LaqfcR5s9LGTXj6AKof7rh

Refactor the editor-only build-status hover card so the admin flags are editable in place, instead of living as separate toggle items in the right-click "Admin options" submenu. - The hover card now shows: last-loaded time, Build checks, an Admin section of Switches, and (insertables) a read-only Parsed section. - Insertable switches: Visible to users, Insert and fasten, Part number search, and Open composite (part studios only). Group switch: Sort alphabetically. Each is disabled while its mutation is in flight; the hide toggle keeps its favorites-removal confirm. - Parsed (read-only) shows Vendors and Configurable — Configurable stays automatic, not a toggle. - Bumped the HoverCard closeDelay so the card can be moved into and its switches clicked. - Extracted the toggle mutations into reusable hooks in card-hooks.ts. - The admin context menu keeps the commands: reload thumbnail (both), and for groups show/hide all, reorder, delete, add group. The per-flag toggle menu items were removed since they're switches now. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01LaqfcR5s9LGTXj6AKof7rh

Mantine's notifications.show is a no-op when a toast with the same id is already displayed (notifications.store: returns early on duplicate id), so the "loading spinner → replace with same id" pattern used by add-group, insert, and the error handlers left the loading toast stuck open forever — e.g. the reload/add "Adding document…" toast never closed. Route the show* helpers through a showToast that first hides any existing toast with the given id, so a loading toast is reliably replaced by its final info/success/error state. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01LaqfcR5s9LGTXj6AKof7rh

The admin switches felt frozen: toggling bumped the library cacheVersion, which changed the build-status query key, blanking the hover card (data went undefined, the badge returned null and the card closed) while the switch waited on a sometimes-slow Onshape request. - Toggle mutations now optimistically flip the build-status cache in onMutate (mirroring useSetGroupOrderMutation), show a loading→success/ error toast, and revert on failure — so switches respond instantly. Factored into a shared useOptimisticToggleMutation; visibility gets the same treatment. - getBuildStatusQuery uses keepPreviousData so the card no longer blanks out and closes while the next cacheVersion refetches. - Split the badge into BuildStatusCard (hover content wrapping the menu) and BuildStatusBadge, which takes a semantic `hoverMenu` prop instead of children. - Dropped the disabled-while-pending state on switches (optimistic updates make it unnecessary). Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01LaqfcR5s9LGTXj6AKof7rh

Rework the admin hover card around a proper header and clearer sections: - Header: the group/insertable name in bold, a severity summary of pill badges (e.g. "1 error", "1 warning", or a green "All checks pass"), and the last-loaded time on the right. - Build checks: each issue rendered as a tinted callout box in its severity color; the section is hidden entirely when everything passes. - Admin: each toggle now has a short description under its label. - Parsed: unchanged (Vendors, Configurable), under an uppercase header. Threads the entity name into InsertableStatusBadge/GroupStatusBadge (no thumbnail, per the earlier decision). Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01LaqfcR5s9LGTXj6AKof7rh

The open-composite and sort-order toggle toasts reported the opposite action (turning "open composite" on said "Removed open composite"). Both were <void> mutations whose messages branched on the captured current value, which the optimistic update flips before onSuccess fires. Convert them to <boolean> mutations that receive the target value, matching insert-and-fasten and part-number-search. Also: "info" instead of "note" for the info count badge, nudge the issue callout icon down to align with the first text line, drop trailing periods from build-check messages, and rename the header time label to "changed". Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01LaqfcR5s9LGTXj6AKof7rh

- Move CLAUDE.md to AGENTS.md and add a comment-brevity guideline; CLAUDE.md now just imports AGENTS.md. - access-level-cache: drop the defensive KV try/catch and shorten the doc comments to a single sentence each. - services/load-insertable/workflows: trim or delete narrating comments. - build-status: carry elementType on InsertableBuildStatus so the badge reads it from the query instead of a threaded prop; move getIssueMessage into build-checker as getIssueDescription so build-issue definitions live in one file. - card-hooks: replace the monolithic useOptimisticToggleMutation config with a small patchBuildStatus helper and explicit per-toggle hooks; drop the loading toast for the trivial (DB-only) open-composite and sort-order toggles; standardize error messages on "Unexpectedly failed to update X". - theme: set cursorType "pointer" globally instead of a per-switch style. - reload-refresh: take the baseline cacheVersion from the caller's loader data instead of reading the query cache; tighten comments. - build-status.test: name the seeded last-loaded constants. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01LaqfcR5s9LGTXj6AKof7rh

- Add a small p-limit (createLimiter, LOAD_CONCURRENCY=15) threaded through LoadContext so a whole-library reload loads at most 15 insertables at once. This keeps Onshape pressure down; because each insertable's save is its own durable step, a rate-limit burst only affects the in-flight few and earlier work is preserved. - Track the active reload workflow instance per library in KV (reload-lock). /reload-groups now skips creating a second reload while one is live and returns { status: "already-running" }; add-group jobs stay unrestricted. The reload button shows an info toast instead of starting a second poll. KV has no CAS, so a sub-second double-trigger can still race — acceptable for an admin-only action. - Raise limits.subrequests to 200,000 in the cert and production env blocks so a large reload doesn't hit the per-invocation subrequest cap. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01LaqfcR5s9LGTXj6AKof7rh

Hiding pops a confirm modal, but Mantine's HoverCard only closes on mouse-leave — which never fires when the modal overlay merely covers the dropdown, so the card was left stranded behind the modal. Expose a close function via context (implemented by remounting the uncontrolled HoverCard) and call it from the visibility switch before the hide modal opens. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01LaqfcR5s9LGTXj6AKof7rh

Editors can trigger a reload that runs for minutes with no persistent signal. Expose the per-library reload state (isReloadRunning) via a polled GET /job-status endpoint and show a spinner while it runs — in the navbar (editors only) and in the build-status hover card, where it replaces the last-modified line. Both are tooltipped "A job is running". The reload trigger and completion invalidate the job-status query so the spinner appears/clears promptly rather than waiting for the 10s poll. Also rename the card's "Changed" label to "Last modified" and give it a clock icon. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01LaqfcR5s9LGTXj6AKof7rh

… into claude/workflow-state-ui-qme9hy

- Align the header's "Last modified" line with the title by putting them on a shared centered row (badges move below). - Add a tooltip clarifying "Last modified" is the last time changes were pulled from Onshape (distinct from a reload merely checking the doc). - Size the navbar spinner to match the other navbar icons (18px). - Turn each row's build-status severity icon into a spinner while a job runs. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01LaqfcR5s9LGTXj6AKof7rh

Replace the frontend reload-refresh polling globals with a hook that refreshes the library when the polled job status flips from running to idle (useRefreshLibraryOnJobFinish, mounted once in the editor-gated navbar indicator). Removes the redundant reload loading/success toasts — the spinner now signals a running load. Generalize the reload-lock into a job tracker (renamed job-tracker.ts) that records both reload and add-group workflow instances per library, so /job-status reports any running load job (isAnyJobRunning) and add-group keeps auto-refreshing when it finishes. Reloads stay a singleton via isReloadRunning. Bump the tracked-job KV TTL to 24h since a reload can run for hours. Also: reword the spinner tooltip to say the library is loading from Onshape in the background; tidy the concurrency limiter (no p-limit jargon, clearer queue drain); trim the reload-singleton comment. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01LaqfcR5s9LGTXj6AKof7rh

Each workflow now removes its own tracked-job entry in a final untrack-job step (it knows event.instanceId), so the running-job state clears promptly instead of lingering until the TTL. Only its own entry is touched, so concurrent jobs are unaffected; the status-check in activeJobs stays as the backstop for a job that crashes before untracking, and the 24h TTL is now purely that crash safety net. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01LaqfcR5s9LGTXj6AKof7rh

Show a transient "started" toast when a reload or add-document trigger call returns, and a success toast when the job-status poll sees the job finish (alongside the library refresh). No persistent running toast — the spinner covers that. Drop the now-unused hideToast helper. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01LaqfcR5s9LGTXj6AKof7rh

The "refetch context-data then re-run the loaders" dance was duplicated across the job-finish watcher and the card admin mutations, and the "invalidate favorites then re-run loaders" dance across the three favorites mutations. Pull them into useRefreshLibrary / useRefreshFavorites and use them everywhere those patterns appeared. useRefreshLibrary also drops the now-redundant explicit library invalidate: every library-mutating endpoint bumps cacheVersion, which re-keys the version-keyed library/search/build-status queries, so refetching context + re-running loaders is enough to pull fresh data. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01LaqfcR5s9LGTXj6AKof7rh

Group delete cascade-removes insertables and their favorites, and reorder bumps the version, so both now use useRefreshLibrary (which pulls the new version's library/search/build-status and re-prefetches favorites) instead of a bare library invalidate. Reorder keeps its optimistic update and its error-path rollback. Also extract useUpdateContextData (optimistic context patch + loader re-run) and use it for the settings save and the access-level switch. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01LaqfcR5s9LGTXj6AKof7rh

Extract a refetchFavorites helper and call it directly from useRefreshLibrary (and useRefreshFavorites) so a library refresh explicitly refetches favorites — a group delete or element hide cascades into them — rather than relying on the /app loader's incidental favorites prefetch. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01LaqfcR5s9LGTXj6AKof7rh

Reordering bumps the library version, so onSettled's useRefreshLibrary reconciles the optimistic order on success. Remove the special-case error invalidate (and the now-unused libraryQueryMatchKey import). Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01LaqfcR5s9LGTXj6AKof7rh

Mirror the card-toggle pattern: capture the previous library data in onMutate (and cancel in-flight fetches) and restore it in onError, so a failed reorder reverts immediately instead of relying on a version bump that never happens on error. onSettled still refreshes on success. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01LaqfcR5s9LGTXj6AKof7rh

- Fold the KV access-level cache into access-level-utils.ts (with the rest of the access-level code); delete access-level-cache.ts. - Rename load/concurrency.ts -> load/load-common.ts. - Collapse the job-status query + refresh-on-finish into a single useJobStatus hook in refresh.ts; delete job-refresh.ts and drop the completion toast (we don't reflect the outcome). The navbar just calls useJobStatus(). - Drop the manual optimistic-undo in onError across the card mutations and the group reorder; useRefreshLibrary now invalidates the library and build-status queries so a failed mutation rolls back via refetch (these routes aren't CDN-cached, so a same-version refetch returns fresh data). onError is now just the error handler; the onSettled var is named refreshLibrary. - build-status: CountBadge takes a severity and maps to color/noun itself; drop the uppercase section headers for normal capitalization. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01LaqfcR5s9LGTXj6AKof7rh

AlexKempen

Left comments

AlexKempen · 2026-08-10T04:16:33Z

+}
+
+/**
+ * Refreshes the whole library view: refetch the context (its cacheVersion keys


Oh hell naw this comment should be two lines max

AlexKempen · 2026-08-10T04:17:14Z

+}
+
+/**
+ * Polls whether a load job is running and refreshes the library once it


Just the first sentence

AlexKempen · 2026-08-10T04:19:33Z

+export function useJobStatus(): boolean {
+    const refreshLibrary = useRefreshLibrary();
+    const running = useJobStatusQuery().data?.running ?? false;
+    const wasRunning = useRef(running);


Let's use a useState instead of a ref for this I think.

If there's a reason we need a ref specifically, add a comment inline explaining why

AlexKempen · 2026-08-10T04:21:08Z

+) => void {
+    const router = useRouter();
+    return useCallback(
+        (recipe: (data: ContextData) => void) => {


Make a custom type for the callback function.

Give it a better name than recipe, like updateFunction or something.

AlexKempen · 2026-08-10T04:21:29Z

+    const jobRunning = useJobStatus();
+    if (!jobRunning) return null;
+    return (
+        <Tooltip label="The library is being loaded from Onshape in the background">


Add withArrow

AlexKempen · 2026-08-10T04:24:20Z

+}
+
+/**
+ * Shows a toast. When an `id` is given, any existing toast with that id is


REally clean this up, move the comment about existing toasts to the if statement in-line

AlexKempen · 2026-08-10T04:28:03Z

-                "add-group"
-            );
+        onSuccess: () => {
+            // Replace the loading toast with a transient "started" toast; the


Delete this comment

AlexKempen · 2026-08-10T04:28:55Z

+            // Raise the per-invocation subrequest cap so a full library reload
+            // (many Onshape calls per insertable) doesn't hit the default limit.
+            "limits": {
+                "subrequests": 200000


Actually make this 500

AlexKempen · 2026-08-10T04:29:01Z

                "ADMIN_TEAM": "6a62e6efcc21741bea57362c",
                "NODE_ENV": "development"
+            },
+            // Raise the per-invocation subrequest cap so a full library reload


Delete this comment

AlexKempen · 2026-08-10T04:29:07Z

                "ADMIN_TEAM": "5b620150b2190f0fca90ec10",
                "NODE_ENV": "production"
+            },
+            // Raise the per-invocation subrequest cap so a full library reload


Delete this comment

- refresh.ts: shorten the useRefreshLibrary/useJobStatus doc comments; keep the finished-transition ref (set-state-in-effect is disallowed) with an inline note; give useUpdateContextData a named ContextDataUpdate callback type. - navbar: add withArrow to the running-job tooltip. - notifications: slim showToast's doc and move the Mantine no-op note inline. - card-hooks/add-group: drop now-superfluous comments. - wrangler: drop the subrequest-limit comments and raise it to 500000. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01LaqfcR5s9LGTXj6AKof7rh

…frcdesign/FRCDesignApp into claude/workflow-state-ui-qme9hy

PR #37 added `limits: { subrequests: 500000 }` to the cert and production environments. Wrangler limits are only supported on the Standard Usage Model; on this Free-plan account the deploy is rejected at the versions endpoint with Cloudflare error 100328 ("CPU limits are not supported for the Free plan"), so the post-merge cert deploy failed even though CI (type-check/lint/test) passed. Comment out both limits blocks to restore deploys, keeping them documented so they can be re-enabled after moving to a paid plan. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01BSxhQes1omyhxKXe9uHY4K

…#39) PR #37 added `limits: { subrequests: 500000 }` to the cert and production environments. Wrangler limits are only supported on the Standard Usage Model; on this Free-plan account the deploy is rejected at the versions endpoint with Cloudflare error 100328 ("CPU limits are not supported for the Free plan"), so the post-merge cert deploy failed even though CI (type-check/lint/test) passed. Comment out both limits blocks to restore deploys, keeping them documented so they can be re-enabled after moving to a paid plan. Claude-Session: https://claude.ai/code/session_01BSxhQes1omyhxKXe9uHY4K Co-authored-by: Claude <noreply@anthropic.com>

claude and others added 30 commits August 6, 2026 05:38

Fix event propagation

9976901

Minor refactor

4f88544

Minor UI tweaks

ac3d9fe

Rename to build-issues

04f6549

Tweak descriptions

c892bac

Merge remote-tracking branch 'origin/claude/workflow-state-ui-qme9hy'…

3b252d9

… into claude/workflow-state-ui-qme9hy

Tweak error message

b4f150c

Fixup tooltip

86c5a2c

Combined load-common and load-context

d2b29a7

AlexKempen commented Aug 10, 2026

View reviewed changes

claude and others added 7 commits August 10, 2026 04:39

Remove sort toast

2a1f7c8

Merge branch 'claude/workflow-state-ui-qme9hy' of https://github.com/…

b0a971e

…frcdesign/FRCDesignApp into claude/workflow-state-ui-qme9hy

Create patchQuery util

7402a35

Fix hyphen

48fede6

Tweak build issues

7338591

Refactor mutation

e54339c

AlexKempen merged commit 2d05e98 into cert Aug 10, 2026
1 check passed

AlexKempen deleted the claude/workflow-state-ui-qme9hy branch August 10, 2026 05:03

AlexKempen mentioned this pull request Aug 10, 2026

Comment out unsupported wrangler limits blocks that broke cert deploy #39

Merged

Navigation Menu

Uh oh!

Add library job tracking and status monitoring - #37

Add library job tracking and status monitoring#37
AlexKempen merged 38 commits into
certfrom
claude/workflow-state-ui-qme9hy

AlexKempen commented Aug 8, 2026

Uh oh!

AlexKempen left a comment

Uh oh!

AlexKempen Aug 10, 2026

Uh oh!

AlexKempen Aug 10, 2026

Uh oh!

AlexKempen Aug 10, 2026

Uh oh!

AlexKempen Aug 10, 2026

Uh oh!

AlexKempen Aug 10, 2026

Uh oh!

AlexKempen Aug 10, 2026

Uh oh!

AlexKempen Aug 10, 2026

Uh oh!

AlexKempen Aug 10, 2026

Uh oh!

AlexKempen Aug 10, 2026

Uh oh!

AlexKempen Aug 10, 2026

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

Uh oh!

Conversation

AlexKempen commented Aug 8, 2026

Summary

Key Changes

Implementation Details

Uh oh!

AlexKempen left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants