The Elastic Beanstalk nginx proxy kept its 1 MB default body limit, so every
media upload over ~1 MB got an nginx 413 before reaching the API. Ship a
.platform nginx override (120M) in the bundle and assert it in the bundle
contract.
Apply the client-confirmed contract: photos up to 10 MB (JPEG/PNG/HEIC),
videos up to 100 MB (MP4/MOV), at most 10 photos and 3 videos per work order,
with stable generic rejection messages. The request ceiling (110 MB) sits
between the per-kind caps and the proxy so oversize files get the generic
message. The vendor portal accepts the same photo/video types and caps.
Editing a work order's POC never reached the backend: no update path wrote
PocName/PocPhone/PocNotes, so the optimistic UI edit was lost on refetch and
the completion freeze captured the Site contact instead of the manual value,
and nothing was audited.
- Add tenant-scoped PATCH api/workorders/{id}/poc via new WorkOrderPocService
+ WorkOrderPocDataService: persists the override, stages FieldChanged audit
entries (which also write field locks so sync never overwrites a manual POC),
and enforces row-version concurrency and terminal-status read-only rules.
- Lock semantics (SH-190): a manual POC away from the Site's live contacts is
stored WO-level; an edit equal to a live Site contact (or blanking name+phone)
stores nothing so the WO follows the Site. PocCustomized exposes the state.
- Board projection, completion freeze and create now share one Site-contact
fallback (first non-deleted contact by SiteContactOrder) so a never-overridden
WO keeps following the Site, including at create when the wizard prefills it.
- Route contract baseline gains PATCH {id:int}/poc.
Adds the personal producers behind GET /api/notifications without changing its shape:
new assignments (SH-288), unanswered comments on work the user takes part in (SH-289),
@mentions, and decisions on uplifts the user requested (SH-215).
No Vendor treated any non-terminal dispatch as an assigned vendor without
checking IsDeleted, so a soft-deleted dispatch hid the work order from both
No Vendor and Vendor Conflict (the conflict query already drops deleted
dispatches). GetNoVendorAsync and the vendor-reminder assigned-work-order rule
now skip soft-deleted dispatches, keeping the asserted parity between the feed
and the reminders.
Vendor Conflict applied the section cap in vendor-sweep order, so when overlaps
exceeded the limit newer conflicts could be dropped while Count still counted
every work order. VendorConflictsAsync now orders pairs by recency before the
cap, matching the other per-work-order sections.
The uplift detail modal needs the work order's assigned dispatcher, the
requesting vendor's technician and the scheduled date. They now resolve
from the same effective work order and vendor as the existing queue row,
so the modal no longer depends on a separate work-order fetch that
account-scoped staff cannot read.
Adds GET /api/notifications, a per-user read model derived from live
work-order state: Unassigned (grouped, High), No Vendor and Aveta Missing
(per work order, Medium) and Vendor Conflict, account-scoped from claims
and ordered by section severity with a fixed reason tie-break.
#144 branched from the SH-210 decision-actions commit before the SH-207/SH-208
read-contract corrections landed, so it re-implemented workOrderClosed,
attachmentCount, decidedByName, and pendingExposureTotal with stale semantics:
it projected WorkerOrderNumber (the CRM external id) instead of InternalWONumber
(what the board renders) and summed raw RequestedNTE, which double-counts the new
NTE total on vendor-portal rows across sequential approvals.
Merge origin/feat/ab/sh-210-uplift-decisions (#141, what lands) in and resolve
every conflict in #141's favour, so those fields and their granted-amount
exposure math now come from #141 rather than being duplicated here. Keep only the
two deltas #144 actually adds on top of #141:
- attachmentCount counts non-deleted UpliftEvidence documents on the dispatch,
not every completion document, so completion photos no longer inflate the chip;
- the queue read resolves the effective work order via the primary-plus-linked
(DispatchWorkOrders) convention the sibling reads use, so a dispatch linked only
through that table surfaces its WO context, closed flag, and exposure. Covered
by an in-memory test and a SQLite relational test that proves the fallback
translates to SQL.
Drop the superseded attachment-count test that asserted completion documents
count, and align the relational test to seed InternalWONumber.
GetPagedAsync resolved WorkOrderNumber/Site/Service, WorkOrderClosed, and
WorkOrderId only through Dispatch.WorkOrderId, so a dispatch linked to its
work order solely through DispatchWorkOrders surfaced blank WO context,
workOrderClosed:false, and zero exposure. Resolve the effective work order
using the same primary-plus-linked convention as GetWorkOrderIdForUpliftAsync
and GetApprovedExposureForWorkOrdersAsync via a single work-order lookup.
The approvals queue frontend needs four list-contract additions the read
contract PRs do not carry yet: workOrderClosed so the Approved tab can
disable Revoke on terminal work orders (mirroring the SH-196 revoke
guard), attachmentCount from non-deleted UpliftEvidence documents so the
+N chip renders, decidedByName for the Approved By column, and the
queue-wide pendingExposureTotal for the header total.
DeleteUserWithCascadeAsync never removed UserPermissionOverrides rows, and
the FK on UserId is Restrict. Once an override was saved for a member, the
admin hard-delete (DeleteUserAsync -> DeleteUserWithCascadeAsync) failed the
foreign key inside the transaction and the controller surfaced it as a 400,
so the user was never deleted. Add an explicit ExecuteDelete on
UserPermissionOverrides before the user is removed, matching how UserRoles is
already cleared in the same method (no schema change).
SetOverrideAsync was check-then-insert on the composite key with no
DbUpdateException handling, so two concurrent PUTs for the same
(UserId, PermissionKey) let the loser violate PK_UserPermissionOverrides and
return 500. Catch the conflict, detach the pending insert, and converge by
updating the persisted row to the caller's requested state.
