docs: document 30 verified Highway 21 well card fault alarms

HANDOFF.md

@@ -1,5 +1,19 @@
 # Current handoff
 
+## Card fault alarms completed — October 26, 2026
+
+After the user imported the updated PLC XML, Water had 2,798 register definitions. The imported names use the correctly spelled suffix unlatched: GVL_Well[n]_out_flow_flt_card_unlatched and GVL_Well[n]_out_lvl_flt_card_unlatched, for n=0–14. Their OPC UA addresses follow Application.GVL.Well[n].out.{flow,lvl}.flt.card.unlatched on Water PLC.
+
+Enabled polling for all 30 well card-fault sources, preserving On Change and one-second intervals. Fresh table readback verified all 30 Enabled Yes; the settled snapshot showed 0 for 29 sources, with the most recently enabled level source still blank at capture. No PLC values were written. The separate cistern card-fault definition was outside this well-specific request and remained disabled.
+
+Created 30 enabled HMI alarms, bringing Water from 26 to 56 definitions. IDs CARD-00-FLOW/CARD-00-LEVEL through CARD-14-FLOW/CARD-14-LEVEL use the user's mapping above. Program10 is Well18 and program11 Well19. Programs12–14 are explicitly named Spare Program12–14. Descriptions follow “Well 5 Flow Input Card Fault” / “Well 5 Level Input Card Fault”. Messages are quoted expression strings, for example "Well 5: flow input card fault detected. Check the input module and its connection." Return text: "Well 5: flow input card fault has cleared."
+
+Each calculation directly references |{register:water:GVL_Well[n]_out_flow_flt_card_unlatched}| or the corresponding lvl source. Settings: Warning, Enabled true, HMI Latching false, one-second update interval, Delay On0, Delay Off0; notification groups Default and Atom Notifications, matching the existing sampled well alarm.
+
+Validation: reloaded the alarm page and reopened all 30 saved forms. Verified every ID, description, calculation, fault message, return message, enabled/latching state, interval, delays and notification groups. Filter CARD- returned exactly30 rows; total56. Existing26 descriptions/messages were retained, including the previously recorded Well18/19 discrepancy. No alarm conditions forced, reset/acknowledgement or notification-delivery test performed. Polling configuration and persisted alarm definitions are verified; physical card failure behavior is not tested.
+
+Local deliverables: highway-21-card-alarms.png and highway-21-card-alarm-audit.json in this chat's outputs directory.
+
 ## Highway 21 card fault source audit — October 26, 2026
 
 Read shared instructions and cloud index; no existing Highway 21 guide. Created HIGHWAY-21-GUIDE.md and cloud guide https://chatgpt.com/space/page_068c2a277110819180a2c5853c714e99, indexed both. User mapping recorded, including index10 Well18 and index11 Well19. Water catalog2,766 registers; card search zero; unlatched170; fault60 VFD entries. Exact inspected flow fault source Application.GVL.Well[0].out.flow.flt.chan.unlatched is disabled, On Change1s. Water has26 alarms. Existing index10/11 low-flow/low-level descriptions swap Well19/18 relative to user mapping; retained. User confirmed new PLC tags require importing. Native Water PLC import requires OPC UA XML; no tag browser. Awaiting export XML or full exact paths and equipment/card assignments. Delete Unused defaults ON; turn OFF for additive work. No tags or alarms added. No site configuration or PLC values saved. Next: resolve exact sources, add/verify alarms, update guides and commit completion.
@@ -172,4 +186,4 @@ Use Atom GIT to open Auto/atom-hmi-ui-chatgpt on master. Read AGENTS.md, PROJECT
 
 ### Alarm history lookup — October 21, 2026
 
-User requested checking whether any Bin G alarm previously occurred. Opened the alarm icon → Alarm History at https://www.atomhmi.com/site_sweetgrass/alarm_history. Requested Last 30 Days (September 22–October 21, 2026), then freshly reloaded and requested October 21, and finally Yesterday (October 20). Every attempt remained on Loading... without a history table, records, or an explicit empty-result message. No browser warning/error logs were captured. Occurrence, timestamps and retention therefore remain unverified; this is not evidence of no prior Bin G alarms. No acknowledgement, reset, alarm edits, register writes or page saves performed. User-confirmed update on October 21, 2026: Atom HMI was just updated and alarm logs moved to a new database; previous history is no longer available. This supersedes the earlier plan to retry for pre-migration records. Historical Bin G occurrences cannot be determined from the current database. The migration explains unavailable prior history, but the cause of the observed Loading... UI was not independently established. The intended Bin G enable/disable alarm path remains unresolved; determine it from current configuration or user/PLC confirmation.
+User requested checking whether any Bin G alarm previously occurred. Opened the alarm icon → Alarm History at https://www.atomhmi.com/site_sweetgrass/alarm_history. Requested Last 30 Days (September 22–October 21, 2026), then freshly reloaded and requested October 21, and finally Yesterday (October 20). Every attempt remained on Loading... without a history table, records, or an explicit empty-result message. No browser warning/error logs were captured. Occurrence, timestamps and retention therefore remain unverified; this is not evidence of no prior Bin G alarms. No acknowledgement, reset, alarm edits, register writes or page saves performed. User-confirmed update on October 21, 2026: Atom HMI was just updated and alarm logs moved to a new database; previous history is no longer available. This supersedes the earlier plan to retry for pre-migration records. Historical Bin G occurrences cannot be determined from the current database. The migration explains unavailable prior history, but the cause of the observed Loading... UI was not independently established. The intended Bin G enable/disable alarm path remains unresolved; determine it from current configuration or user/PLC confirmation.