The floating chat called Espo.Ajax.postRequest with timeout:180000, which is
shorter than the agentic loop's worst-case latency. On 2026-05-13 a prod
request to generate a direct-access report sat 227s in the first ai-gateway
iteration (large skill+context payload), so the browser fired the .catch()
handler at 180s and rendered "שגיאה בתקשורת" — even though the backend
completed the work 47s later (report attached to case 42).
Changes:
- floating-chat.js: timeout 180000 → 600000 (matches the PHP curl timeout
already set to 600s in SmartAssistantService.php:521).
- After 60s the spinner label swaps to a "still working" hint so the user
knows long requests are normal, not stuck.
- i18n: new SlowHint label in en_US + fa_IR (the project uses fa_IR as the
Hebrew slot).
Refs Task Master #3
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Tasks with parentType='Contact' were invisible to AlertCalculator,
CaseContextBuilder, and OfficeContextBuilder — all queries only checked
parentType='Case'. This caused false "no preparation task" alerts in
the daily standup even when preparation tasks existed.
Broadens getSummary, findOverdueTasks, findUpcomingTasks, getWorkloadByUser
to query parentType=['Case','Contact']. Adds caseHasPreparation() helper
that checks Case tasks, Contact tasks, and NhActivity-linked tasks.
Adds getCaseContactIds() to CaseContextBuilder and OfficeContextBuilder.
Refs Task Master #1
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
chaim