Skip to main content

Alert for KB uploads stuck in Processing beyond 5 hours

Context​

Follow-up to the Fast Track chatbot-ai recurring-issues report (2026-07-27-fast-track-chatbot-ai-recurring-issues.md), specifically its Cluster C — KB ingestion lifecycle finding "Stuck Uploading Knowledge (Processing)".

Notes​

  • Cluster C covers ~9 Fast Track tickets; the "stuck in Processing forever" symptom is evidenced by BOT-4746 (AI Resource Upload Stuck in Processing, In Progress) and BOT-3635 (AI Source PDF Stuck in Processing, cannot Resync/Delete).
  • Per the report, V1 AiKnowledgeSource has no worker, no timeout, and no terminal-failure state; status refreshes lazily only when the list is opened (get.rb:33 → update_status.rb), and unknown AI-service statuses are swallowed by the rescue in update_status.rb, so a row silently stays IN_PROGRESS.
  • The report's own recommendation (a scheduled poller with a poll budget forcing IN_PROGRESS → FAILED, mirroring the V2 path, plus FE polling with a client-side give-up) is context only — it is a proposal from the report, not what was decided here.

Decisions / changes​

  • Build an alert that detects a knowledge-upload process stuck in Processing for more than 5 hours. Decided 2026-07-27; owner Eko Aprianto (BE).
  • Tracked as bucket story A6 in Chatbot 26Q3 Engineering Initiatives (Epic BOT-4620) as Jira Story BOT-4796, Class 🟡 Standard, DRI Eko Aprianto, pulled into Chatbot 26Q3 Sprint 2 (2026-07-27 → 2026-08-10).

Follow-ups​

  • Confirm the A6 per-story QA lane (bucket default is Lane B) — TBD.