Bomet runs stock pgr-services-dev — no domain-event producer, real complaints don't notify

[KDwevedi]

Summary

On Bomet (bometfeedbackhub.digit.org) a real citizen complaint does not trigger the Novu/Twilio lifecycle notification, because the deployed PGR image lacks the CCRS domain-event producer.

The notification stack itself is proven working end-to-end (see Evidence) — this is purely a PGR build/deploy gap.

Evidence (2026-05-16)

• Citizen UI "File a Complaint" correctly calls POST /pgr-services/v2/request/_create → 200 (complaint PG-PGR-2026-05-16-048405).
• Deployed container digit-pgr-services-1 = image registry.preview.egov.theflywheel.in/pgr-services-dev:latest.
• That image publishes only to the legacy save-pgr-request topic (and its NotificationService errors there). It does not emit COMPLAINTS.WORKFLOW.* to complaints.domain.events — kafka high-watermark unchanged after the UI submit; bridge logs show no dispatch for the real complaint.
• A synthetic COMPLAINTS.WORKFLOW.APPLY event published directly to complaints.domain.events delivered a real WhatsApp (Twilio status delivered) via the approved Content template — so bridge → Novu → Twilio + the seeded bindings/provider are all good.

Root cause

backend/pgr-services in this repo contains ComplaintDomainEventService (publishes the COMPLAINTS.WORKFLOW.<ACTION> envelope the novu-bridge consumes). Bomet runs the stock pgr-services-dev image, which does not include it.

Required action

Build and deploy the CCRS pgr-services image (the fork in this repo, with ComplaintDomainEventService) to Bomet — same upstream-source → rebuilt-image → redeploy pattern; do not patch in place. After deploy, verify a real UI complaint produces an offset on complaints.domain.events and the bridge dispatches it.

Related

• Notif seed / bindings: PR fix(notif-seed): align seed.sh + schemas with live config-service behavior (corrections for #37) #41 (lifecycle contentSids + upsert seed.sh)
• Bridge channel-aware prefix: PR fix(novu-bridge): channel-aware whatsapp: prefix — unblock SMS delivery #43
• Runtime env/nginx: PR fix(notif): runtime env defaults + nginx workarounds from bomet bring-up #42
• otp-publisher: PR feat(notif): otp-publisher service — real OTPs over Novu + Twilio (implements #38 approach A) #39

This is a deployment-tracking issue, not a code change in the notif PRs.

[KDwevedi]

Deployed + verified (2026-05-16) — not closing per dev/QA norm

CCRS pgr-services is now live on Bomet, independent of PR merges.

• Built pgr-services:ccrs-nairobi on-host from origin/nairobi (b4b0d99) via local-setup/docker/pgr-services/Dockerfile; swapped /opt/digit/docker-compose.egov-digit.yaml (line 1387) and recreated only pgr-services. Rollback anchored (old image pgr-services-dev:latest + compose .bak-preccrspgr).
• Real UI complaint now emits COMPLAINTS.WORKFLOW.APPLY to complaints.domain.events (Kafka HWM moved; Producer ... topic : complaints.domain.events).
• Full chain verified: bridge resolves → Novu → Twilio WhatsApp delivered to a real handset via approved Content template HX67fae….

Two real bugs surfaced & fixed in PR #41

1. Tenant scoping: config-service resolves TemplateBinding/ProviderDetail by exact tenantId, no hierarchy fallback. The citizen UI files complaints under the root tenant ke, not the city tenant ke.bomet. Bindings seeded only at ke.bomet → every real complaint failed CONFIG_NOT_RESOLVED. Re-seeded at ke; README now documents "seed at the event tenant".
2. provider-details.json had channel=sms while the bridge dispatches whatsapp → provider never resolved (NB_PROVIDERS_NOT_FOUND). Fixed to whatsapp.

seed.sh is now a true upsert (PR #41) so live tenants can be corrected, not just initially created.

Leaving this issue open for the owner to close per the dev→To-Review / QA-owns-Done convention.