SYLVA Labs · AI Pulse

Operations & support guide

How AI Pulse stays healthy day-to-day — backups, updates, health checks, and how to reach us.

Companion to Azure services, data & security Audience IT / Cloud operations Status Pre-release draft

Contents

On this page

1Backup & recovery

Your data — governance records, audit log, attestations, AI-usage traces — lives in a Postgres database inside your own resource group.

WhatDetail
Backup methodAutomatic, built into Azure Database for PostgreSQL Flexible Server — no action required.
Retention7 days of point-in-time restore.
Geo-redundancyNot enabled by default — backups stay in the same region as your deployment.
RestoringThrough the Azure Portal (your Postgres Flexible Server → Restore) or az postgres flexible-server restore, to any point within the retention window.
If you need longer retention or cross-region backup, these are standard Azure Postgres Flexible Server settings you can raise directly in the Azure Portal — they aren't controlled by AI Pulse and can be changed without contacting SYLVA.

2Updates

The portal updates itself. There's no patch to schedule or button to click.

3Health checks & monitoring

CheckHow
Is the portal up?GET https://<your-portal-fqdn>/healthz — returns a simple liveness response. Point your own uptime monitor at this if you'd like independent alerting.
Container-level healthAzure Portal → your resource group → the portal Container App → Log stream / Metrics, same as any other Container App you operate.
Application-level signalsThe dashboards inside AI Pulse itself (cost, security posture, admin activity) are, in effect, your ongoing operational view of the AI estate it's watching.

4Capacity & availability

An honest note on what to expect at this stage of the product.

Single instance, not yet horizontally scaled. The portal runs as one replica (no automatic scale-out, no multi-region failover). This is adequate for the typical single-tenant AI-governance workload, but if your organisation needs guaranteed high availability, raise it with SYLVA before going live — this is a roadmap item, not a hidden limitation.

The self-hosted tracing component (Langfuse) scales to zero when idle and starts on demand — expect a brief cold-start delay on its first request after a quiet period. This never affects the main portal.

5Getting help

NeedContact
Product questions, onboarding help, general supporthello@sylva.sg
Marketplace / commercial / billing follow-uphello@sylva.sg

SYLVA does not have standing access into your deployment (see Azure services, data & security) — if we need to look at something specific to help you, we'll ask you to show us or grant time-boxed access, not assume it.

SYLVA Labs Pte Ltd · AI Pulse — Operations & Support Guide. Pre-release draft.

Companion to Azure services, data & security. Reflects the current shipping install (single-replica portal, 7-day Postgres backup retention).