Backup & Restore
Backing up and restoring every persistent store in a GENIE.AI deployment — ArangoDB knowledge base, PostgreSQL (Kong + Keycloak), Redis cache, document-repository uploads, and supporting volumes.
Operations
Day-to-day operations for a running GENIE.AI deployment — backup, restore, updates, scaling, troubleshooting, and operator-facing diagnostic surfaces.
Deployment gets GENIE.AI running; this section keeps it running. It is for operators and administrators responsible for a live deployment: backing up data, applying updates, scaling, and diagnosing problems.
For real-time health, dashboards, and alerting, see Observability — this section assumes you already have the observability stack on and are watching it.
Pages in this section
- Backup & restore — ArangoDB dump/restore and Kong configuration restore, including the network-name and password gotchas in the shipped scripts.
- Updates — updating models, service images, and the stack; what forces re-ingestion; the rollback procedure.
- Scaling — horizontal replicas and vertical GPU memory; what scales well and what does not.
- Troubleshooting — common problems, trace-first / logs-second methodology, and pointers to deeper notes.
- Admin Logs — the operator-facing Logs tab,
the
/api/admin/logs/*endpoints, the MELT seam, and the VL fallback configuration. - Ingestion Log — per-chunk ingestion progress feed, canonical messages, ArangoDB AQL access, and trace correlation.
- Health Checks —
/api/healthand/api/admin/system-healthendpoints, Docker Swarm healthcheck wiring, Kubernetes probe recipes, and failure-mode troubleshooting. - Security Hardening — Helmet defaults, CORS, CSP, and rate-limiting configuration; per-service transport security knobs.
Distributed tracing (operator how-to)
W3C traceparent propagation across the RAG pipeline, the span taxonomy, and automatic PII filtering.
Updates
Updating models, service images, and the GENIE.AI stack — and which changes require re-ingesting the knowledge base.
Troubleshooting
Common GENIE.AI operational problems, where to look first, and pointers to the deeper troubleshooting notes in the deployment guides.
Admin Logs
Operator-facing Logs tab in the GENIE.AI admin UI — VictoriaLogs-backed log search, summaries, and the MELT seam.
Ingestion Log
The per-chunk ingestion progress log surfaced by dataprep — how to read it, where it lives, and what its messages mean.
Health Checks
Operational health endpoints — the public /api/health probe and the admin /api/admin/system-health diagnostic — with wiring examples for Swarm, Kubernetes, and Prometheus.
Security Hardening
Defense-in-depth controls built into the GENIE.AI backend and frontend — the sensitive-path blocker, Helmet security headers, robots.txt policy, and the user-profile static file surface.
Scaling GENIE.AI
How to scale every service in the GENIE.AI stack — stateless services scale freely, stateful services need replicas with shared storage, AI services are GPU-bound.