Operations - Enterprise AI Platform (English)
French version: ../.
| Document | Content |
|---|---|
| installation.md | Local compose, Helm on-prem GPU, air-gapped install with Harbor mirror |
| secrets.md | OpenBao secrets inventory, ExternalSecrets and bootstrap order |
| configuration.md | Configuration of every component |
| platform-integration.md | Control plane and infrastructure alignment: background worker, Prometheus metrics, OPA, OpenFGA, Keycloak, EAI_* variables |
| security-hardening.md | mTLS, SPIFFE, network policies, secrets, image signing |
| network-dns-hardening.md | Default deny, flow matrix, per workload L7 DNS, cloud metadata, CoreDNS and NodeLocal DNSCache |
| waf-gateway.md | APISIX: TLS, Coraza WAF and OWASP CRS, internal headers, limits, CORS, portal ingress |
| pod-security.md | Pod security context, GPU exceptions, Pod Security Admission, Kyverno policies |
| media-workers.md | Multimodal GPU workers (python -m eai.media_worker): operations, KEDA, NATS subjects, lease and requeue |
| mcp-runtime.md | MCP Runtime (ToolHive Operator): ephemeral MCP instances isolated per tenant, project and user, hardening, separate audit logs |
| data-services-hardening.md | PostgreSQL, Redis, NATS, OpenBao, Qdrant, ClickHouse |
| ci-security-gates.md | CI security gates and failure conditions |
| ha-dr.md | HA, DR, RPO/RTO targets, Velero, CNPG backups, multi-AZ |
| load-testing.md | Load testing plan (k6 scripts in deploy/loadtest/) |
| upgrade.md | Upgrade procedure |
| runbooks.md | Runbooks: provider outage, cost spike, key leak, prompt injection, GPU saturation, Control Plane down, worker down, media backlog, database unavailable, kill switch |
| administration.md | Administrator guide: roles, first start, day to day tasks |
| compliance-mapping.md | Compliance mapping: control families, audits and evidence sets |
| phase11-readiness.md | Phase 11 requirements built in from phase 1 |
| installation-handbook.md | Customer installation from the signed release bundle, isolated site |
| demarrage-local.md | Local start: backend and portal, sign in, install a service from the portal |
| alerting.md | Prometheus signals, alert rules, alerts received by the Control Plane, broken audit chain |
| slo.md | SLO, error budgets and trace propagation |
| performance-results.md | Measured performance, quotas, RTO and RPO |
| analytics-export.md | Analytics export to ClickHouse |
| siem-audit.md | SIEM forwarding, mutation audit, WORM export and erasure |
| postgresql.md | Central PostgreSQL: logins, backup and restore |
| redis.md | Central Redis |
| licensing.md | Platform licences |
| third-party-licences.md | Exact licences of PostgreSQL, Redis and the telephony bricks |
| voice-runtime.md | Voice runtime: consent, identity, interruption, personal data, engines |
Installation and reference
Step by step installation guides (Kubernetes with Helm, OS packages, Docker Compose, make, air-gapped site, sizing,
ports, TLS, first configuration, verification, upgrade, uninstall, hosting): ../../en/install/index.md.
Generated references: configuration,
Helm values, OS compatibility.
Sécurité / Security
Security documentation: ../../security/en/README.md (threat model, injection matrix, configuration, runbooks, testing).