Operations
Component health
Recent errors
Throughput & Latency
TPSTTFT p95
Requests / min · peak—
Queue depth · peak—
KV cache % · peak—
CPU % · peak—
Memory GB · peak—
Network Mb/s · peak—
Recent events
Business Overview
Private enclaves & metered consumption across all organizations.
Top organizations
| Organization | Requests | Tokens | Share of tokens |
|---|
Fleet
DGX Spark + Apple Silicon nodes serving the shared inference gateway
LLM · Qwen3.8-27B
Embedding · voyage-4-nano
Rerank
API Servers
Workers
Agents · Hermes engine
Agentic engine (celestis-agentic-engine) on Rancher — deployment health, resource use & the live agent catalog.
Agent catalog
| Agent | Owner | Registered |
|---|
Enclaves · org routing
Per-customer private infra. Auth0 org_id → the org's backend IPs (load-balanced).
Hosts · MAC registry
Physical hosts enrolled by MAC; live IP resolved from ARP (survives DHCP changes).
Usage & Billing
Metered requests & tokens per organization, recorded at the enclave router.
| Organization | Requests | Prompt tok | Completion tok | Total tok | Lifetime tok |
|---|
Logs
Fleet logs aggregated in Loki — ingest, inference & services.
API Keys
Gateway auth for inference.celestis.ai — minted into the nginx map, validated, auto-rollback.
Keys are minted into the gateway auth map · validated with nginx -t · auto-rollback on error
Compliance · SOC 2 Type 2
Continuous controls monitoring & audit evidence — a readiness view, not a certification.
Documentation
Platform guides, read right here — self-hosted behind this console's login.
Loading…
