This guide explains the practical workflow for founders, operators, site owners, enterprise customers, customer-success teams, and launch owners.
Each step tells the user what to do, who owns it, and what evidence confirms it is complete.
Owner: Owner or operator
You enter the platform with a valid Supabase session.
Owner: Account owner
A tenant is created for your team, customer, or pilot network.
Owner: Operations lead
A physical or commercial host location is registered.
Owner: Operator + infrastructure engineer
A GPU or inference node exists in the registry with a secure telemetry key.
Owner: Operator
The team can understand health, incidents, usage, and readiness from dashboards.
Owner: Customer-success lead
Pilot customers, site owners, and partners have clear workflows.
Owner: Admin + customer engineer
A customer can test the OpenAI-compatible gateway safely.
Owner: Launch owner
A pilot can move from demo to controlled production readiness.
Online, degraded, offline, and retired node states show operational availability and maintenance risk.
GPU utilization, memory, temperature, latency, power draw, and heartbeat events indicate whether a node is safe for inference workloads.
The router selects nodes based on health, latency, available power, residency rules, runtime readiness, and tenant quota.
Warnings and critical telemetry should create or update incident records for operator triage and audit visibility.
Usage events, committed capacity, node revenue, and site-owner payouts define pilot economics.
Compliance packs summarize node authentication, RLS, data residency, SLA policy, and incident response posture.
Create the organization, approve billing/commercial model, assign administrators, and approve go-live.
Manage sites, nodes, runtimes, customer keys, compliance evidence, organization settings, and team access.
Monitor telemetry, respond to incidents, schedule maintenance, run failover checks, and validate SLA posture.
Read dashboards, reports, guides, executive status, and launch-readiness evidence without changing settings.
Track site earnings, power consumption, uptime, maintenance windows, and contract status.
Use OpenAI-compatible APIs, follow quotas, review model endpoints, and escalate support through agreed channels.
Open /status, /operator, /observability, and /incidents. Confirm no critical telemetry or offline production nodes.
Review /usage, /billing-control, /customer-success, and /support. Update customer or partner next actions.
Review /risk-register, /deployments, and /go-live. Close completed gates and assign blockers before the next day.
Review /board-metrics, /executive, /finance, /partners/pipeline, and /commercial to prepare investor or leadership updates.
Google/GitHub/magic-link sign-in works and stale session recovery is documented.
Organization, role, site, and node setup are complete for the pilot customer.
At least one node sends heartbeat, GPU, temperature, latency, and power metrics.
Node API keys, RLS, customer access model, and compliance evidence have been reviewed.
Pilot plan, pricing, contract pack, procurement path, and site-owner economics are documented.
Escalation owner, response path, risk register, and training materials are ready.