Enterprise AI, optimised for value, control and scale.Discover AI Economics

Tenant Plane

Manage the customer environment.

The Tenant Plane is the customer view of a T-Flux Ultra environment. It brings practical measures of operational health, workload, governance, and customer value into one enterprise dashboard.

For a single-tenant on-premises or private-cloud installation, it makes the operating picture clear: is the environment healthy, controlled, responsive, and ready to scale?

Customer environmentHealthy
ServiceSLA and incidents
WorkloadRequests and queues
GovernancePolicy and audit
CapacityGPU, storage, cost
Ingestion to retrieval healthModel runtime and answer qualitySecurity events and retention

Customer operating view

Bring the measures that matter to every T-Flux Ultra dashboard.

The Tenant Plane gives customer teams a clear view of status, SLA, workload, incidents, cost trend, data processing, model performance, and governance. Visual widgets make operational signals easy to understand and drill into when action is needed.

What it measures

See performance, control, and value across the environment.

01

Platform health and SLA

Service status, 24-hour uptime, availability and latency SLOs, error-budget burn, active incidents, time to restore, and time to detect.

02

Workload overview

Concurrent sessions, active conversations, queued jobs, requests by UI, API, batch, and agent channels, workload mix, and pipeline backlog.

03

GPU and model runtime

GPU and VRAM use, temperature and power, model concurrency, tokens per second, time-to-first-token, context length, and scheduling fairness.

04

CPU, memory, and resources

CPU use by service, RAM and GC pressure, disk and network throughput, plus container and pod health.

05

Storage and retention

Object storage, metadata size, growth rate, retention compliance, deletion failures, backup success, RPO/RTO posture, and restore testing.

06

Ingestion and processing

Documents by connector, OCR and parsing success, re-indexing events, dead-letter queues, oldest-message age, and failure reasons.

From data to outcomes

Monitor the work that turns customer data into reliable AI results.

Vector database and retrieval

Index size, shard health, embedding throughput, retrieval latency, cache performance, and permission-filtered retrieval with ACL correctness.

RAG quality and answer integrity

Citation coverage, grounding confidence, low-evidence refusals, hallucination flags, answer revision rate, and review outcomes.

Agentic workflows and batches

Jobs running, completed, or failed; duration and throughput; step timings; tool-call metrics; retries; and poison-pill detection.

API management

API calls by key and client, authentication failures, rate-limit events, key lifecycle, webhook and callback health.

Security, governance, and audit

RBAC and ABAC events, policy changes, break-glass use, corpus approvals, DLP events, audit integrity, and egress control.

Multi-LLM orchestration

Model routing, ensemble outcomes, arbitration time, quality and refusal rates, latency, tokens per second, and GPU seconds per request.

Customer benefits

Give every customer the information to operate with confidence.

Customer confidence

Clear visibility of service health, workloads, model performance, and resource utilisation across the customer environment.

Operational control

Early visibility of incidents, queue backlogs, runtime constraints, ingestion issues, and the conditions that require action.

Governance evidence

Auditable oversight of policy, access, corpus change, DLP, retention, audit logs, and controlled outbound activity.

Capacity and value

Cost proxies, capacity headroom, saturation forecasts, and what-if planning to support informed investment decisions.

Cost, capacity, and forecasting

Plan before demand becomes a constraint.

Track GPU-hours, CPU-hours, storage growth, egress, cost per 1,000 tokens, and cost per exported deliverable. Capacity views show headroom, days until saturation, and the likely impact of adding a GPU node on p95 latency.

Reporting and exports

Deliver evidence that is ready for operational, governance, and security review.

Monthly SLA report

Uptime, latency, incidents, and error budget.

Governance and audit report

Policy and corpus changes, approvals, and privileged actions.

Security posture summary

Authentication failures, DLP, egress blocks, and key rotations.

Exports are available in PDF and CSV where applicable, with timestamp, tenant, and signature hash.