Platform health and service status
See uptime, SLO attainment, end-to-end latency, error rates, queues, and the oldest outstanding jobs.
Control plane
Manage the platform.
The Control Plane is the central command and governance layer for the entire T-Flux Ultra environment. It provides a single, auditable view of platform health, workload, performance, governance, security posture, and capacity and cost.
It answers one operational question: is T-Flux operating securely, efficiently, and within policy?
One operating view
Control every aspect of a customer’s T-Flux Ultra environment.
For single-tenant T-Flux installations in a private cloud or on-premises, the Control Plane brings service status, infrastructure performance, policy enforcement, and governance evidence together for platform administrators, enterprise IT, security, risk, and T-Flux operators.
Core functions
Operate with visibility, control, and evidence.
Workload and capacity management
Monitor concurrent sessions, storage growth, compute and GPU capacity, workload queues, and resource utilisation.
Security and governance
Bring security posture, access controls, governance policies, audit controls, alerts, incidents, and exceptions into one view.
Infrastructure performance
Review CPU, RAM, disk IOPS, network throughput, GPU utilisation, VRAM, temperature, and power across the environment.
Cost and resource utilisation
Track GPU-hours, CPU-hours, storage growth, cost proxies, cost per deliverable, and available headroom.
Configuration and policy enforcement
Manage platform configuration, administrative controls, cross-tenant monitoring, and policy enforcement.
Views by role
Information for the people who operate the environment.
Customer Admin
Service status, consumption, API keys, governance, and billing or cost proxies.
Operations
Incidents, queues, latency, infrastructure, logs, and remediation actions.
Security and Risk
Policy changes, access events, DLP events, audit integrity, and retention compliance.
Data Steward
Ingestion quality, re-indexing, retrieval health, and corpus change control.
Model Admin
Model routing, throughput, quality KPIs, GPU scheduling, and model failures.
Dashboard coverage
See the whole operating picture.
Operational tiles update within ten seconds and governance tiles within sixty seconds, giving teams current, auditable visibility across the platform.
Overview
Uptime across 24 hours, 7 days, and 30 days; SLO attainment; latency; requests; concurrent sessions; error budget; queues; and storage growth.
Compute and models
GPU and VRAM utilisation, CPU and memory, disk and network performance, tokens per second, routing distribution, and model concurrency.
Data plane
Connector throughput, parse and OCR success, embedding queues, vector index health, retrieval latency, cache hit rate, and ACL-filtered retrieval.
Quality and governance
Citation coverage, grounding and verification scores, low-evidence refusals, revision rate, corpus changes, policy changes, and guardrail rule hits.
Security and audit
Authentication failures, privileged actions, break-glass usage, DLP and PII events, egress controls, audit integrity, and retention compliance.
API and integrations
API usage, rate limiting, API-key lifecycle, webhook and callback health, retry counts, and integration latency.
Batches and agents
Batch jobs, duration distribution, workflow step timings, tool-call failures, retry rate, and poison-pill detection.
Board and audit-ready controls
Turn operating data into assurance and forward-looking management.
Security and audit
Surface privileged actions, break-glass use, policy changes outside workflow, DLP events, egress controls, and tamper-evident audit integrity.
Storage and retention
Track storage by artefact class, retention compliance, deletion failures, backup and restore status, RPO/RTO posture, and restore-test history.
Cost and capacity
Forecast GPU-hours, CPU-hours, storage growth, cost proxies, headroom, saturation, and the impact of capacity changes.
Alerts and thresholds
Make exceptions visible before they become business risk.
Default SEV1, SEV2, and SEV3 baselines cover availability, latency, error rate, queues, capacity, audit integrity, retention, and security. Every threshold can be configured per tenant and audited.
Availability
Below 99.5% daily is SEV2; below 99.0% is SEV1.
Latency and errors
p95 latency above target for a sustained period and elevated 5xx errors trigger escalation.
Queues and capacity
Queue depth, oldest-job age, GPU VRAM, disk utilisation, and vector index health are monitored against thresholds.
Security and retention
Audit-integrity failures, break-glass usage, unauthorised policy changes, and retention failures are surfaced for action.
Drill down and report
Trace the signal, investigate the cause, and export the evidence.
Every widget can drill down to request traces, token statistics, model selection, retrieval hits, citations, job details, retries, infrastructure events, and correlated request, job, and document IDs.
One-click exports provide monthly SLA reports, governance and audit reports, and security posture summaries in PDF and CSV formats, each with timestamp, tenant, and signature hash.