Workspace

The landing page: KPI cards, system health, the model list, and recent experiments — all scoped to the selected workspace. Use the Overview and Dashboard tabs, start a New Experiment, or click a model row for its report.

Executive Dashboard

A board-level rollup across financial, compliance, explainability (XAI), and operational dimensions. Click Re-scan to recompute the compliance score, or Generate Executive Summary PDF to export.

Governance Canvas

A visual, node-based view of your AI estate. Pan/zoom to explore; click nodes to inspect governance state. Use Import to bring code graphs in as governance nodes.

Model Registry

The system of record for every governed model. Search and filter by status/framework, import from MLflow, and open any model for metrics, lineage, and the safety-gated promotion flow.

Prompt Registry

Version-controlled prompts with built-in safety evaluation. Filter by lifecycle status, register a prompt, and open a prompt to run its safety eval — promotion to production is blocked until the safety gate passes.

LLM Gateway

Central routing for LLM providers with data-classification enforcement. Add providers (with API-key env-var indirection), add routing rules (e.g. confidential traffic to on-prem), and test connectivity.

LLM Analytics

Cost, usage, and quality analytics. The LLM Usage tab covers tokens, cost, budget, and chargeback; the Analytics tab covers project spend, latency, drift, and infra health. Click a KPI to drill into the cost breakdown.

Feedback

Aggregated end-user feedback to drive fine-tuning. Review KPIs and the tag cloud, filter by rating, and export a curated dataset for fine-tuning.

Monitoring

Real-time monitoring of deployed agents and models — predictions/hour, GPU, P99 latency, a live status badge, and the drift monitor. Click an agent for its trace; use the per-agent kill switch to halt it.

Compliance

Framework-by-framework compliance posture. Switch frameworks with the tabs, review each model's compliant/non-compliant status, and open a model for its control breakdown and recommendations.

Policy Center

Policy-as-code bundles and their evaluation results. Add a bundle from a git source, click Sync to pull the latest, and open a failed evaluation to see the violating policy file and rule.

Audit Log

A tamper-evident, hash-chained record of every governance action — actor, event type, severity, and category filters. Click a row to expand its detail; Export to download; the chain is verifiable end-to-end.

AI Robustness

Adversarial robustness evaluation (NIST / EU AI Act). Select a model and attack types, set the epsilon budget, and run in simulation or full mode. The History tab lists past runs.

LLM Output Evaluation

LLM-as-judge scoring of prompt/response pairs for hallucination, toxicity, and factual consistency. Fill in the prompt and response, submit, and review the overall result and per-check breakdown. The History tab lists prior evaluations.

Bias Evaluation

Fairness metrics across protected attributes (EU AI Act / NIST). Select a model and dataset, define protected attributes and groups, and run the evaluation. The History tab lists prior runs with disparate-impact metrics.

Federation

Privacy-preserving federated learning across participants — round progress, differential-privacy parameters, and a convergence chart. Create a federation, open one for its round history, and start the next round.

Data Sources

The registry of data sources feeding your models, with lineage. Register a source (S3, Snowflake, BigQuery), then pick it in the left list to load its URI, metadata, and linked models in the right pane.

External AI Governance

Governance over external/third-party AI providers, with per-provider connection status. Add a provider, test its connection, and browse its available models.

Governance Marketplace

Pluggable governance extensions with versions and active/inactive status. Install a plugin, toggle it on/off, edit its config, and Reload to apply.

GitOps & Deployment Pipeline

GitOps-driven, compliance-gated deployments. The Manifests tab lists deployment manifests (traffic splits, commit SHAs); the Runs tab runs the Validate → Compliance Scan → Deploy → Activate pipeline with a live console.

User Management

Visible to admin/ops. Workspace users with persona badges, status, last-seen, and SOC 2 access-review (CC6) due flags. Click a user for detail; Revoke to remove access.

Settings

Workspace and platform configuration across tabs — Cloud Provider, Deployment, Agent Gateway, Budget & Efficiency, Private Clusters, MLflow, API Keys, General, and Infrastructure (clusters, service health, DR policy).