Admin Guide

The Turqoa Admin panel provides system administrators with tools for managing users, configuring the platform, monitoring system health, and maintaining compliance through audit logs.

Admin Panel Overview

The Admin panel is accessible to users with the admin role. It is organized into the following sections:

Section	Purpose
Dashboard	System health overview, service status, and key metrics
User Management	Create, edit, and deactivate user accounts; assign roles
System Configuration	Gate lane settings, AI model thresholds, decision rules, and integration parameters
Audit Logs	Searchable audit trail of all system and user actions
License	License status, feature entitlements, and renewal information
Integrations	Connection status and configuration for TOS, cameras, and SSO

System Configuration

Administrators manage Turqoa's operational parameters through the Admin panel. Key configuration areas include:

Decision Engine Thresholds

decision_engine:
  auto_approve_threshold: 0.95    # Confidence above this: auto-approve
  auto_reject_threshold: 0.20     # Confidence below this: auto-reject
  referral_timeout_seconds: 120   # Time before referred transaction escalates
  shadow_mode: false              # When true, decisions are logged but not enforced

Gate Lane Configuration

Setting	Description	Default
`lane_count`	Number of active gate lanes	4
`direction`	Inbound, outbound, or bidirectional per lane	Inbound
`cameras_per_lane`	Number of cameras per lane	6
`barrier_integration`	Enable/disable automatic barrier control	Enabled
`max_transaction_time`	Maximum seconds before a transaction times out	300

AI Model Management

Administrators can view and manage the AI models deployed in the inference pipeline:

# List deployed models
turqoa admin models list

# Output:
# MODEL                    VERSION   STATUS    GPU    VRAM
# ocr-license-plate        v2.4.1    active    0      2.1 GB
# ocr-container-number     v3.1.0    active    0      2.3 GB
# damage-detection         v2.8.0    active    0      4.2 GB
# chassis-classification   v1.5.2    active    1      1.5 GB
# seal-detection           v1.2.0    active    1      1.0 GB
# hazmat-placard           v1.0.3    active    1      1.1 GB

Model updates are deployed via the Admin panel with zero-downtime rolling updates. The previous model version is retained for instant rollback.

Health Monitoring

The Admin dashboard provides real-time visibility into system health:

Service Status Panel

Service	Health Check	Alert Threshold
API Server	HTTP 200 on `/health`	3 consecutive failures
AI Inference	GPU utilization + queue depth	GPU > 90% for 5 min
Database	Connection pool + query latency	Latency > 500ms
Event Bus (Kafka)	Consumer lag	Lag > 1000 messages
Camera Streams	FPS per stream	FPS < 10 for 60s
TOS Connection	API response time	Timeout or 5xx for 3 min

Alerting

Alerts are delivered through multiple channels:

Command Center --- Banner notification visible to all logged-in users
Email --- Configurable recipient list per alert severity
Webhook --- POST to external systems (Slack, PagerDuty, SIEM)

alerting:
  channels:
    - type: email
      recipients: ["admin@terminal.example.com"]
      severity: [critical, high]
    - type: webhook
      url: https://hooks.slack.com/services/T00/B00/xxx
      severity: [critical, high, medium]
    - type: pagerduty
      routing_key: ${PAGERDUTY_KEY}
      severity: [critical]

Next Steps

User Management --- Create accounts and manage roles
Audit Logs --- Query and export audit data