Features Overview

AI Admin Panel ships with 156 features across deployment, management, security, and monitoring. This page provides a scannable overview of everything included.

Deployment Methods

Four ways to deploy services, each suited to different workflows:

See the Deployment section for detailed guides on each method.

Template Catalog

45 templates organized into 4 categories:

Every template includes:

• Docker Compose spec with health checks
• Configurable variables (ports, passwords, resource limits)
• Minimum resource requirements
• Security profile (secure/advanced/raw)
• Traefik routing (no exposed ports)

Browse the full catalog: Template Overview

Authentication and RBAC

• Keycloak 26 as the sole identity provider (OIDC, no JWT fallback)
• Role-based access control: Admin, Operator, Viewer
• Per-customer isolation — users only see their own services
• Session management with secure cookies
• API key support for programmatic access
• Dark-branded Keycloak login page matching panel theme

Multi-Tenant Customer Management

• Create and manage customer accounts
• Assign service plans with resource quotas
• Per-customer service isolation
• Customer portal with self-service capabilities
• GDPR-compliant data handling (export, delete)
• Usage tracking and billing-ready metrics

See Customer Management for details.

Service Plans and Quotas

• Define plans with CPU, memory, storage, and service count limits
• Assign plans to customers
• Enforce quotas at deploy time
• Overage alerts and notifications

See Service Plans for details.

Domain and DNS Management

• Automatic subdomain provisioning: {service}.panel.example.com
• Custom domain support with CNAME verification
• Cloudflare DNS integration for automated record management
• Wildcard SSL via Let's Encrypt (Traefik)
• Traefik-based routing — services never expose ports directly

See Domains and DNS for details.

Monitoring and Observability

• Real-time service status dashboard
• Container resource usage (CPU, memory, network)
• Service health checks with auto-restart
• Deploy progress via WebSocket
• Container log streaming
• Service event timeline
• Honest per-model GPU/CPU status on the AI Models page (⚡ GPU when the model uses the card, 🖥 CPU otherwise)
• AI Models cockpit: an overview with fleet status, a global GPU-utilization roll-up, and a live indicator, plus a per-instance detail/telemetry sub-page (hardware health, p50/p95 latency sparklines, running models)

GPU Support

• Host GPU auto-detection (nvidia-smi, with a PCI-bus scan fallback)
• ~36-model GPU catalog: NVIDIA datacenter (H100, H200, A100, L4, L40S, B200…), workstation/RTX (RTX 6000 Ada, A6000, Quadro RTX 5000…), GeForce (RTX 4090/3090…), and AMD Instinct (MI300X/MI325X…)
• Hardware health on the AI Models detail sub-page: each GPU with model, VRAM, driver, compute capability, and live utilization, with a global GPU-utilization roll-up on the overview (the former standalone GPUs page is merged into AI Models; /hardware/gpus redirects there)
• Model-fit hints — VRAM mapped to runnable model size (7B / 13B / 34B / 70B at Q4_K_M)
• NVIDIA Container Toolkit detection with a warning when it is missing

GPU-aware deploy

• A GPU access picker on every deploy path (Compose, Git, AI, Template): None (CPU-only), All GPUs, or Specific GPUs chosen from the detected cards
• Specific-GPU options are labelled with model, VRAM, and the largest model that fits
• Behind the scenes the panel passes the chosen GPUs to the container (Docker device requests + NVIDIA_VISIBLE_DEVICES), so Ollama / vLLM / ComfyUI actually use them
• The picker is hidden on CPU-only hosts and warns when the NVIDIA Container Toolkit is missing

Backup and Restore

• On-demand service backups
• Scheduled backup policies
• S3-compatible storage targets
• One-click restore
• Database-specific backup strategies

See Backup and Restore for details.

Notifications

• Multi-channel alerts: Email, Slack, Discord, Telegram, Webhooks
• Configurable notification triggers (deploy success/failure, health alerts, quota warnings)
• Per-customer notification preferences
• Webhook integration for custom automation

See Notifications for details.

Infrastructure

• Traefik v3 reverse proxy with automatic SSL certificate management
• PostgreSQL 16 for persistent data storage
• Valkey 8 for caching and session storage
• River async job queue for background operations
• Docker container orchestration via Docker API
• Single-binary deployment — frontend embedded in the Go binary

Security

• No ports exposed except 80/443 — all traffic routes through Traefik
• Secure-by-default template profiles
• Secret management via Docker secrets
• OIDC-only authentication (no password-based fallback)
• Automatic SSL/TLS for all services
• Container isolation per service