Ops-Center
The AI-Powered Infrastructure Command Center
βββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ
β β
β βββββββββββββββββββββββββββββββββββββββββββββββββββββββββ β
β β β β
β β You: "Colonel, spin up a new org with Pro access β β
β β and notify the team on Slack." β β
β β β β
β β Colonel: Done. Created "Acme Corp" with 3 roles, β β
β β assigned Pro tier, SSO configured. β β
β β Slack notification sent to #ops. β β
β β β β
β βββββββββββββββββββββββββββββββββββββββββββββββββββββββββ β
β β
β Users ββββββββββββββββββββ 847 active β
β API ββββββββββββββββββββ 62% quota β
β Rev ββββββββββββββββββββ $12.4k MRR β
β GPU ββββββββββββββββββββ 38% utilization β
β β
βββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ
Manage users, billing, AI models, organizations, and infrastructure β all from one place.
Talk to The Colonel, your AI platform engineer, to orchestrate it all with natural language.
Get Started Β· The Colonel Β· Features Β· API Reference Β· Documentation
What is Ops-Center?
Ops-Center is a full-stack operations dashboard for managing AI-powered infrastructure. It combines the capabilities of an AWS Console, Stripe Dashboard, Auth0 admin panel, and LLM gateway into a single, self-hosted platform β with an AI agent (The Colonel) that can operate it all through conversation.
Starting with v3.0, Ops-Center also supports federation β connect multiple instances into a distributed mesh network, route inference across nodes, and manage GPU resources across cloud providers.
βββββββββββββββββββββββββββββ
β USERS β
β Browsers Β· APIs Β· Apps β
ββββββββββββββ¬βββββββββββββββ
β
ββββββββββββββΌβββββββββββββββ
β β CLOUDFLARE β
β CDN Β· DDoS Protection β
β DNS Β· WAF Β· Edge Cache β
ββββββββββββββ¬βββββββββββββββ
β
ββββββββββββββΌβββββββββββββββ
β π TRAEFIK β
β Reverse Proxy Β· SSL/TLS β
β Let's Encrypt Β· Routing β
β Web Hosting (sites) β
ββββββββββββββ¬βββββββββββββββ
β
ββββββββββββββββββββββββββββββββββΌβββββββββββββββββββββββββββββββββ
β β β
βΌ βΌ βΌ
ββββββββββββββββββββ ββββββββββββββββββββββββββ ββββββββββββββββββββββ
β USER DASHBOARD β β OPS-CENTER β β ADMIN DASHBOARD β
β β β (FastAPI + React) β β β
β Credits Β· Usage β β β β Services Β· GPUs β
β Subscription βββββββββββΊβ The Colonel (AI Agent) βββββββββΊβ Users Β· Billing β
β Apps Β· API Keys β β "Deploy the service" β β Orgs Β· Analytics β
ββββββββββββββββββββ βββββββββββββ¬βββββββββββββ ββββββββββββββββββββββ
β
ββββββββββββββββ¬ββββββββββββββββββΌββββββββββββββββββ¬βββββββββββββββ
β β β β β
βΌ βΌ βΌ βΌ βΌ
ββββββββββββββββ βββββββββββββββ ββββββββββββββββ ββββββββββββββββ ββββββββββββββ
β APPS β β AUTH & SSO β β AI & LLM β β BILLING β β MONITORING β
β MARKETPLACE β β β β β β β β β
β β β Keycloak β β LiteLLM Proxyβ β Stripe β β Prometheus β
β Open-WebUI β β SSO β β 100+ cloud β β (payments) β β (metrics) β
β Bolt.diy β β β β models β β β β β
β Forgejo β β βββββββββββ β β β β Lago β β Grafana β
β Center-Deep β β β Google β β β Ollama β β (metering & β β (dashbds) β
β Presenton β β β GitHub β β β (local LLM) β β invoicing) β β β
β Web Hosting β β β MS 365 β β β vLLM β β β β Umami β
β β β βββββββββββ β β (GPU infer.)β ββββββββββββββββ β (web β
ββββββββββββββββ β β β llama.cpp β β analytics)β
β MS 365 Emailβ β β ββββββββββββββ
β (SMTP / β β TTS models β
β Graph API)β β STT models β
βββββββββββββββ β Embeddings β
β Reranking β
β Image Gen β
β DALL-E Β· SD β
β Imagen β
ββββββββ¬ββββββββ
β
ββββββββββββββββββββββββββββββββΌβββββββββββββββββββββββββββββββ
β β β
βΌ βΌ βΌ
ββββββββββββββββββββ ββββββββββββββββββββββ ββββββββββββββββββββ
β PostgreSQL β β Redis β β AI Memory β
β β β β β β
β Users Β· Orgs β β Sessions Β· Cache β β Kuzu (Graph DB) β
β Billing Β· Tiers β β Rate Limiting β β Colonel memory β
β Audit Logs β β Usage Counters β β β
β API Keys β β β β Mem0 (Vectors) β
β App Permissions β β β β Semantic search β
ββββββββββββββββββββ ββββββββββββββββββββββ ββββββββββββββββββββ
Screenshots
Admin Dashboard β System health, services, GPU status, hosted sites |
User Dashboard β Credits, usage, subscription, spending breakdown |
AI Model Management β 100+ LLMs, curated lists, BYOK configuration |
Service Management β Docker containers, health checks, logs |
|
System Monitoring β CPU, RAM, disk, GPU metrics, real-time graphs |
Apps Marketplace β Tier-based app access, org grants, SSO integration |
Settings β Email providers, system configuration, admin controls |
|
Federation Mesh
Ops-Center v3.0 introduces Federation β connect multiple Unicorn Commander instances into a distributed mesh network. Route inference requests across nodes based on hardware capability, manage GPU resources across cloud providers (RunPod, Lambda, Vast.ai), and aggregate metering across your fleet.
- Node Registry β Discover and monitor peer nodes automatically
- Inference Routing β Smart routing based on model availability, GPU memory, and load
- Distributed Metering β Aggregate usage and billing across federated nodes
- Cloud Provisioning β Bootstrap GPU instances on RunPod, Lambda, Vast.ai
- Resilience β Automatic failover when nodes go offline
- Access Control β Per-node, per-model access policies
The Colonel β AI Platform Engineer
The Colonel is Ops-Center's flagship feature: an AI-powered infrastructure agent that lives inside your dashboard. Built on Claude/GPT with real-time WebSocket streaming, tool-use capabilities, and a modular skill system, The Colonel can execute actual operations on your infrastructure through natural language.
βββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ
β THE COLONEL β Online β
βββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ€
β β
β You: Show me which services are unhealthy β
β β
β Colonel: Running health checks... β
β βββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ β
β β PostgreSQL ........... healthy (2ms) β β
β β Redis ................ healthy (1ms) β β
β β Keycloak ............ healthy (45ms) β β
β β vLLM ................ degraded (GPU temp 82C) β β
β β Traefik ............. healthy (3ms) β β
β βββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ β
β vLLM is running but the GPU is warm. Want me to check β
β the inference queue depth? β
β β
β You: Yes, and restart it if the queue is backed up β
β β
β Colonel: Queue depth is 47 (threshold: 20). β
β Requesting confirmation to restart vLLM... β
β ββββββββββββββββββββββββββββββββββββββββ β
β β Restart vLLM inference server? β β
β β [ Approve ] [ Deny ] β β
β ββββββββββββββββββββββββββββββββββββββββ β
β β
β [Type a message...] [Send] β
βββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ
Colonel Skills
| Skill | Capabilities | Confirmation Required |
|---|---|---|
| System Status | CPU, RAM, disk, GPU metrics, uptime | No |
| Docker Management | List, start, stop, restart containers, view logs | Destructive ops only |
| Service Health | Check all services, latency, connection status | No |
| Log Viewer | Tail logs, search patterns, filter by service | No |
| PostgreSQL Ops | Query stats, table sizes, active connections, vacuum | Write ops only |
| Keycloak Auth | User lookup, session management, realm status | Write ops only |
| Forgejo Git | Repo stats, user management, org operations | Write ops only |
| Bash Execution | Run shell commands with safety controls | Always |
How It Works
User Message βββΊ WebSocket Gateway βββΊ LLM (Claude/GPT)
β β
β tool_calls
β β
βΌ βΌ
Stream chunks βββ Skill Router βββΊ Skill Executor
to browser β
βΌ
Safety Layer
(confirmation for
destructive ops)
β
βΌ
ββββ Memory Layer ββββ
β Kuzu (Graph DB) β
β Mem0 (Vectors) β
β Redis (Sessions) β
ββββββββββββββββββββββ
β
βΌ
Audit Log
(every action recorded)
Key Design Decisions:
- Human-in-the-loop: Destructive operations require explicit user approval
- Streaming: Real-time WebSocket delivery β see The Colonel think and act
- Auditable: Every skill execution logged with parameters, results, and duration
- Extensible: Add new skills by dropping a
SKILL.mdfile inbackend/colonel/skills/ - Configurable: Adjust personality, enabled skills, and LLM model via admin UI
Features
Platform Management
User Management
|
Billing & Subscriptions
|
Organizations
|
LLM Gateway
|
Apps Marketplace
|
Monitoring & Analytics
|
Billing Flexibility
Ops-Center adapts to your deployment scenario:
# Personal server β no billing at all
BILLING_ENABLED=false
# Internal company β everyone gets free access
CREDIT_EXEMPT_TIERS=*
# SaaS platform β full billing with custom exempt tiers
CREDIT_EXEMPT_TIERS=free,enterprise,staff,beta_tester
Quick Start
Option 1: Full Stack (Recommended)
Use the Unicorn Commander umbrella repo for one-command setup with Keycloak SSO, Brigade, and all infrastructure:
git clone --recursive https://github.com/Unicorn-Commander/Unicorn-Commander.git
cd Unicorn-Commander
./setup.sh
This auto-imports the Keycloak uchub realm with pre-configured OAuth clients and identity provider stubs (Google, GitHub, Microsoft).
Option 2: Standalone Docker Compose
git clone https://github.com/Unicorn-Commander/Ops-Center-OSS.git
cd Ops-Center-OSS
# Copy environment template and configure
cp .env.example .env.auth
nano .env.auth
# Start everything
docker compose -f docker-compose.direct.yml up -d
# Verify
curl http://localhost:8084/api/v1/system/status
Option 3: Bare Metal
# Run the installer (Python, Node.js, Docker, all dependencies)
sudo ./install.sh
# Configure
cp .env.example .env.auth && nano .env.auth
# Start
sudo systemctl start ops-center
sudo systemctl enable ops-center
Option 4: Development Mode
npm install && pip install -r backend/requirements.txt
# Terminal 1: Backend
cd backend && uvicorn server:app --reload --port 8084
# Terminal 2: Frontend
npm run dev # β http://localhost:5173
# Build for production
npm run build && cp -r dist/* public/
First Login
- Navigate to
http://localhost:8084 - If using the umbrella repo, the Keycloak realm is already configured
- If standalone, set up Keycloak and create your admin user
- Configure SSO providers (Google, GitHub, Microsoft) if desired
- Visit
/adminto access the full dashboard
Keycloak Realm
A scrubbed Keycloak uchub realm export is included at keycloak-theme/realm/uchub-realm.json. It contains:
- Pre-configured OAuth clients (
ops-center,brigade) - Identity provider stubs (Google, GitHub, Microsoft) β add your own OAuth credentials
- Client scopes, roles, and authentication flows
When using the umbrella repo's docker-compose.yml, this realm is auto-imported on first boot via Keycloak's --import-realm flag.
API at a Glance
624+ endpoints across 12 API domains.
User Management β CRUD, bulk ops, roles, API keys, impersonation
GET /api/v1/admin/users List users (10+ filters)
POST /api/v1/admin/users/comprehensive Create user (full provisioning)
GET /api/v1/admin/users/{id} User details
POST /api/v1/admin/users/bulk/import CSV import (up to 1,000)
GET /api/v1/admin/users/export CSV export
POST /api/v1/admin/users/bulk/assign-roles Bulk role assignment
POST /api/v1/admin/users/{id}/api-keys Generate API key
POST /api/v1/admin/users/{id}/impersonate/start Admin impersonation
GET /api/v1/admin/users/analytics/summary User statistics
Organizations β Multi-tenant with feature grants
GET /api/v1/organizations List organizations
POST /api/v1/organizations Create organization
GET /api/v1/organizations/{id}/members List members
POST /api/v1/organizations/{id}/invite Invite member
POST /api/v1/admin/orgs/{id}/features Grant feature to org
DELETE /api/v1/admin/orgs/{id}/features/{key} Revoke feature
GET /api/v1/admin/features/available List grantable features
LLM & Credits β OpenAI-compatible chat, images, model catalog
POST /api/v1/llm/chat/completions Chat completion (OpenAI-compatible)
POST /api/v1/llm/image/generations Image generation
GET /api/v1/llm/models List all models
GET /api/v1/llm/models/categorized BYOK vs Platform breakdown
GET /api/v1/llm/models/curated?app=bolt-diy Per-app curated lists
GET /api/v1/llm/usage Usage statistics
Billing & Subscriptions β Stripe + Lago, self-service management
GET /api/v1/billing/plans List subscription plans
POST /api/v1/subscriptions/upgrade Upgrade tier
POST /api/v1/subscriptions/downgrade Downgrade tier
POST /api/v1/subscriptions/cancel Cancel subscription
GET /api/v1/subscriptions/preview-change Preview cost changes
GET /api/v1/payment-methods List payment methods
POST /api/v1/payment-methods/setup-intent Add new card
GET /api/v1/usage/current Current usage stats
The Colonel β AI agent via WebSocket + REST
WS /api/v1/colonel/ws WebSocket (streaming chat)
GET /api/v1/colonel/config Current configuration
PUT /api/v1/colonel/config Update configuration
GET /api/v1/colonel/status Health + session stats
GET /api/v1/colonel/sessions List chat sessions
DELETE /api/v1/colonel/sessions/{id} Delete session
GET /api/v1/colonel/audit Audit log
Full API documentation: docs/API_REFERENCE.md
Technology Stack
FRONTEND BACKEND INFRASTRUCTURE
βββββββββββββββββββββ βββββββββββββββββββββ βββββββββββββββββββββ
React 18 + Vite FastAPI (async Python) Docker + Compose
Material-UI v5 PostgreSQL + asyncpg Traefik (SSL/TLS)
React Router v6 Redis (cache + sessions) Keycloak SSO
Chart.js Lago (billing engine) Let's Encrypt
Tailwind CSS Stripe (payments) NVIDIA GPU support
WebSocket LiteLLM (LLM proxy)
Kuzu + Mem0 (AI memory)
Performance
API Response Time ββββββββββββββββββββββββββββββ 2-8ms avg
Database Queries ββββββββββββββββββββββββββββββ <1ms execution
Redis Cache ββββββββββββββββββββββββββββββ ~1ms lookups
Usage Tracking ββββββββββββββββββββββββββββββ <5ms overhead/request
Container Footprint βββββββββββββββββββββββββββββ 0.66% RAM, 0.20% CPU
Project Structure
ops-center/
βββ backend/ # FastAPI backend
β βββ server.py # Main application
β βββ colonel/ # The Colonel AI Agent
β β βββ websocket_gateway.py # WebSocket streaming
β β βββ skill_router.py # Tool-call routing
β β βββ skill_executor.py # Skill execution engine
β β βββ skill_loader.py # SKILL.md parser
β β βββ safety.py # Confirmation + safety
β β βββ system_prompt.py # Dynamic prompt builder
β β βββ a2a_server.py # Agent-to-Agent protocol
β β βββ memory/ # Kuzu graph + Mem0 vectors
β β βββ skills/ # 8 skill definitions
β βββ routers/colonel.py # Colonel REST API
β βββ litellm_api.py # LLM proxy + credits
β βββ user_management_api.py # User CRUD + bulk ops
β βββ billing_analytics_api.py # Billing + analytics
β βββ org_api.py # Organization management
β βββ my_apps_api.py # Tier-based app access
β βββ keycloak_integration.py # SSO integration
β βββ lago_integration.py # Lago billing
β βββ migrations/ # SQL schemas
β
βββ src/ # React frontend
β βββ pages/
β β βββ admin/
β β β βββ ColonelChat.jsx # Colonel chat interface
β β β βββ ColonelStatus.jsx # Colonel health dashboard
β β β βββ ColonelOnboarding.jsx
β β βββ Dashboard.jsx
β β βββ UserManagement.jsx
β β βββ AppsMarketplace.jsx
β β βββ subscription/ # Self-service billing
β β βββ organization/ # Org management
β βββ components/colonel/ # Colonel UI components
β βββ hooks/ # useColonelWebSocket, etc.
β βββ contexts/ # React contexts
β
βββ public/ # Static assets + logos
βββ docker-compose.direct.yml # Docker configuration
βββ install.sh # Bare-metal installer
βββ package.json # Frontend dependencies
βββ .env.example # Configuration template
Security
| Layer | Implementation |
|---|---|
| Authentication | Keycloak SSO with Google, GitHub, Microsoft; custom themed login, error, password reset, profile, and IDP linking pages |
| Authorization | 5-tier role hierarchy (admin, moderator, developer, analyst, viewer) |
| API Keys | bcrypt hashing, secure storage |
| Sessions | Redis-backed with configurable TTL |
| Input Validation | Pydantic models throughout |
| SQL Protection | Parameterized queries via asyncpg |
| XSS Protection | React built-in escaping |
| HTTPS/TLS | Traefik with Let's Encrypt auto-renewal |
| PCI Compliance | Stripe Elements (no raw card data touches your server) |
| Colonel Safety | Human-in-the-loop confirmation for destructive operations |
| Audit Trail | Immutable log of all operations and Colonel actions |
Ecosystem
Ops-Center is the control plane for a full AI infrastructure stack:
| Service | Role | Integration |
|---|---|---|
| Unicorn Brigade | AI agent platform (47+ agents) | Shared SSO, LLM routing |
| Open-WebUI | AI chat interface | SSO, credit billing |
| Center-Deep | AI metasearch (70+ engines) | SSO, cross-domain auth |
| Bolt.diy | AI dev environment | Curated model lists |
| Presenton | AI presentations | Image generation API |
| Unicorn Orator | Text-to-Speech service | SSO, credit billing |
| Unicorn Amanuensis | Speech-to-Text service | SSO, credit billing |
| Forgejo | Self-hosted Git | SSO, tier-based access |
| Keycloak | Identity provider (Google, GitHub, MS) | SSO backbone |
| Lago + Stripe | Billing engine | Metering, invoicing, payments |
| LiteLLM | LLM proxy | 100+ model routing, BYOK |
| Prometheus | Metrics collection | /metrics endpoint |
| Grafana | Observability dashboards | Metrics visualization |
| Umami | Web analytics | Privacy-focused tracking |
| Traefik | Reverse proxy + web hosting | SSL/TLS, Let's Encrypt |
| Cloudflare | CDN + DDoS protection | DNS, WAF, edge cache |
| Docker | Container orchestration | All services containerized |
Documentation
| Document | Description |
|---|---|
| CLAUDE.md | Complete technical reference (production context) |
| API Reference | All 624+ REST endpoints |
| Admin Handbook | Operations guide |
| Deployment Guide | Production deployment |
| Integration Guide | Connect your apps |
| Troubleshooting | Common issues and fixes |
| Architecture | System design diagrams |
| Contributing | How to contribute |
| Security Policy | Vulnerability reporting |
| Roadmap | What's coming next |
Contributing
We welcome contributions! See CONTRIBUTING.md for guidelines.
git checkout -b feature/amazing-feature
npm test && cd backend && pytest
git commit -m 'feat: add amazing feature'
git push origin feature/amazing-feature
# Open a Pull Request
We use Conventional Commits: feat, fix, docs, refactor, test, chore
Star History
License
MIT License β see LICENSE for details.
Copyright (c) 2025-2026 Magic Unicorn Unconventional Technology & Stuff Inc
Acknowledgments
Built with these excellent open-source projects:
FastAPI Β· React Β· Material-UI Β· Keycloak Β· Lago Β· LiteLLM Β· Traefik Β· Prometheus Β· Grafana Β· Umami Β· Forgejo Β· Vite Β· Chart.js Β· Kuzu Β· Ollama
If Ops-Center helps you, consider supporting development:
GitHub Β· Issues Β· Discussions
Built with care by Magic Unicorn Tech