Platform Overview
Aethyr is a sovereign AI orchestration platform designed for organizations that demand complete control over their AI infrastructure, data, and deployment models.
Three-Tier Architecture
Aethyr is built on a flexible three-tier architecture that allows you to deploy AI capabilities at any level—from edge devices to cloud infrastructure—while maintaining data sovereignty and operational control.
Edge Tier
Local inference on user devices—laptops, desktops, or embedded systems. Perfect for air-gapped environments and real-time applications.
- • Sub-10ms inference latency
- • Full offline capability
- • Data never leaves the device
- • Supports GGUF/ONNX models via Ollama
Node Tier
On-premises or private cloud deployment for team and organization-wide AI services. Handles multi-agent orchestration and RAG pipelines.
- • Multi-agent coordination
- • RAG with pgvector semantic search
- • Self-hosted model serving
- • Organizational data sovereignty
Core Tier (Optional)
Cloud connectivity for premium API access, global synchronization, and managed services. Fully optional—air-gapped deployments work without it.
- • Premium API gateway (OpenAI, Anthropic, Google)
- • Global sync and state management
- • Centralized model registry
- • Managed updates and support
Key Concepts
Agents
AI agents are configurable entities with specific models, system prompts, and tool access. Multiple agents can collaborate on complex tasks through orchestration.
Conversations
Persistent chat threads that maintain context across messages. Conversations can involve one or more agents working together.
MCP Tools
Model Context Protocol integrations that extend agent capabilities—file access, database queries, API calls, and domain-specific tools.
Knowledge Bases
RAG-powered document stores for semantic search. Upload documents and agents can retrieve relevant context automatically.
Organizations
Multi-tenant isolation for teams and companies. Each organization has separate agents, data, and access controls.
Audit Logs
Comprehensive logging of all operations for compliance and debugging. Track every model invocation, tool use, and data access.
Deployment Models
Aethyr supports multiple deployment models to match your security requirements and operational constraints.
| Model | Best For | Network Required |
|---|---|---|
| Cloud Managed | Teams wanting managed infrastructure | Always connected |
| Hybrid | On-prem data with cloud AI | Intermittent OK |
| On-Premises | Full data sovereignty | Optional |
| Air-Gapped | Classified/secure environments | Never |