Deployment
Four models, one codebase
The same application across all four, including airgapped.
Deployment models
- SaaS
- Multi-tenant cloud in a US or EU region. Most customers
- Hybrid
- Application in cloud, your LLM calls routed through your own VPC. Strict AI-boundary policies
- Private cloud
- Single-tenant deployment in your account, HIPAA BAA available. Regulated data, dedicated tenancy
- On-prem airgapped
- Docker and Helm, your own model serving, no egress. Sovereign or airgapped sites
What airgapped actually means
- Your models, your hardware
- vLLM, Ollama or TensorRT-LLM serving open-weight models you choose.
- No customer data egress
- Nothing leaves your network, prompts included.
- Regulatory intel still arrives
- Public data ships as a quarterly-updated bundled database.
- You control the update cadence
- Signed Helm charts delivered quarterly; you decide when to apply them.
Talk through your deployment model
Including fully airgapped, with your own model serving.