System Requirements

NCP offers three deployment options based on where the LLM endpoint runs and who manages it. NCP compute always runs on-premises, only the LLM endpoint location varies.

Option 1: Everything On-Premises

Both NCP compute and the LLM run inside your data center on your own GPU server.

• You provide: GPU server, NCP compute host.

• Next step: See On-Prem Deployment for hardware and GPU requirements.

Option 2: Compute On-Premises, LLM EP in Cloud

NCP compute runs on-premises. The LLM endpoint runs in the cloud and is managed by Aviz, you don't procure GPUs or manage the LLM lifecycle.

• You provide: NCP compute host.

• Aviz provides: Managed LLM endpoint and credentials.

• Next step: Contact Aviz for LLM endpoint credentials, then proceed with NCP installation.

Option 3: Bring Your Own LLM Endpoint

NCP compute runs on-premises. You provide the LLM endpoint, either a private LLM on your own cloud GPU instance (AWS, GCP, Azure) or a public LLM-as-a-Service (Gemini, Claude, ChatGPT, pay-per-token).

• You provide: NCP compute host and the LLM endpoint (private cloud instance or public API key).

• Next step: Onboard the LLM endpoint in NCP.