Deployment Guide

Hosting
Options

Personal Hosting

Local Device

Run Ona on your own machine for maximum privacy and direct local file access.

Personal Server

Use a VPS or home server for always-on personal runtime and multi-device access.

Hybrid

Keep sensitive context local and route selected heavy workloads to cloud compute.

Family Hosting

  • Host on a Raspberry Pi, NAS, or an old gaming computer for low-cost shared usage.
  • Set role-based permissions for adults, guests, and child-safe automations.
  • Use backup and health-check routines to keep household operations reliable.

Company Hosting

  • Deploy on cloud infrastructure, local servers, or a hybrid model.
  • Use Sphere governance for shared files, shared models, approvals, and audit trails.
  • Standardize observability, incident response, and rollout policies from day one.

AI Model Hosting

Ona can run cloud-hosted models, self-hosted local models, or hybrid routing across both. This lets you tune privacy, cost, and reasoning depth per mission.

Supported Backends

  • Cloud APIs: Anthropic, OpenAI, Groq, Gemini, OpenRouter.
  • Local model server: Ollama.
  • Docker-hosted model services (self-managed containers).
  • OpenAI-compatible: llama.cpp, LM Studio, vLLM.

Routing Modes

  • Cloud-only for low-maintenance operation.
  • Self-hosted only for maximum local control.
  • Hybrid for local-first privacy with cloud escalation.

Recommended Pattern

  • Keep Solin and core workflows on local models.
  • Escalate deep research and online tasks to larger cloud models.
  • Store accepted escalated results in memory with metadata.

For full model architecture details, role routing, dual-bot separation, and production validation, see the dedicated Models page.

Open Models Guide