1. Offline
Disconnect internet — assistant still responds from local model + RAG index.
Illustrates the Sovra value proposition for enterprise buyers. Phase 2 connects live on-device inference via FastAPI.
Static demonstration. Live inference connects in Phase 2 (FastAPI + Ollama on target hardware).
Configure your deploymentDisconnect internet — assistant still responds from local model + RAG index.
Ask about manuals or internal docs — accurate answers with source citations.
Natural language maps to validated JSON intents — never free-form dangerous actions.
Sub-second feel matters more than model size for buyer confidence.
We will map demo scenarios to your hardware tier and compliance requirements.