SSovAIHub
Products
Phase 1: Offline Runtime DemoOpen Source

SovAI Air-Gap AI Starter

Prepare while connected. Run while disconnected. Consume only local artifacts.

View GitHub RepositoryRequest implementation help

Expected outcomes

Prepare approved artifacts while connected, then run the app without runtime internet
Demonstrate local document retrieval, approved agent tools, and audit logging
Show the core sovereign AI principle that runtime workloads consume only approved internal artifacts

What it includes

Connected preparation script
Offline bootstrap workflow
Local document RAG pattern
Approved tool demo
JSONL audit logging
Docker Desktop setup notes

What this kit proves

These are the practical claims the kit demonstrates when run locally.

Application runtime can work after internet is disconnected.
Runtime dependencies can be packaged before going offline.
Local documents can be retrieved without external services.
Agent tools can be restricted to an approved list.
Unapproved tools such as web search can be blocked.
Important actions can be written to audit logs.

Offline runtime flow

Step 1

Build the Docker application image while internet is available.

Step 2

Download required Python wheels into the local artifact hub.

Step 3

Save the built image as a local tar artifact.

Step 4

Disconnect internet and load the image from the local artifact store.

Step 5

Run local document retrieval, approved tools, and audit logging without runtime internet.

Quick start

Copy these commands into a terminal and follow the connected/offline steps described in the repository README.

git clone https://github.com/sovaihub-lab/sovai-airgap-ai-starter
cd sovai-airgap-ai-starter
chmod +x scripts/*.sh
./scripts/prepare-online.sh
# Disconnect internet
./scripts/bootstrap-offline.sh
open http://127.0.0.1:8080

Configuration

SOVAI_OFFLINE_MODE
true
APPROVED_TOOLS
calculator,document_search,ticket_classifier
DOCUMENTS_DIR
/app/data/documents
AUDIT_LOG_PATH
/app/data/audit/audit-log.jsonl

Ports and services

Service
Port
Purpose
SovAI app
8080
Offline runtime web UI and API

Troubleshooting notes

scikit-learn failed to build on Windows

The early version used compiled dependencies. The kit was simplified to pure-Python retrieval and classification so Windows users do not need a local compiler.

Container was running but browser could not reach it

Check docker compose ps, docker port sovai-airgap-app, and docker logs. Use http://127.0.0.1:8080 instead of localhost when local DNS behaves differently.

Docker network was too restrictive

For the laptop demo, use a normal bridge network. Avoid internal: true until the runtime and host access pattern are working.

Limitations

This phase is an offline runtime proof, not a full production TL4 platform.
The first phase does not use a local LLM for answer generation.
The demo focuses on runtime behavior, not enterprise scanning, signing, or RBAC.

Recommended next steps

Move to Phase 2

Add Ollama and llama3.2:1b to turn local document retrieval into local LLM RAG.

Move to Phase 3

Introduce an internal artifact hub so new apps can be built from approved images and wheels.