Air-Gapped Deployment

Deploy ReductrAI with zero network egress. For financial services, healthcare, government, and defense.

Zero data leaves your network. Raw telemetry stays local. AI runs locally. No cloud connection required.

Architecture Overview

+-----------------------------------------------------------------+
|                    YOUR INFRASTRUCTURE                          |
|                                                                 |
|  +------------------+     +------------------+                  |
|  |   ReductrAI      |     |   Local LLM      |                  |
|  |   Binary         |---->|   (Ollama/vLLM)  |                  |
|  |                  |     |                  |                  |
|  |  - Ingestion     |     |  - Runs locally  |                  |
|  |  - DuckDB        |     |  - CPU or GPU    |                  |
|  |  - Detection     |     |  - Any OpenAI-   |                  |
|  |  - Dashboard     |     |    compatible    |                  |
|  +------------------+     +------------------+                  |
|           |                                                     |
|           |  License validated locally (signed JWT)             |
|           |                                                     |
|  +--------v---------+                                           |
|  |   Your Apps      |                                           |
|  |   (telemetry)    |                                           |
|  +------------------+                                           |
|                                                                 |
|  ============================================================   |
|              NO OUTBOUND NETWORK CONNECTIONS                    |
+-----------------------------------------------------------------+

Requirements

ReductrAI binary (download once, copy to air-gapped network)
Local LLM runtime - Ollama, vLLM, LocalAI, or any OpenAI-compatible server
An open-weight model file (e.g. Llama 3.2, Mistral 7B, Qwen 2.5)
Enterprise license (contact sales@reductrai.com)
8 GB+ RAM for typical 7B models (less for smaller quantized variants)

Step 1: Obtain License

Contact sales@reductrai.com for an enterprise license. For air-gapped networks, we provide offline license validation.

Step 2: Download a Local LLM & Model

On an internet-connected machine, install Ollama and pull a model:

        # Install Ollama

        curl -fsSL https://ollama.com/install.sh | sh

        # Pull a model (choose based on your hardware)

        ollama pull llama3.2

        # Models are stored in ~/.ollama/models/

Transfer to Air-Gapped Network

        # On internet-connected machine:

        tar -czf reductrai-bundle.tar.gz \

          ~/.ollama /usr/local/bin/reductrai /usr/local/bin/ollama

        # Transfer via USB/approved media to air-gapped machine

        # On air-gapped machine:

        tar -xzf reductrai-bundle.tar.gz -C /

Step 3: Install ReductrAI Binary

Download the binary on an internet-connected machine, then transfer to air-gapped network:

        # Download binary

        wget https://github.com/reductrai/reductrai/releases/latest/download/reductrai-linux-amd64

        # Transfer to air-gapped machine and install

        chmod +x reductrai-linux-amd64

        mv reductrai-linux-amd64 /usr/local/bin/reductrai

Step 4: Configure ReductrAI

Set environment variables for air-gapped operation:

        # License key

        export REDUCTRAI_LICENSE=your-license-key

        # Point at your local LLM

        export REDUCTRAI_LLM_ENDPOINT=http://localhost:11434/v1

        export REDUCTRAI_LLM_MODEL=llama3.2

        # Standard settings

        export REDUCTRAI_PORT=8080

        export REDUCTRAI_DATA_DIR=~/.reductrai

As long as REDUCTRAI_LLM_ENDPOINT points at an internal address, no traffic leaves your network.

Step 5: Start ReductrAI

        # Start Ollama (or your chosen LLM runtime) first

        ollama serve &

        # Start ReductrAI

        reductrai start --local

        # Verify

        reductrai status

Step 6: Access Dashboard

The dashboard runs locally on the ReductrAI binary:

http://localhost:8082

In air-gapped mode, the Copilot dashboard is served directly from the binary on port 8082. No external cloud connection.

Recommended Models for Air-Gapped

Model	Size	RAM Required	Notes
Llama 3.2 8B (Recommended)	4.7 GB	8 GB	Strong reasoning, open weights, runs well on CPU
Qwen 2.5 7B	4.4 GB	8 GB	Excellent at structured output / tool use
Mistral 7B Instruct	4.1 GB	8 GB	Fast, permissive license
Llama 3.2 3B (quantized)	2.0 GB	4 GB	Smallest footprint; acceptable for summarization

Security Considerations

Binary is obfuscated (garble) to protect proprietary code
License file uses cryptographic signatures - cannot be forged
All data stays in ~/.reductrai - audit and control access
Dashboard can be restricted to localhost or internal network
No telemetry or analytics sent anywhere

License Renewal

Enterprise licenses are issued for a fixed term (typically 1 year). Before expiration, contact sales@reductrai.com for renewal.

Troubleshooting

License validation failed: Verify your license key is correct. Contact sales@reductrai.com for enterprise air-gapped licensing.

LLM not reachable: Confirm your local LLM (Ollama, vLLM, etc.) is running and listening on the address in REDUCTRAI_LLM_ENDPOINT. Test with curl $REDUCTRAI_LLM_ENDPOINT/models.

Need Help?

For enterprise and air-gapped deployments, contact our team:

Email: sales@reductrai.com
Security: security@reductrai.com