On-Premise Deployment

Host BriqMind infrastructure in your own data center. Keep 100% control of your data while running our high-performance AI models inside your internal network, including air-gapped environments.

01 Isolation 02 Hardware 03 Deployment 04 Monitoring

01Enterprise Isolation

In highly regulated sectors such as finance, healthcare, defense, and the public sector, preventing data from leaving the organization is a critical requirement. BriqMind's on-premise architecture lets models run fully on your own servers. No data packet is sent to the outside world or the public internet.

Air-Gapped Compatibility

The system can run fully on a closed local network, or intranet, without external internet access.

Data Sovereignty

Training, fine-tuning, and inference processes are kept entirely on internal enterprise servers.

02Hardware Requirements

AI models require specialized hardware to run efficiently. The requirements below are prepared for a 50 concurrent-user scenario using Birk-Agent-Heavy as the reference model.

GPU

Minimum 2x NVIDIA A100 (80GB) or 4x L40S. Model weights and KV cache are kept in VRAM.

Critical

CPU & RAM

Minimum 32 cores, for example AMD EPYC or Intel Xeon, and at least 256GB ECC DDR4/DDR5 system memory.

Recommended

Storage

At least 2TB NVMe SSD. Required to improve model loading speed and for database logs such as PostgreSQL / Vector DB.

Required

03Architecture and Deployment Process

Docker & K8s

To simplify deployment, the system is fully containerized. For setup, you need to pull images from the secure Docker Registry we provide, or transfer .tar files to the server for air-gapped systems.

Before starting installation, make sure docker, docker-compose, and NVIDIA drivers, including NVIDIA Container Toolkit, run correctly on your server.

Step 1: Log In to the Registry and Pull Images

Terminal

$ docker login registry.briqmind.com -u ENTERPRISE_USER
> Password (Token): ******************
> Login Succeeded
$ docker-compose pull
> Pulling briq-api-gateway ... done
> Pulling briq-model-inference ... done
> Pulling briq-vector-db ... done

Step 2: Configure and Start

After editing your .env file according to your organization's network settings and license key, start the services.

Terminal

$ docker-compose up -d
> Starting briq-vector-db ... done
> Starting briq-postgres ... done
> Starting briq-model-inference ... done
> Starting briq-api-gateway ... done
$ docker logs -f briq-api-gateway
> [INFO] BRIQ Enterprise System Ready.
> [INFO] Listening on 0.0.0.0:8000
> [INFO] Models loaded into VRAM successfully.

04Monitoring and System Management

After the system is running, you can monitor hardware consumption, model latency, and user logs from a central panel.

Grafana & Prometheus

Ready-made Grafana dashboards are provided for GPU memory usage, tokens-per-second statistics, and real-time API response times.

SIEM Integration

All admin and user logs are forwarded to your organization's central SIEM system, such as Splunk or QRadar, through Syslog or Webhook.