Overview

The On-Prem Workstation offering is designed to enable secure deployment of Magnus within your organization's private environment. This version of Magnus runs on consumer-grade hardware, providing an accessible on-premises solution tailored to meet localized design needs.

The on-premises workstation delivers the core functionalities of Magnus while maintaining data privacy and supporting local deployment. However, due to the inherent limitations of consumer hardware, users may encounter several restrictions compared to the cloud-based version.

Networking

PrimisAI is providing a physical server which runs the Magnus software in an offline environment. Magnus starts automatically on server startup and is running on port 8080. The port can be changed in the docker-compose.yml file in the ~/magnus directory. The customer is responsible for adding the server to their existing network and configuring a hostname/URL/IP address and associated firewall rules for the server. Please refer to the Getting Started section for setting the client up with this URL/IP address.

RAG

Warning

Adding an excessive volume of files to the knowledge base may slow performance or degrade performance by providing irrelevant information to the AI models. If the customer chooses to add additional files to the PrimisAI supplied knowledge base, they should be chosen carefully for relevance to the workload and placed in their own subdirectory to ease administration and allow an easy return to factory configuration.

Info

The IP catalog’s default path is set to ~/magnus/rapidgpt-api/data/xilinx. You can add or remove IP cores as needed. However, preserving the original directory structure is strongly recommended.

Magnus includes a Retrieval Augmented Generation (RAG) capability to improve LLM performance and provide domain-specific knowledge. The system is pre-loaded with a PrimisAI supplied knowledge base. Additional customer-specific files may also be added by the system administrator. The files for the knowledge base are stored in the ~/magnus/rapidgpt-api/data directory and the database is rebuilt when the system is restarted. To rebuild the knowledge base after modifying the contents, run sudo docker compose down && docker compose up –d. The service will be available after the RAG database rebuilds, approximately ten minutes using the PrimisAI supplied knowledge base.

Update Policy

We are committed to continuously improving Magnus. Accordingly, we will update this page as new bugs and limitations are identified. Users are encouraged to review this document periodically for the latest information.

For any questions or to report additional issues, please contact our support team at support@primis.ai.