This is relevant ONLY to those deciding to go fully "local" - ie not choosing a private cloud or Managed Services option of deployment. Please check out our summary of deployment options to learn more. The requirements are divided into 2 Sections:

• Flows Requirements

• Threads Requirements

• Whisper Requirements

Flows Requirements

Requirements for a client-owned data center installation of Flows:

• Linux computing resources

  • These can be virtual machines or bare metal hardware.

  • Linux OS - Ubuntu / Debian. We provide support and documentation for Ubuntu specifically

  • The Linux kernel must be version 6.5.x or newer

  • Total RAM and CPU are usage dependent, 6 virtual CPUs at ~2.0Ghz and 64 Gb total RAM minimum

    
    

• Network-attached storage

  • Storage size and IOPS requirements are usage dependent. We recommend 80Gb per node as a minimum

  • Backups of the stored files and database must be handled by the customer

    
    

• Permit outgoing HTTPS connections to Authentise

  • These connections allow us to bill the installation on a per-transaction basis as well as send emails/SMS notifications using our infrastructure

  • Standard SSL over port 443 at periodic intervals, no customer-owned data is transmitted.

    
    

• Incoming HTTPS requests over intranet

  • All services are exposed via HTTPS, so client software must be able to reach the servers over port 443 internally.

    
    

If the customer has a particular database provider requirement, Flows can be integrated with an existing postgresql (14.x and above) database server and license. Otherwise Flows will allocate some of the NAS and compute resources to running custom postgres on Linux for all of it’s RDBS needs.

If you're intending to use Machine Data as part of your application installation, then you will need Echo. Echo is described here, and needs minimal hardware: a very simple windows computer should do the trick. The only important mark is that this computer must sit on the same network as your printer. For certain printers you may need to install it directly on the PC controlling your printer.

Additionally if the deployment will be using slicing or design streaming to Stratasys printers:

• Windows computing resources

  • Virtual machines or bare metal hardware

  • Windows Server 2012 or Windows 10

  • 8 Gb RAM, 1 2.0Ghz CPU minimum

    
    

Flows is architected as a system of microservices that are deployed via containers. This allows for flexibility in allocate computing resources and easy scalability. All deployments of Flows come with a dashboard for monitoring resource utilization to make informed decisions about adding additional capacity to the cluster.

This architecture diagram might help:

Threads Requirements

The technical requirements to host our Threads application locally are as follows: 

• Threads Primary Server

  • Minimum: 4 vCPU, 16GB RAM

    
    

• Threads DB

  • Minimum: 2 vCPU, 8GB RAM

    
    

• AI Service 

  • Minimum: 8 vCPU, 32GB RAM

  • GPU: 1+ Nvidia hardware needs discussion/trial

    
    

• 3D Viewer 

  • Server: 4 vCPU, 8GB RAM

  • Database: 2 vCPU, 4GB RAM

    
    

• Storage:

  • Shared between all instances

  • 700GB-1TB

  • 200+ GB is taken up by AI

Whisper Requirements

Whisper is deployed as a set of containerized services via Docker Compose on a single Linux host.

Linux OS

Ubuntu / Debian. The Linux kernel must be version 5.10 or newer. Docker Engine 24+ and Docker Compose v2 are required.

Whisper Application Server

• Whisper Primary Server Minimum: 2 vCPU, 4GB RAM; Storage minimum: 80gb

• Whisper DB Minimum: 2 vCPU, 4GB RAM; Storage minimum: 80gb

• Whisper Vector DB Minimum: 2 vCPU, 4GB RAM; Storage minimum: 80gb

Storage

Minimum: 50 GB

PostgreSQL data, Weaviate vector index, and Redis persistence share the host filesystem. Storage requirements grow with the volume of ingested documents. Backups of PostgreSQL and Weaviate data directories must be handled by the customer.

LLM Provider (external)

Whisper requires access to an external LLM for answer generation. This can be hosted locally, We currently utilize these models:

• OpenAI API (standard), or

• Azure OpenAI (customer-managed deployment)

The embedding model (sentence-transformers/all-MiniLM-L6-v2) runs locally on CPU. No GPU hardware is needed.

Network — Outgoing HTTPS (if required)

Aside from the model, Whisper requires outgoing HTTPS (port 443) to the services you choose to integrate. At present, these are some examples:

No customer-owned document content is transmitted to Authentise. LLM calls send document chunks and user queries to the configured LLM provider only.

Network — Incoming HTTPS

All user access is via HTTPS on a single port (default: 3000). Caddy handles TLS termination. Internal service-to-service communication stays within the Docker network and is not exposed externally.

Integration credentials

Credentials for external integrations (GitHub, Slack, Shortcut, Threads) are entered by the customer in the Whisper UI and stored encrypted in the local PostgreSQL database. No integration secrets are stored in configuration files or transmitted to Authentise.