Skip to content
Copyright © 2022-2026 basebox GmbH, all rights reserved.
Licensed to be used in conjunction with basebox, only.

// installation

Installation & Operations

Applies to

Product: Cloud · Server · Audience: Platform Operator · Administrator

This section answers one question: I have this infrastructure – how do I install and operate basebox on it correctly?

It is written for Platform Operators and DevOps. It is not a hardware buying guide; sizing material exists elsewhere. And it contains no API tutorials – those are under Developer → API.

Start here

  1. Understand the architecture – basebox platform, service models and inference are three separate layers with very different resource needs.
  2. Deployment models – Demo, Cloud, Server, and why a customer server hosted at basebox is still Server.
  3. basebox components – the services an installation consists of and how they talk to each other.
  4. Responsibilities – who does what in self-managed vs. basebox-operated setups.

Then choose your product.

basebox Cloud

basebox operates the infrastructure; you administer the application. The Cloud part of this section describes what your environment contains, how users and your systems reach it and how you go live: basebox Cloud.

basebox Server

A dedicated server for your organisation – at your site or at basebox, operated by you or by basebox. The Server part is the main part of this section:

Section Content
Overview Hardware, hosting, operating and installation-path decisions
Architecture Platform, service models, inference, resources & scaling, deployment topologies
Reference configurations Concrete, known hardware setups with status Validated / Supported / Experimental / Custom
Bare-metal installation The end-to-end path from an empty server to a validated installation
Helm Chart reference, quick start, production installation, connectors, services
Models & inference Backends, configuration, tested and recommended models, benchmarks
Operations Monitoring, logging, backup, updates, scaling, troubleshooting
Advanced architectures Multi-GPU, dedicated service GPU, multiple inference instances, multi-node

Three things to know before anything else

basebox itself needs little GPU. A server's GPUs serve the service models (RAG, OCR, speech-to-text) and inference (the language model). A sentence like "basebox needs two H200s" describes inference, not the platform.

A server at basebox is Server, not Cloud. Hardware source, location and operation are properties of a server, not products. This documentation describes every server deployment along exactly these dimensions.

Buying the software does not include operations. Who operates the server – you or basebox – is a separate agreement. The table: Responsibilities.