Choose your path

Two ways to run your own Data Vault — Desktop app or Docker. Pick the one that fits you.

There are two ways to run your own Data Vault. Both connect to meinGPT and keep your data in your environment — they differ only in effort and control.

In short

Just want it running on a workstation? → Desktop app. Running on your own Linux server, or need full control? → Docker.

The two paths at a glance

Desktop appDocker / Linux server
For whomAdmins without deep ops knowledgeSelf-hosting experts
SetupGuided wizard, pairing codeDocker Compose, config files
Runs onWindows workstationYour own server (Linux)
EmbeddingsOffline model built inYour choice (Ollama, OpenAI, …)
TunnelBuilt in (Go Bridge agent)Built-in tunnel sidecar
UpdatesAutomatic (signed)Manual (docker compose pull)

The Vault Desktop app is the simplest way. You download an installer, install it on a workstation, and pair the vault with a pairing code from meinGPT — no config files, no command line.

  • Guided setup — a wizard walks you through pairing step by step.
  • Embeddings built in — an offline, German-optimized model ships with the app. No GPU, no Ollama, no setup required.
  • Tunnel built in — the app spawns the bundled Go Bridge agent itself and connects the vault to meinGPT without opening any ports.
  • Automatic updates — the app updates itself via signed packages.

Download coming soon

The signed installer will be published at vault-download.meingpt.com. The download is not public yet — reach out to enterprise@meingpt.com or check the release notes to get the link as soon as it is available.

Path 2 — Docker / Linux server (self-hosting)

The Docker path is for teams that run the vault on their own infrastructure and want full control over every building block. You run the services with Docker Compose on a Linux server and pick your embedding provider yourself.

  • Full control — every service, resource limit, and embedding model is configurable.
  • Runs server-side — ideal when the vault must run continuously, independent of any workstation.
  • Your choice of embeddings — from the built-in Ollama (bge-m3) to your own OpenAI-compatible endpoint.

Hinweis

In the default setup you need neither path: you can use data pools in meinGPT directly, without operating your own vault.

Was this page helpful?