Migrate to the new Outpost

Move a previous-generation Docker Outpost to the new Outpost application — guided by us, with a full re-index of your sources

The on-premise Outpost has been rebuilt from the ground up. If you still run a previous-generation Outpost on Docker, moving over is not an option but the way forward — and we do it together with you.

Public Docker distribution has ended

We no longer publish Docker images of the Outpost. A new Outpost can no longer be set up through Docker, and your existing installation receives no new tags — no updates, no new source types, no fixes.

That does not stop it working overnight: the version you are running today keeps running on the images already on Docker Hub. But it only gets older from here, and we no longer keep it current alongside meinGPT. You can schedule the move to the Outpost application, then — but not sit it out.

Get in touch at enterprise@meingpt.com and we will set up the new Outpost together with you — your data is re-indexed in the process.

What's new

New embedding models

Newer, considerably more efficient models — the reason a re-index is needed

New database, new indexer

The search index was rebuilt from scratch

Simpler setup

Connect with a pairing code instead of manually managed credentials, no Docker

No inbound exposure

Outbound Bridge connection — no open firewall ports, no public IP, no reverse proxy

Why a re-index is unavoidable

The new version is nine months of complete redevelopment: new and considerably more efficient embedding models, a new database, a new indexer.

An embedding model translates text into vectors, and every model does that in its own way. The vectors the old model wrote into the old store therefore cannot be carried over — held against the new index they simply mean nothing. They have to be recomputed. Re-indexing is not extra work we impose on you; it is precisely what the move consists of.

What cannot happen along the way: losing original data. An Outpost does not store original files — it holds a derived index of your own source systems (SharePoint, S3, local files, Confluence, and more). Your original data stays untouched in those systems, and the index is rebuilt from them.

How the move works

Get in touch

Write to enterprise@meingpt.com or talk to your contact. We go through the sources and services your current Outpost serves and plan the move around them. To help us start faster, tell us the number of connected sources, the approximate data volume, and whether more than one data pool is affected — we'll discuss the concrete timeline for your move with you directly from there.

Create the new Outpost

You create the new Outpost in Company Settings → Outposts. See Create Outpost for details.

Install and pair the Outpost

The new Outpost application is installed on your infrastructure — on Windows, macOS (Apple Silicon) or Linux — and connected with the pairing code. It opens the outbound Bridge connection on its own — no inbound firewall access required. The per-system requirements are under Set up an Outpost.

Reconnect the sources and re-index

The same sources you used before are reconnected (SharePoint, S3, local files, …) and indexed from scratch. See Data sources for the full list of supported sources.

Verify the results

We run a test search together to confirm your content is found again.

Decommission the old Outpost

Only once the new Outpost is connected and returning results does the old Docker installation go out of service.

Note

You do not have to do this on your own — and you should not. Write to enterprise@meingpt.com, we'll agree the setup package and a date and do the move together.

Was this page helpful?