Why on-premise only

Your documents stay where they belong.

We thought hard about offering a cloud option. We chose not to — on purpose. If your DMS is the backup of your business, it doesn't belong in someone else's data center.

No data leakage.

No LLM request leaves your network. We ship the models — local, quantized, GPU-optional.

Air-gapped ready.

Install without internet. Updates as signed tarballs. Attestation check verifies the origin.

Docker Compose or Kubernetes — you choose.

Standard deployment is Docker Compose: one command, one Docker host, done. For HA setups and larger scale we ship a Helm chart. No 40-container wiring marathon either way.

What you actually need.

A Docker host — or a Kubernetes cluster.
For smaller sites a Docker host with Docker Compose is enough — one command, one system, done. Need more scale? K3s, RKE2, OpenShift, EKS-on-prem — doesn't matter. We ship a Helm chart that runs anywhere.
Postgres 16 with pgvector.
From us or your own. HA setup with Patroni is in the chart if you want it.
Object storage.
MinIO, Ceph, NetApp StorageGRID — S3-compatible is enough. Content-addressable storage means duplicate documents hit disk only once.
GPU — optional but recommended.
For local LLM inference. One RTX 4090 handles a site with ~500 docs/day. Without a GPU, everything works except the fastest classification.

We set it up with your team.

Two days. One migration engineer from us, someone from your ops team. On day three, RecordTailor is running in your data center.

Request an install slot