Skip to main content

​
Overview

On-premise deployment gives you full infrastructure sovereignty. Enclave runs on your servers, connects to your HSM, and stores data in your storage systems. No traffic leaves your network perimeter.

​
System requirements

The content-inspection (DLP) subsystem — Microsoft Presidio (NLP) + Apache Tika (JVM) — sets the memory floor: 16 GB is the minimum with DLP enabled. For millions of files / thousands of users, size to 16 vCPU / 64 GB / NVMe. See the private appliance sizing profiles for the full tier table; the appliance auto-tunes to the VM it’s given, while a hand-rolled on-premise Compose deployment must set container memory limits itself.

​
Components

​
Installation

Enclave ships as a Docker Compose stack. You receive a private registry token when you purchase an on-premise licence.

​
Environment configuration

Key variables in .env:

​
Object storage CORS

Enclave’s web client encrypts files in the browser and uploads ciphertext directly to your bucket via pre-signed URLs — plaintext never transits the application server. For this to work, your bucket must allow PUT requests from the Enclave web origin(s). Apply a CORS configuration like the following, replacing https://enclave.yourcompany.com with your deployment’s domain:
Via the AWS console:
1

Open the bucket

S3 → click into your bucket
2

Go to Permissions

Open the Permissions tab
3

Edit CORS

Scroll to Cross-origin resource sharing (CORS) and click Edit
4

Paste the rules and save

The console expects the rules array directly (no CORSRules wrapper):
Click Save changes. The policy takes effect immediately.
Only add http://localhost:3000 to AllowedOrigins when developing against this bucket from a local frontend build. Production deployments do not need it.
ExposeHeaders: ETag is required for multipart uploads — the client reads the ETag returned for each uploaded part. Without a matching CORS rule, browser uploads fail with a CORS policy error on the pre-signed PUT and the file remains stuck in pending/scanning until the abandoned-upload cleanup worker removes it.

​
Updates

Database migrations run automatically on startup. Back up PostgreSQL before each update.

​
High availability

For HA deployments, run multiple application instances behind a load balancer. The application is stateless — all state is in PostgreSQL and object storage. Use PostgreSQL streaming replication for database HA. Contact hello@kvelden.com for the HA architecture guide.