Deploy on Salad

Run Laboratory OS on Salad, a distributed GPU cloud that taps into idle gaming PCs and workstations worldwide. Salad offers GPU compute at up to 90% lower cost than traditional clouds.

1. Create a Container Group

2. Configure the Container

Image Source:

openlaboratoryorg/laboratory-os

Resource Class: Select a GPU tier.

Choosing a GPU: More VRAM lets you run larger models — 16 GB handles most image generation, 24-48 GB covers mid-size LLMs, and 100 GB+ is needed for large LLMs.

Environment Variables:

Key	Value
`UPLINK_API_KEY`	Your account key from openlaboratory.com. The lab connects through it and serves on your `*.laboratory.computer` URL.

No inbound ports or firewall rules are needed — Laboratory OS uses an outbound tunnel. You can block all inbound traffic entirely.

Note: Salad nodes are distributed — your container runs on community hardware. This is ideal for batch jobs and inference but may not suit interactive sessions that need consistent uptime.

3. Deploy and Connect

Click Deploy. Salad schedules your container across available nodes. Startup time varies — it can take a few minutes for a suitable node to become available.

Check the container logs for your assigned slug — it appears within the first 30 seconds. Then sign in to your account at:

https://app.laboratory.computer

From there you'll see your running instance and can open the Laboratory OS desktop. Access is gated by your SSO login rather than a shared token, so only authenticated account holders can connect.

4. Persistence

Salad containers are ephemeral. When a container stops or is rescheduled, the filesystem is lost. To persist data:

Store models and outputs in external storage (S3, Google Drive, etc.)
Download models on container startup via a custom entrypoint script

Tip: Salad’s distributed model means nodes can go offline unexpectedly. Best suited for workloads that can tolerate occasional restarts — like background rendering or batch inference.