🚧Documentation is under active development — content may be incomplete or change frequently.
Skip to Content
Cloud Computing

Cloud Computing

Run GPU-intensive workloads on cloud hardware without managing infrastructure. Salpa’s cloud nodes give you access to NVIDIA A100 and H100 GPUs through a simple, pay-per-use model.

How it works

Cloud nodes behave exactly like local nodes — they appear in the Node Library, connect in workflows, and execute with a single click. The difference is that computation happens on remote GPUs instead of your local machine.

Your Desktop App
  └→ Cloud node starts execution
       └→ Request sent to API Gateway (Google Cloud)
            └→ Routed to GPU compute (Modal)
                 └→ Result returned to your workflow

Available cloud nodes

NodeGPUUse CaseTypical Runtime
ESMFoldA100 (40GB)Single-chain protein structure prediction from amino acid sequence30-120 seconds
Boltz-2H100 (80GB)Multi-chain biomolecular structure prediction (proteins, RNA, ligands)2-10 minutes

Getting started with cloud

1. Sign in

Click the Cloud icon in the sidebar or go to Settings → Cloud. Sign in with your Google account via Firebase Authentication.

2. Browse cloud nodes

Cloud nodes appear in the Node Library with a cloud badge. They’re organized under the “Cloud” category.

3. Run a cloud node

  1. Drag a cloud node (e.g., “ESMFold”) onto the canvas
  2. Configure it — paste an amino acid sequence or select an input file
  3. Click Execute — the job is sent to the cloud
  4. Monitor progress — real-time status updates stream back via WebSocket
  5. View results — output files (PDB structures, confidence scores) appear in your workflow

Authentication

Cloud computing requires a Salpa account:

  • Sign in with Google — One-click authentication via Firebase
  • JWT tokens — Automatically managed by the app (7-day expiration)
  • Secure — Tokens are sent via X-Cloud-Auth-Token header, never stored in plain text

Pricing model

Salpa uses a credit-based pay-per-use model:

  • No cloud accounts needed — You don’t need AWS, GCP, or Modal accounts
  • Credits — Purchase Salpa credits and spend them on cloud node executions
  • Transparent pricing — Cost per execution is shown before you run
  • Quotas — Plan-based limits to prevent unexpected charges

Infrastructure

Salpa’s cloud infrastructure runs on Google Cloud Platform:

ComponentServicePurpose
API GatewayGoogle Cloud RunAuthentication, routing, quota enforcement
User DatabaseFirestoreProfiles, usage logs, credit balances
AuthenticationFirebase AuthGoogle sign-in, JWT management
GPU ComputeModalESMFold (A100), Boltz-2 (H100)
CPU ComputeCloud RunLightweight cloud tasks

Limitations

  • Internet required — Cloud nodes need a network connection
  • Queue times — GPU instances may take 10-30 seconds to cold-start
  • File size limits — Large input files may increase upload time
  • Timeout limits — ESMFold: 600s, Boltz-2: 1800s per execution

Next steps

Build Custom Nodes