Skip to main content

NovaPod CLI (np) Reference

The NovaPod CLI (np) is a lightweight command-line tool designed to manage GPU instances, upload code, stream remote logs, and orchestrate clusters directly from your local terminal.

Installation & Setup​

curl -fsSL https://cli.novapod.io/install.sh | bash

Authenticating​

Generate an API key in your Console and run:

np setup --api-key np_live_your_secret_key

This saves your credentials securely in ~/.novapod/config.json.

Core Commands​

Query available GPU offers, spot prices, and reliability scores across the network:

# Search for single-GPU H100 offers sorted by lowest price
np search --gpu H100_SXM --count 1 --sort price

# Search for high-VRAM machines with at least 80GB VRAM and >99% uptime
np search --min-vram 80 --reliability 0.99

2. np launch​

Spin up a new GPU container pod or Root KVM VM:

np launch \
  --gpu RTX_4090 \
  --count 1 \
  --template pytorch \
  --disk 100GB \
  --name my-experiment-01

Common Flags:

  • --gpu <model>: e.g. H100_SXM, H200, B200, RTX_4090, MI300X.
  • --count <n>: Number of GPUs (1, 2, 4, 8).
  • --template <id>: pytorch, vllm, comfyui, ollama, unsloth.
  • --vm: Launch as a full Root KVM VM instead of a Docker container.
  • --disk <size>: NVMe allocation (e.g. 100GB, 500GB).
  • --vpc <id>: Attach to a private VPC.

3. np run​

Upload a local project, execute a command on a fresh GPU instance, stream stdout/stderr back to your local terminal, and automatically pause the instance when complete:

np run train.py \
  --gpu H100_SXM \
  --requirements requirements.txt \
  --auto-pause

4. np ssh​

Open an interactive SSH session into a running instance:

np ssh <instance_id>

5. np pause & np resume​

Pause instances to stop compute billing while keeping storage intact:

# Pause instance
np pause <instance_id>

# Resume instance
np resume <instance_id>

6. np destroy​

Permanently tear down an instance and eliminate all storage fees:

np destroy <instance_id>