CLI Tool | RSP Cloud

Command Line
Interface

The rsp command-line tool is the fastest way to interact with RSP Cloud. Spin up B200 clusters, manage NVMe storage, and monitor your AI training directly from your terminal.

1. Installation

Install the RSP CLI on Linux, macOS, or Windows Subsystem for Linux (WSL). Sab dependencies automatically install ho jayengi.

BASH 
# Yahan hum curl ka use karke official script run kar rahe hain
$ curl -sL https://cli.rsplatforms.com/install.sh | bash

Downloading rsp-cli-v2.4.1-linux-amd64...
[████████████████████] 100%
Extracting binary to /usr/local/bin/rsp...
✓ Installation complete.

# Verify the installation
$ rsp --version
rsp version 2.4.1 (build 8a9b7c)

2. Authentication

Apne RSP Cloud account ko terminal se connect karne ke liye CLI ko authenticate karein. Aap browser ke through ya API key se login kar sakte hain.

BASH 
# Interactive login (Browser open hoga)
$ rsp auth login
Opening browser to https://cloud.rsplatforms.com/auth/device...
✓ Successfully authenticated as admin@rsplatforms.com

# Ya fir headless servers ke liye API Key use karein
$ rsp auth login --api-key rsp_live_xYz123aBc456...
✓ API Key authenticated and stored locally.

3. Compute Operations

Sabse zyada use hone wali commands instances launch, list, aur destroy karne ke liye.

rsp compute launch

Naya AI cluster deploy karne ke liye (e.g., 8x H100 GPUs).

# Yahan hum 8 nodes ka H100 cluster bana rahe hain PyTorch image ke sath
$ rsp compute launch \
    --name "llm-trainer-1" \
    --gpu-type H100_SXM5 \
    --nodes 8 \
    --image "pytorch-latest"

Provisioning cluster (ID: cl_9x8y7z)...
✓ Cluster 'llm-trainer-1' is up and running.
IP: 192.168.10.45

rsp compute list

Apne account ke saare active instances dekhne ke liye.

$ rsp compute list

ID         NAME           TYPE       NODES  STATUS   IP
------------------------------------------------------------------
cl_9x8y7z  llm-trainer-1  H100_SXM5  8      RUNNING  192.168.10.45
cl_1a2b3c  inference-api  A100_80GB  2      RUNNING  192.168.10.88
cl_5p6q7r  batch-jobs     B200_NVL   32     BOOTING  --

rsp compute destroy

Billing rokne aur instance delete karne ke liye.

$ rsp compute destroy cl_9x8y7z

Warning: This will permanently delete 'llm-trainer-1' and all local NVMe data.
Are you sure? [y/N]: y
Terminating instance...
✓ Destroyed successfully. Billing stopped.

4. Storage Operations

High-speed parallel NVMe volumes create aur attach karne ke commands.

# 10 TB ka naya NVMe volume create karna
$ rsp storage create --name "dataset-vol" --size 10TB --type nvme
✓ Volume 'vol_ab12cd' created.

# Volume ko running cluster ke sath attach karna
$ rsp storage attach vol_ab12cd --cluster cl_1a2b3c
✓ Volume attached. Available at /mnt/dataset-vol inside the instance.

5. Monitoring & Logs

Training ke dauran GPU temperatures, VRAM usage aur logs terminal me live dekhne ke liye.

# Live GPU telemetry stream karne ke liye
$ rsp top --cluster cl_1a2b3c

Connecting to telemetry server...
-------------------------------------------------------
GPU 0 (H100) | Util: 98% | Mem: 76GB/80GB | Temp: 68°C | Pwr: 650W
GPU 1 (H100) | Util: 97% | Mem: 76GB/80GB | Temp: 69°C | Pwr: 645W
-------------------------------------------------------
Press Ctrl+C to exit.

# Instance ke raw logs fetch karne ke liye
$ rsp logs cl_1a2b3c --tail 50
[12:45:01] Epoch 14/100 completed in 320s. Loss: 0.1205
[12:45:05] Checkpoint saved to /mnt/dataset-vol/chkpt_14.pt

Need Higher CLI Limits?

Agar aapko automation ke through 100+ GPUs spin up karne hain, toh aapko enterprise limits aur custom SLA ki zaroorat padegi.

Contact Sales Team