manage-prime-gpu-pods

To find and rank on-demand GPU pods by price, use the Prime CLI with the select_on_demand_gpu.py script to filter non-spot configurations and output the cheapest options as deterministic create commands. You can generate results in json, table, id, or create-cmd formats with top-N filtering.

Yes, you can automate reproducible GPU provisioning by generating deterministic create commands via the Prime CLI. The workflow supports filtering by GPU type, count, provider, and location, ensuring deployments are consistent and repeatable across automation agents.

The Prime CLI supports filtering non-spot GPU configurations for pod creation. The bundled select_on_demand_gpu.py script applies non-spot filters to the available inventory, ensuring you only provision on-demand instances without spot pricing volatility.

The best way to manage GPU pod status, SSH access, and termination is using the Prime CLI workflow, which operates end-to-end from discovery to creation, monitoring, SSH access, and clean termination directly from the terminal.

To generate a deterministic create command for the cheapest non-spot GPU, run the select_on_demand_gpu.py script with the Prime CLI to rank candidates by price and emit a reproducible create command format for immediate pod deployment.

Available output formats for filtering on-demand GPU configurations include json, table, id, and create-cmd. These formats support top-N filtering to limit results, enabling seamless integration into automation pipelines and agent-driven workflows.