RunPod vLLM Deployer

Overview

Use this repository's runpod-vllm CLI to plan, create, inspect, validate, and remove OpenAI-compatible vLLM GPU Pods. Keep configuration validation offline until the user explicitly authorizes a live operation, and preserve the local state needed for verified teardown.

Use and non-use triggers

Use this skill when the current request involves:

• Installing or checking the runpod-vllm-deployer repository.
• Producing an offline, redacted deployment plan.
• Creating a RunPod vLLM Pod after clear user approval.
• Listing Pods or inspecting the Pod recorded in local state.
• Waiting for vLLM readiness or running the deterministic endpoint smoke test.
• Recovering after Pod creation when local state could not be saved.
• Destroying a specifically verified Pod and confirming provider-side absence.

Do not use this skill to operate a different RunPod client, manage Serverless endpoints, alter accounts, estimate costs, or claim that a model will fit a GPU. A request to explain, plan, inspect code, run tests, or validate configuration does not authorize any live resource mutation.