vllm plugin
vLLM operator reference suite — deployment, configuration, quantization, caching, KV, tool parsers, reasoning parsers, chat templates, benchmarking, performance tuning, observability, omni, input modalities, speculative decoding, NVIDIA hardware, and a Gemma 4 31B operating-point serve recipe.
How to install the vllm plugin
Run both commands inside Claude Code. The first adds the marketplace once; the second installs the plugin.
/plugin marketplace add air-gapped/skills
/plugin install vllm@air-gapped-marketplaceWhat's inside vllm
15 parts read from the plugin's own files.
Skills 15
Claude loads it when the task matches- vllm-benchmarking
Run production-grade benchmarks for vLLM to measure latency, throughput, and SLO compliance using various subcommands and methodologies.
- vllm-caching
Configures tiered KV cache offloading for vLLM production clusters using CPU, NVMe, and RDMA backends.
- vllm-chat-templates
Provides a reference for managing Jinja-based chat templates, resolution precedence, and troubleshooting prompt-rendering issues in vLLM deployments.
- vllm-configuration
Configure vLLM deployment settings via YAML, CLI arguments, and environment variables, including specialized setups for air-gapped environments.
- vllm-deployment
Helps agents author, review, and fix vLLM deployment manifests for Kubernetes, Docker, and OpenShift, including multi-node and MoE configurations.
- vllm-gemma-4-31b
Provides optimized deployment configurations and performance benchmarks for serving Gemma 4 31B on vLLM hardware.
- vllm-input-modalities
Configure and deploy non-chat inference endpoints for text embeddings, reranking, speech-to-text, and document OCR using vLLM.
- vllm-nvidia-hardware
Provides technical specifications and deployment guidance for NVIDIA AI hardware ranging from Hopper to Vera Rubin architectures for vLLM workloads.
- vllm-observability
Monitor and troubleshoot production vLLM deployments using Prometheus metrics, Grafana dashboards, and OTLP tracing.
- vllm-omni
Extends vLLM to support multimodal generation for images, video, audio, and any-to-any omni models via a plugin layer.
- vllm-performance-tuning
Optimizes vLLM performance through MoE kernel autotuning, parallelism configuration, scheduler knob adjustment, and distributed system tuning.
- vllm-quantization
Configures and troubleshoots GPU quantization formats and KV-cache dtypes for high-performance LLM serving on NVIDIA and AMD hardware.
- vllm-reasoning-parsers
Enables agents to split model output into distinct reasoning and content fields using specialized parsers for thinking models.
- vllm-speculative-decoding
Configure, tune, and monitor speculative decoding methods in vLLM production environments to improve inference throughput.
- vllm-tool-parsers
Provides a navigation map for selecting, configuring, and debugging tool-calling parsers within vLLM deployments.
Using vllm in Claude Code
Skills need no command. Claude reads them when your request matches what they describe.
Update, turn off or remove
- Update the plugin (shell)
- claude plugin update vllm@air-gapped-marketplace
- Refresh the marketplace
- /plugin marketplace update air-gapped-marketplace
- Turn it off for now
- /plugin disable vllm@air-gapped-marketplace
- Remove it
- /plugin uninstall vllm@air-gapped-marketplace
Skills in vllm
- vllm-benchmarkingair-gapped/skills
Run production-grade benchmarks for vLLM to measure latency, throughput, and SLO compliance using various subcommands and methodologies.
11 files6stars - vllm-configurationair-gapped/skills
Configure vLLM deployment settings via YAML, CLI arguments, and environment variables, including specialized setups for air-gapped environments.
12 files6stars - vllm-deploymentair-gapped/skills
Helps agents author, review, and fix vLLM deployment manifests for Kubernetes, Docker, and OpenShift, including multi-node and MoE configurations.
18 files6stars - vllm-gemma-4-31bair-gapped/skills
Provides optimized deployment configurations and performance benchmarks for serving Gemma 4 31B on vLLM hardware.
8 files6stars - vllm-observabilityair-gapped/skills
Monitor and troubleshoot production vLLM deployments using Prometheus metrics, Grafana dashboards, and OTLP tracing.
9 files6stars - vllm-quantizationair-gapped/skills
Configures and troubleshoots GPU quantization formats and KV-cache dtypes for high-performance LLM serving on NVIDIA and AMD hardware.
13 files6stars - vllm-speculative-decodingair-gapped/skills
Configure, tune, and monitor speculative decoding methods in vLLM production environments to improve inference throughput.
15 files6stars - vllm-cachingair-gapped/skills
Configures tiered KV cache offloading for vLLM production clusters using CPU, NVMe, and RDMA backends.
8 files6stars - vllm-chat-templatesair-gapped/skills
Provides a reference for managing Jinja-based chat templates, resolution precedence, and troubleshooting prompt-rendering issues in vLLM deployments.
10 files6stars - vllm-input-modalitiesair-gapped/skills
Configure and deploy non-chat inference endpoints for text embeddings, reranking, speech-to-text, and document OCR using vLLM.
11 files6stars - vllm-nvidia-hardwareair-gapped/skills
Provides technical specifications and deployment guidance for NVIDIA AI hardware ranging from Hopper to Vera Rubin architectures for vLLM workloads.
12 files6stars - vllm-omniair-gapped/skills
Extends vLLM to support multimodal generation for images, video, audio, and any-to-any omni models via a plugin layer.
9 files6stars - vllm-performance-tuningair-gapped/skills
Optimizes vLLM performance through MoE kernel autotuning, parallelism configuration, scheduler knob adjustment, and distributed system tuning.
8 files6stars - vllm-reasoning-parsersair-gapped/skills
Enables agents to split model output into distinct reasoning and content fields using specialized parsers for thinking models.
10 files6stars - vllm-tool-parsersair-gapped/skills
Provides a navigation map for selecting, configuring, and debugging tool-calling parsers within vLLM deployments.
9 files6stars
vllm plugin questions
How do I install the vllm plugin?
In Claude Code run /plugin marketplace add air-gapped/skills, then /plugin install vllm@air-gapped-marketplace.
What does the vllm plugin include?
vllm bundles 15 skills.
How do I update or remove the vllm plugin?
Update it with claude plugin update vllm@air-gapped-marketplace in your shell, or from the Installed tab of /plugin. Remove it with /plugin uninstall vllm@air-gapped-marketplace.
What license is the vllm plugin under?
MIT, as stated in the plugin's manifest.