sparkrun
Get Running in 3 Steps
Install sparkrun
One command installs sparkrun, sets up a managed environment, configures tab completion, and starts the setup wizard.
uvx sparkrun setupConfigure your cluster(s)
Setup your DGX Spark or Sparks with our best practices by following the wizard.
Run inference
Pick a recipe and follow startup until readiness is confirmed. Use --follow to keep watching logs; Ctrl+C detaches from the workload.
sparkrun run qwen3-1.7b-vllmOne Command Launch
Pick a recipe, run it. sparkrun handles container orchestration, model distribution, and networking automatically.
Multi-Node Tensor Parallel
Scale across DGX Sparks with --tp. Each Spark contributes one GPU — sparkrun handles InfiniBand/RDMA and NCCL configuration.
Recipe System
YAML configs capture model, container, runtime, and defaults. Override anything at launch time — no config files to hunt down.
Recipe Registries
Share and collaborate on recipes via git registries. Add community or private registries and search across all of them.
VRAM Estimation
Estimate model weights and cache demand before launching. Unknown or lower-bound estimates need validation against your hardware and runtime.
Multiple Runtimes
First-class support for vLLM, SGLang, llama.cpp, atlas, and more. Same CLI, same recipe format, different engines under the hood.
Claude Code Plugin
AI-assisted inference management. Claude learns your cluster and helps run, monitor, and stop workloads conversationally.
CLI Tab Completion
Rich shell completions for Bash, Zsh, and Fish. Tab-complete commands, recipe names, cluster names, and options instantly.