# Warpscale

## Docs

- [What is Warpscale?](https://docs.warpscale.ai/index.md): SRE platform for GPU infrastructure.
- [Quickstart](https://docs.warpscale.ai/quickstart.md): Sign in, create an API key, and install Warpscale.
- [What to install](https://docs.warpscale.ai/install/index.md): The four Warpscale components, what each one is for, and the requirements they share.
- [Install the probe](https://docs.warpscale.ai/install/probe/index.md): The per-node collector. Required for every workload — choose where it runs.
- [Install the probe on a Linux host](https://docs.warpscale.ai/install/probe/linux-host.md): Download the binaries and run the probe on a machine with GPUs.
- [Install the CLI](https://docs.warpscale.ai/install/cli.md): Install the warpscale command.
- [Install the Warpscale Python SDK](https://docs.warpscale.ai/install/python-sdk.md): Install the warpscale Python package.
- [Install the Warpscale vLLM plugin](https://docs.warpscale.ai/install/vllm-plugin.md): Install the warpscale-vllm package into your vLLM environment.
- [Set up training](https://docs.warpscale.ai/training/wrap.md): Prefix your launch command with warpscale run to capture GPU, CUDA, and NCCL telemetry.
- [Report training progress](https://docs.warpscale.ai/training/python-sdk.md): Report run intent and step, epoch, and checkpoint progress from your training script.
- [Set up vLLM](https://docs.warpscale.ai/inference/setup.md): Point the probe at vLLM, start the server with the plugin enabled, and confirm telemetry arrives.
- [Grafana](https://docs.warpscale.ai/observability/grafana.md): The bundled Grafana instance and the dashboards it ships with.