What problem does it solve? Deploying a Boundless prover to a GPU server involves many error-prone manual steps: installing NVIDIA drivers, configuring Docker with the NVIDIA Container Toolkit, writing Ansible inventories, managing secrets, and tuning broker.toml settings. This Skill guides the entire deployment, redeployment, and troubleshooting workflow so nothing is missed. ## Core Features & Use Cases - Full-stack deployment: Creates Ansible inventory files with env-var-based secrets, tests SSH connectivity, and runs the prover playbook to install NVIDIA drivers, Docker, and the bento compose stack. - Flexible image modes: Supports pre-built nightly images, pinned SHA tags, release versions, or building from source, plus custom broker.toml via URL or local file. - Operations & troubleshooting: Covers teardown, selective deployment with --skip-tags, log inspection, GPU verification, postgres password fixes, and scaling the bento-worker-overflow AWS Auto Scaling group. - Use Case: You have a new Latitude.sh GPU server and need a production prover running. The Skill walks you through creating the inventory, exporting PROVER_PRIVATE_KEY and PROVER_RPC_URLS, running the playbook, and verifying the broker is healthy. ## Quick Start Ask the assistant to deploy a Boundless prover to your GPU server at a given IP address using Ansible with the nightly image tag.
