remote-bench

What problem does it solve?

Producing reliable benchmark numbers is hard when your execution machine is shared or noisy, causing run-to-run variance that makes comparisons untrustworthy.

Core Features & Use Cases

• Run noise-free EC2 benchmarks: Executes Barretenberg/BB benchmark targets on the dedicated remote benchmarking instance for isolated, publishable single-run results.
• Safe remote orchestration: Validates required environment variables, enforces a remote lock to prevent concurrent benchmark corruption, transfers the needed binary and inputs, and retrieves benchmark outputs.
• Workload-agnostic transport: Does not decide what to benchmark; it’s designed to pair with other benchmark skills (e.g., chonk/native/A-B comparison workflows) that define the actual BB command and inputs.

Quick Start

Ask the AI to run the remote benchmark for target bb with the command you intend to measure, after confirming BB_SSH_KEY, BB_SSH_INSTANCE, and BB_SSH_CPP_PATH are configured for the remote machine.