Loading…
Guides for wiring up a model, running a benchmark end-to-end, and reading the data and costs that come out the other side.
How Benchwright works. Pipeline phases, sandboxes, events, fingerprints, scorers.
Public HTTP API. Resources, auth, conventions, errors.
Scenario-driven walkthrough. Run benchmarks, iterate on failures, author your own, build composites.
How your vaulted secrets stay out of reach of the agent that runs your benchmarks.
The benchwright binary. Launch runs, author benchmarks, manage secrets and billing from the shell.