Documentation

Choose the path that matches what you are trying to do.

Get started

Project README: preview the results site and verify the pipeline without credentials.
Run the stages yourself: inspect planning, execution, saved state, and evaluation with the small example.

System-under-test requirements: find out what your generation system must support, and at which conformance level, before a benchmark can measure it.
Protocol schemas: connect a generation system through the language-neutral adapter interface, including generation protocol v2.
Artemis integration: understand the accepted Artemis and Hyperion boundary.
Process evaluators: run an out-of-process evaluator under its own version, bounds, and secret handling.
Results-site guide: validate, build, preview, and publish release data.

Hyperion development pack: the 19 public exercise briefs, their provenance, coding rules, and limitations.
Hyperion development study: the runnable smoke configuration and the live Artemis template.

System design: components, stored records, resume behavior, and release process.
Methodology: what is measured and how formal comparisons are designed and reported.
Telemetry profile: what process evidence is captured, which claim it can support, and when a capture fails closed.
Reference pricing: report a dated OpenRouter catalog estimate without confusing it with provider billing.

Glossary: project terms and their corresponding protocol fields.
Changelog: changes to the generation, evaluation, dataset, and release contracts, and what an adapter or evaluator author has to do about them.
Restricted archives: packaging, verifying, and storing private run evidence.
Security policy: trust boundaries, isolation requirements, and private vulnerability reporting.
Licensing and third-party material: how project code, dependencies, and benchmark or result material are licensed, and what may be redistributed.
Citation metadata: authorship and software citation information.