Module baseline 

A shipped, ready-to-run baseline capability suite — basic checks any agent can be measured against in one call.

let report = eval_core::run_suite(&my_agent, &eval_core::baseline());
println!("{report}");

The suite covers three capability areas, authored as RON files grouped by capability and EMBEDDED into this crate (so they ship with the published crate — nothing is read from disk at runtime):

• arithmetic.ron — basic math (number assertions, plus a clearly-labelled opt-in tool-use subset),
• language.ron — instruction following over the agent’s final text (fully portable),
• tool_use.ron — tool-calling with parameters (assumes a DOCUMENTED tool-name convention — adapt it).

Portable vs. adapt-me

The number/text assertions are PORTABLE: they inspect only the agent’s final reply, so they hold whether the agent uses tools or reasons inline. The tool-use cases (and a small subset of the arithmetic file) additionally assert specific tool names/args, which a user must adapt to their own agent — every such case is clearly commented at the top of its file.

Use it, or fork it

Run it directly via baseline, OR dump the raw embedded RON via baseline_files as a starting template you copy into your own suite directory and edit. The same files back both paths, so the template you copy is exactly what baseline runs.

Functions

baseline

The ready-made baseline capability suite: every embedded case, parsed and concatenated.

baseline_files

The raw embedded baseline files as (file_name, ron_contents) pairs, in sorted (filename) order.