Evals are apte's main differentiator but the README mostly points to the docs
site. A reader skimming the README should understand the eval model without
leaving the page.
Add directly in the README:
- What an eval is in apte terms: a test that returns a value, scored not
asserted (the one-liner exists; expand it with a minimal but complete example).
- The scoring/evaluators model: built-in evaluators, custom evaluators,
LLM-as-judge, what a score looks like.
- The history/runs story:
apte eval, .apte/history.jsonl, how trends are
tracked across runs.
- When to reach for evals vs tests.
Keep the deep reference on the docs site, but the README should stand on its
own for the eval pitch.
Evals are apte's main differentiator but the README mostly points to the docs
site. A reader skimming the README should understand the eval model without
leaving the page.
Add directly in the README:
asserted (the one-liner exists; expand it with a minimal but complete example).
LLM-as-judge, what a score looks like.
apte eval,.apte/history.jsonl, how trends aretracked across runs.
Keep the deep reference on the docs site, but the README should stand on its
own for the eval pitch.