Jidoka.Eval (Jidoka v0.8.0-beta.1)

Copy Markdown View Source

Small deterministic eval runner for Jidoka harness flows.

The runner intentionally delegates execution to Jidoka.Harness. It adds no new runtime path; it only packages an agent/request pair with assertions that are useful for examples, regression tests, and optional live smoke checks.

Summary

Types

case_input()

Functions

evaluate(case, result)

Evaluates supported assertions against a completed turn result.

run_case(eval_case_input, opts \\ [])

Runs one eval case through the harness.

Types

case_input()

@type case_input() :: Jidoka.Eval.Case.t() | keyword() | map()

Functions

evaluate(case, result)

@spec evaluate(Jidoka.Eval.Case.t(), Jidoka.Turn.Result.t()) :: [
  Jidoka.Eval.Run.assertion()
]

Evaluates supported assertions against a completed turn result.

run_case(eval_case_input, opts \\ [])

@spec run_case(
  case_input(),
  keyword()
) :: {:ok, Jidoka.Eval.Run.t()} | {:error, term()}

Runs one eval case through the harness.