What I tried
Replay a captured run with a stricter policy to demonstrate the "what if Guard had been tougher" workflow the README pitches.
const replay = runFromEvent({
ledger,
eventId,
// overrides: ... ← only memory-related overrides documented
// policy: tighten(...) ← exported but the README example never wires it through
});
const diff = diffReplayAgainstOriginal(ledger, replay);
What broke
Even with tighten(demoPolicy, { deny: [...] }) passed, the diff comes back with every entry tagged identical. The override doesn't change verdicts during replay.
Expected
The README's "hero diff" section explicitly shows require_approval (constrain) → allow flipping under a stricter policy. That capability needs a tested path through runFromEvent + diffReplayAgainstOriginal.
Repro
/tmp/actantdb-real-project/replay-override.mjs.
What I tried
Replay a captured run with a stricter policy to demonstrate the "what if Guard had been tougher" workflow the README pitches.
What broke
Even with
tighten(demoPolicy, { deny: [...] })passed, the diff comes back with every entry taggedidentical. The override doesn't change verdicts during replay.Expected
The README's "hero diff" section explicitly shows
require_approval (constrain) → allowflipping under a stricter policy. That capability needs a tested path throughrunFromEvent+diffReplayAgainstOriginal.Repro
/tmp/actantdb-real-project/replay-override.mjs.