Skip to content

v1.18.0 — Evidence-backed session review

Choose a tag to compare

View all tags
@alloevil alloevil released this 23 Sep 15:16
v1.18.0
c882564
This commit was created on GitHub.com and signed with GitHub’s verified signature.
GPG key ID: B5690EEEBB952194
Verified
Learn about vigilant mode.

Evidence-backed session review

  • Group repeated unresolved failures while retaining first, last and every individual result as evidence.
  • Recognize OMP-native eval/search failures and distinguish completed, running, cancelled and unknown tool outcomes without rewriting raw fields.
  • Record browser-local human review notes; changed full evidence invalidates old reviews. Manual labels never rewrite automatic results.
  • Review sessions on narrow screens with list/content navigation, retained same-session drafts and evidence jumps that keep navigation visible.
  • Run the synthetic terminal walkthrough from a checkout with node scripts/demo-diagnostics.cjs.

Validation: 214 Node tests passed for this release, plus UI build, lint and browser checks for review persistence, isolation, staleness, storage failures, evidence navigation and narrow screens. Public verification receipt · Usage and limitations.

npx @alloevil/agent-xray@1.18.0

Requires Node 22.13+ (22.15+ for compressed DeepSeek Harness logs). Diagnostics describe loaded records, not task correctness. Review notes are unencrypted browser-local storage, not a backup. No new runtime dependencies, log uploads or automatic execution of logged commands. Personal evaluation artifacts are excluded.

Assets 4
Loading