v3.14.3: Honest failures, Vitest 5 support

What's New

Agentic QE 3.14.3 makes test, security, coherence, and quality-gate results honest about failure, and adds Vitest 5 support everywhere a test runner is spawned. Most of this release was contributed by @rudycelekli.

Test execution

• A broken test run (suite that fails to import, crashing teardown, unhandled rejection, empty suite) now yields aqe test execute exit 1 and MCP success: false instead of a fabricated pass. Ordinary assertion failures, skips, and todos behave as before.
• Vitest 5 is supported in aqe test execute, the test_execute_parallel MCP tool, retry, flaky detection, and scheduled runs.

Security scanning

• Every scan carries per-file and per-engine execution receipts. Unreadable files, unsupported languages, and a missing or failed Semgrep are reported as partial or unavailable in the CLI, MCP, and saved JSON/SARIF/Markdown, never as clean.

Quality gates and coherence

• quality_assess({ runGate: true }) over MCP runs the same fail-closed, timestamped evidence evaluator as aqe quality --gate, and the session cache never replays a gate result.
• The coherence adapter now actually reaches the native sheaf engine for connected graphs.

Learning and verification

• Looking up past guidance no longer counts as a successful reuse.
• Judge responses that reference no recognized checklist item stay inconclusive, and the frontier-judge adapter uses the requested model or returns inconclusive.

CI

• The coherence job really runs its tests, runner exit codes propagate, coverage thresholds read a real summary, and fork PRs no longer fail on the comment step.

Dependencies

• Vitest 5.0.1 with matching coverage provider; adm-zip ≥ 0.6.1; js-yaml 4.3.2.

Upgrade notes

No database migration is required. Scripts relying on aqe security --dast exiting 0 without a target, or on rulesApplied equalling the sum of declared rule counts, should be updated.

Getting Started

npx agentic-qe init --auto

See CHANGELOG and release notes for full details.