Clarify AI sidebar and fix empty Jev evidence reviews

[TerminallyLazy]

Problem and behavior

Jev preparation could show “Ready” with zero executable claims when an answer used a multi-sentence passage with a final citation or grouped citations such as [s1, s2]. New sidebar previews preserve the entire cited passage and judge it separately against each cited abstract. Empty previews now explain that Jev has not run and offer fresh preparation. Versioned validation preserves existing CLI and saved-plan semantics; the original answer remains unchanged.

The AI rail now separates Chat, Research, and Jev review. Optional voice controls are collapsed in Chat. Research cards show formatted answers and a clear review/result action; the dedicated review workspace hides the composer and unrelated results, with abstracts, probabilities and receipts collapsed. Selected public text still requires explicit confirmation before TypeSafe receives it. The muted palette and horizontal separators replace the former side accent; its obsolete detector suppression was removed.

Verification

• 248 backend/evaluator tests and 50 viewer tests passed; viewer type check and production build passed.
• Isolated synthetic Electron acceptance passed: text/research without Realtime, workspace separation, selection and consent, keyboard focus, cancellation and unknown-usage receipt, voice independence, saved-result reopening, stale-response rejection, history cleanup, full fixture NVIDIA catalog, and 280px layout.
• Actual desktop acceptance recovered the reported empty preview and completed one public NCBI/TypeSafe check: resolved jev-1.13.0, one completed assessment, and an exactly unchanged original answer. No images or patient text were sent. This establishes provider execution, not clinical validity or general model accuracy.
• Scoped design audit found no remaining detector issue in the changed rail. The pre-existing grid advisory in the unrelated local-viewer illustration remains unchanged and unsuppressed.

Owning guidance and the Jev implementation runbook are updated with the workflow, extraction version and acceptance boundaries.

Hosted checks passed for a330444d36d8fce91d503847d7edd153e648cbf5: backend regressions/dependency audits, browser builds/tests/audits, and both CodeQL languages. External review limitations: RecurseML returned an analysis error; CodeRabbit skipped review; Qodo reported an inactive subscription. These are not counted as successful code reviews.

[qodo-code-review]

ⓘ Qodo reviews are paused because the subscription is no longer active. Ask your workspace admin to reactivate the subscription to resume reviews. Manage billing

[coderabbitai]

Important

• 🔍 Trigger review

This repository does not receive automatic reviews because it has fewer than 10 stars.

⚙️ Run configuration

Configuration used: defaults

Review profile: CHILL

Plan: Advanced

Run ID: df07d6fd-021d-4571-8ab2-15f3b59d0df5

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands.}