What Fact check found, in three parts: every cited source opened and judged, the reasoning read step by step, and what the answer missed. Named for the AI that ran it.
Each source opens in place to the passage it was checked against, so a verdict of "Not in that source" can be verified by the reader rather than taken on trust. What the answer missed is added after it and marked as the checker's, never merged in.
Fixing rewrites only the sentences flagged above, and says so beside the button.
import { FactCheckReport } from '@redrob-labs/ui';
| Prop | Type | Notes |
|---|---|---|
by |
React.ReactNode |
The AI that ran the check. |
took |
React.ReactNode |
|
summary |
React.ReactNode |
|
claims |
FactClaim[] |
|
reasoning |
FindingProps[] |
|
missed |
Array<{ by?: string; label?: React.ReactNode; text: React.ReactNode }> |
|
fixed |
boolean |
|
onFix |
() => void |
|
onClose |
() => void |
|
defaultOpen |
string | null |
|
verdicts |
Partial<Record<FactVerdict, [NonNullable<BadgeProps['tone']> | string, React.ReactNode]>> |
|
title |
React.ReactNode |
|
label |
string |
|
fixLabel |
string |
|
fixHint |
React.ReactNode |
|
claimLabel |
string |
|
sourcesLabel |
string |
|
reasoningLabel |
string |
|
missedLabel |
string |
|
missedItemLabel |
string |
|
closeLabel |
string |
|
className |
string |