Provenance Trail

AI Governance
Block

Complete data lineage and decision trail showing how AI arrived at conclusions. Tracks data sources, model inference, and human approvals with cryptographic hash verification.

Preview

Track every step from data ingestion to AI inference and human approval

Provenance Trail

3 steps • Complete audit trail with data lineage

Data Pipeline
data ingestion
Task Priority Analysis
12/1/2024, 10:30:00 AM
AI Agent (claude-sonnet-4)
model inference
Task Priority Analysis
12/1/2024, 10:31:15 AM
87%
claude-sonnet-4

"Analysis shows Marketing Strategy project has highest deadline risk based on current velocity"

Olivia Chen
human approval
Task Priority Analysis
12/1/2024, 10:35:22 AM

"Recommendation aligns with Q4 sprint priorities. Approved for implementation."

Props

ProvenanceTrail component API reference

PropTypeDefaultDescription
stepsStep[]—Array of provenance steps
compactbooleanfalseCompact display mode
searchablebooleantrueEnable search functionality
filterablebooleantrueEnable filter button
verifiablebooleantrueShow hash verification
onExport(format: string) => void—Callback when export is clicked
onVerify(stepId: string, verified: boolean) => void—Callback after hash verification

Step Type

PropTypeDefaultDescription
id*string—Unique step identifier
timestamp*string—ISO timestamp
action*string—Action type (data_ingestion, model_inference, human_approval)
actorActor—Who performed the action (type, name, email)
targetstring—Target of the action
sourcesSource[]—Data sources with name, hash, rows, quality_score
outputsRecord<string, unknown>—Step outputs
modelModel—Model info (name, version, parameters)
rationalestring—Explanation for the action
confidencenumber—Confidence score (0-1)
hashstring—Cryptographic hash for verification

Usage

Import and implementation example

import { ProvenanceTrail, type Step } from '@/blocks/ai-transparency/provenance-trail'

const steps: Step[] = [
  {
    id: 'step-1',
    timestamp: '2024-12-01T10:30:00Z',
    action: 'data_ingestion',
    actor: { type: 'system', name: 'Data Pipeline' },
    target: 'Task Analysis Job',
    sources: [
      { name: 'tasks-data.csv', rows: 2500, quality_score: 0.98 }
    ],
    hash: 'sha256:verified...'
  },
  {
    id: 'step-2',
    timestamp: '2024-12-01T10:31:00Z',
    action: 'model_inference',
    actor: { type: 'ai', name: 'AI Agent' },
    model: { name: 'claude-sonnet-4', version: '20250514' },
    confidence: 0.87,
    hash: 'sha256:abc123...'
  }
]

export default function Example() {
  return (
    <ProvenanceTrail
      steps={steps}
      searchable={true}
      verifiable={true}
      onVerify={(stepId, verified) => console.log('Verified:', stepId)}
      onExport={(format) => exportData(format)}
    />
  )
}

Built With

3 components

This block uses the following UI components from the design system:

ButtonInputBadge

Features

Built-in functionality

  • Complete data lineage: Track from source to output
  • Actor visualization: Icons for system, AI, and human actors
  • Expandable details: Click to see sources, outputs, and rationale
  • Model tracking: Model name, version, and parameters
  • Quality scores: Data source quality indicators
  • Hash verification: Cryptographic integrity verification per step
  • Search filtering: Real-time search across steps
  • Export functionality: Export for compliance documentation
  • Dark mode: Full dark mode support

Accessibility

Accessibility considerations

ARIA Attributes

Timeline structure uses semantic markupExpandable steps are clickable regionsSearch input has proper input semantics

Keyboard Navigation

KeyAction
TabNavigate between steps and controls
Enter / SpaceExpand/collapse step details

Notes

  • Actor type indicated by icon and name
  • Timestamps shown in readable format
  • Confidence shown as percentage
  • Hash verification status shown with icon and text

Related Components

Audit LogExplanation PanelConfidence IndicatorModel Registry Table