sdac-analyze-duplicates-context

Purpose

Analyzes duplicate salary/fringe patterns and assigns risk levels (LOW/MEDIUM/HIGH).

Tool type

AI-assisted (LLM; non-deterministic)

Tool ID

sdac-analyze-duplicates-context

Inputs

• report_id (string)
• min_occurrences (number, optional, default: 3)
• fields (array of salary/fringe, optional)

Outputs

• Duplicate groups with risk level and recommendations
• Summary statistics

Example output

{
  "success": true,
  "report_id": "SDAC-2024-Q2-001",
  "duplicate_groups_analyzed": 1,
  "high_risk_groups": 1,
  "groups": [
    {
      "group_id": "dup-001",
      "record_ids": [
        101,
        102
      ],
      "row_numbers": [
        12,
        13
      ],
      "duplicate_type": "SAME_PERSON_SAME_COSTS",
      "risk_level": "HIGH",
      "explanation": "The same employee appears on adjacent rows with identical salary and fringe amounts.",
      "recommendation": "Confirm whether one row should be removed or whether the rows represent separate assignments."
    }
  ],
  "summary": {
    "total_groups": 1,
    "high_risk_groups": 1,
    "medium_risk_groups": 0,
    "low_risk_groups": 0
  }
}

Functional details

• Uses the SDAC Duplicates Analyzer agent to contextualize duplicate patterns.
• Produces explanations and remediation recommendations.
• Highlights streaks and repeated values with risk scoring.

Implementation details

• Groups salary/fringe values across the report and checks consecutive rows.
• Sends grouped patterns to the Duplicates Analyzer agent.
• Risk levels and explanations come from LLM analysis.
• Considers only positive values when grouping duplicates.

Code Location: packages/domain-sdac/src/tools/analyze-duplicates-context.tool.ts