Raw Record Vs Report Consistency

Output Structure

The output will be a JSON object containing:

• overall_status: Summary status (pass, fail, needs_review).
• project_match_summary: High-level status for specific checks.
• checks: Detailed array of individual rule validations.
• summary: A human-readable text summary of the findings.

See output_schema.json for the exact schema definition.

Execution Steps

1. Identify and Load Documents:

   • Ingest raw_records list.
   • Ingest report_document.

2. Extract Key Fields:

   • Dates: Extract inspection dates from raw records and inspection/appraisal dates (ranges supported) from the report.
   • Instrument IDs: Extract equipment identifiers from both sources.
   • Record IDs: Extract record unique identifiers from raw records and citations in the report.

3. Data Normalization:

   • Normalize dates to ISO 8601 (YYYY-MM-DD).
   • Normalize IDs (remove whitespace, standardize case).

4. Execute Validation Rules:

   • Apply R1 (Date Consistency).
   • Apply R2 (Instrument ID Consistency).
   • Apply R3 (Raw Record ID Consistency).
   • See rules.md for detailed logic.

5. Calculate Overall Status:

   • Determine overall_status based on the severity of rule results (see Logic below).

6. Generate Output:

   • Construct the JSON response with findings, evidence locations, and severity levels.

Status Logic