drive_log_parser

Parses pier driving logs from CSV, PDF text, Excel, and handwritten scanned images (Claude Vision). Produces structured data for cert generation.

Connections

• depends on: permit-parser (Vision path uses Opus model)
• feeds into: cert-data-service (sparkline cards, aggregate charts)
• consumed by: cert-generator-v3 (driving log companion document)

Output Format

{
  "type": "psi" | "torque",
  "stats": [PIERS, AVG DEPTH, AVG PSI/TORQUE, MAX, optional EXT SLEEVES],
  "headers": [...],
  "rows": [...],
  "raw_piers": [{ pier_id, location, depth, tubes, final_psi, readings, depth_readings, extension_sleeve }],
  "installation_date": "YYYY-MM-DD"
}

Notes

• Dual variant support: PSI (push/slab) and torque (helical)
• depth_readings array provides depth-associated readings for accurate sparklines (session 39)
  • PSI: 3 ft per tube (standard tube length)
  • Torque: inferred from total_depth / interval_count (Vision) or explicit from log
• extension_sleeve boolean per pier — conditional stat in stats strip
• installation_date extracted from labeled date patterns in logs
• Image preprocessing: grayscale + contrast 1.8 tuned for yellow carbon-copy pads
• Backward compatible: old driving_log.json files without new fields still work