Quality Inspection Knowledge Base Builder

Quality Record Pain Points

Construction quality inspection records often sit in scattered Word, Excel, and PDF files. Some are tabular supervision joint inspections; others are narrative reports from quality supervision stations or construction committees. Manual curation can miss fields, misclassify issues, or reduce solution to “rectified” or “deadline: 3 days” without reusable cause, action, spec basis, or resource notes. This skill turns one inspection cycle into searchable, reusable knowledge entries.

How It Builds the Knowledge Base

• Document conversion and extraction: Batch-converts legacy OLE2 .doc files when detected, then extracts issues and remediation notes from supervision records, supervision responses, quality station reports, and committee reports into structured JSON.
• Pairing and classification: Matches inspection issues with remediation responses, maps them to the 4-category, 75+ subtype taxonomy, and extracts keywords.
• Field completion and validation: AI fills cause, solution, applicable_condition, estimated_time, required_resources, and reference_spec using references/field_mapping.md; clean_data.py and enrich_data.py remove duplicates, repair malformed fields, backfill missing values, and calibrate severity.
• Knowledge base output: Writes the final records to knowledge_base.xlsx with the 13-column template for retrieval, analysis, and reuse.

Fit and Caveats

This skill is best suited to construction quality inspection records where “phenomenon, cause, solution, and spec basis” should be preserved. It depends on the existing taxonomy and field mapping rules; new inspection types may require extending category_taxonomy.md or field_mapping.md. For large batches, persist intermediate JSON checkpoints. solution should be concrete operating steps, not generic closure statements.