catalog-enrichment-pipelinelisted

# Catalog Enrichment Pipeline Takes a raw supplier catalog feed and produces an enriched, normalized, scored catalog: consistent descriptions, mapped categories, flagged gaps, and a supplier-facing remediation report. ## Inputs - Supplier feed: CSV/TSV with whatever columns the supplier sent - Target taxonomy: the category tree items must map into - Enrichment sources allowed: supplier API, manufacturer part number lookup, structured extraction from description text Synthetic sparse feed in `references/sample_feed.csv`. ## Workflow 1. Run `scripts/profile_feed.py` to produce a completeness matrix: per field, fill rate and format consistency. 2. Normalize units of measure, pack sizes, and manufacturer part numbers using the rules in `references/normalization_rules.md`. UOM ambiguity (EA vs CS vs PK) is the highest-frequency ordering-error source; resolve it explicitly. 3. Map items to the target taxonomy. High-confidence mappings apply automatically; low-confidence mappings go to a review file with the top-3 candidate categories and the evidence for each. 4. Score the enriched catalog and emit the supplier remediation report: fields only the supplier can fix, ranked by order-impact. ## Output format Enriched catalog CSV, `## Completeness Report` (before/after matrix), `review_queue.csv`, supplier remediation report (markdown, sendable as-is). ## Guardrails - Never fabricate product attributes. Enrichment fills from a source or flags the gap. - Auto-applied category m