fuzzy-matching
Multi-Stage Fuzzy Matching
Production-grade fuzzy matching for inventory items, products, or entity reconciliation.
When to Use This Skill
- Matching vendor SKUs to inventory items
- Entity reconciliation with varying name formats
- Product deduplication across sources
- Any scenario where exact matching misses valid matches
Core Concepts
Three-stage pipeline: PostgreSQL trigram (fast pre-filter) → Salient overlap (quick filter) → Multi-factor similarity (expensive, accurate). Achieves O(log n) with proper indexing.
Stage 1: PostgreSQL Trigram (fast) → 50 candidates
Stage 2: Salient Overlap Check (fast) → ~20 candidates
More from dadbodgeoff/drift
sse-streaming
Implement Server-Sent Events (SSE) for real-time updates with automatic reconnection and heartbeats. Use when building live dashboards, notifications, progress indicators, or any feature needing server-to-client push.
78oauth-social-login
Implement OAuth 2.0 social login with Google, GitHub, and other providers. Handles token exchange, user creation, and account linking.
48multi-tenancy
Implement multi-tenant architecture with tenant isolation, data separation, and per-tenant configuration. Supports shared database and schema-per-tenant models.
45deduplication
Event deduplication with canonical selection, reputation scoring, and hash-based grouping for multi-source data aggregation. Handles both ID-based and content-based deduplication.
43webhook-security
Implement secure webhook handling with signature verification, replay protection, and idempotency. Use when receiving webhooks from third-party services like Stripe, GitHub, Twilio, or building your own webhook system.
37data-transformers
Centralized transformation logic for consistent data shaping across API routes. Includes aggregators, rankers, trend calculators, and data sanitizers.
36