Documentation
Technical details on how Rebridge processes invoices, matches products, and syncs with Lightspeed.
Invoice Processing Pipeline
When you upload a PDF, Rebridge runs a multi-step pipeline: extract text from the PDF, parse line items using AI, sync your latest catalog from Lightspeed, match each line item against your catalog, classify matches by confidence, and present the results for review.
Each step is checkpointed. If the pipeline fails at any point, it resumes from the last successful step rather than starting over.
Product Matching Engine
The matcher searches your local catalog copy by manufacturer SKU first, then falls back to fuzzy name matching. It distinguishes between four classifications: reorder (exact product exists), new variant (matrix exists but this color/size combination is new), new product (nothing matches), and manual review (ambiguous match).
Each match includes a confidence score and reasoning so you can understand why Rebridge made its decision.
Catalog Audit
On first connection, Rebridge runs a deep audit of your entire Lightspeed catalog. This produces a statistical profile covering naming conventions, attribute usage, category structure, size systems, color formats, SKU patterns, and pricing.
This profile drives how Rebridge formats new products — ensuring they match your existing catalog conventions without manual configuration.
Category Assignment
New products are assigned categories through a four-step process: inherit from matched product, keyword matching against your category index, AI-assisted selection from your valid category list, and finally human review for ambiguous cases. This ensures products always land in the right part of your catalog.
Lightspeed API Integration
Rebridge uses Lightspeed's REST API with OAuth 2.0 authentication. All API calls are rate-limited with a token bucket algorithm that respects Lightspeed's per-account limits. Catalog syncs use cursor-based pagination to handle catalogs of any size.