Partner data ingestion, visualization and branding automation workflow

Project Summary

Built a production-oriented partner mapping platform for 24/7 response operations, with upload-driven ingest, geocoding, media-kit generation, and optional Supabase persistence. The app runs as a single Express service and deploys to Vercel using one serverless entrypoint.

Architecture

• Express app as the primary runtime, shared by local mode and Vercel serverless mode.
• Vanilla front-end with Leaflet map experience plus dedicated ingest and media-kit interfaces.
• File parsing and transformation pipeline for CSV, XLSX, PDF, and DOCX inputs.
• Mapbox geocoding integration with fallback and session-based ingest flow management.
• Media-kit compositing pipeline using Sharp and canvas tooling.

Data and Integrations

• Dual persistence strategy: Supabase-backed mode or local JSON file fallback.
• Supabase storage plus RPC path for confirm/commit ingest operations.
• Schema and migration workflow managed under dedicated SQL migration files.
• Operational scripts for import, pruning, env syncing, and media asset preparation.

Tech Stack

• Backend: Node.js, Express, CommonJS modules.
• Ingest: multer, csv-parser, xlsx, pdf-parse, pdfjs-dist, mammoth.
• Data: Supabase JavaScript client and SQL migrations.
• Frontend: HTML, CSS, JavaScript, Leaflet + MarkerCluster.
• Supplemental analytics and prototyping in Python.

Engineering Risks Identified

• Unauthenticated destructive route (`DELETE /api/stores`) should be gated or removed.
• Large body and upload limits increase abuse surface on public deployments.
• Serverless in-memory sessions can fail across multi-instance invocation boundaries.
• Any non-env token fallback should be eliminated for production safety.