Plan Viewer is a TypeScript pipeline that extracts plan numbers from scanned construction drawing PDFs using OCR. Large sets of drawings are often distributed as a single unlabeled PDF, making it hard to programmatically identify which page corresponds to which plan number. This automates that lookup instead of relying on manual page-by-page inspection. Given a multi-page PDF, it returns a structured array of results indexed by page.
The pipeline loads each PDF with pdfjs-dist, running entirely on bundled WASM builds with no native dependencies. Every page is rendered to a canvas and cropped using a fixed transform to isolate the title-block corner of the sheet where the plan number lives, avoiding the cost of processing the full-resolution page. That cropped region is passed to Tesseract.js, which OCRs the image and parses the plan number out of the recognized text. Pages are processed concurrently via p-limit, capping simultaneous OCR jobs so throughput stays high without exhausting memory.
Foremen can use the program to quickly look up plans, create custom plan sets, and make edits and notes within the pages of the plans. Overall it makes their work more efficient, and provides an organized platform for all of their plan-based needs.