Test coverage: Document processing jobs (AnalyzePdf, OCR)

[AndreRobitaille]

Summary

The document processing jobs (Documents::AnalyzePdfJob and Documents::OcrJob) have no test coverage. These jobs handle PDF text extraction and OCR — the foundation for all downstream AI analysis.

Jobs Needing Tests

Documents::AnalyzePdfJob (app/jobs/documents/analyze_pdf_job.rb)

• Extracts text from PDF documents using pdftotext
• Performs page-level analysis (creates Extraction rows)
• Classifies text_quality (good, poor, no_text)
• Triggers OcrJob for scanned/image-based documents

Test scenarios:

• Text-based PDF → extracts text, sets text_quality: good, creates Extraction rows per page
• Image-based/scanned PDF → detects low text quality, enqueues OcrJob
• Already-processed document (idempotent re-run) → clears and rebuilds extractions
• Handles corrupt/unreadable PDFs gracefully (doesn't crash)
• Updates MeetingDocument fields: extracted_text, text_chars, avg_chars_per_page, page_count

Documents::OcrJob (app/jobs/documents/ocr_job.rb)

• Runs Tesseract OCR on image-based PDFs
• Updates MeetingDocument.ocr_status and extracted text

Test scenarios:

• Scanned PDF → OCR produces text, updates extracted_text and ocr_status
• PDF with mixed text/image pages → handles correctly
• Tesseract unavailable → graceful failure with appropriate status
• Idempotent re-run

Approach

• Create small test PDF fixtures (test/fixtures/files/): one text-based, one image-based
• Stub system calls to pdftotext and tesseract where appropriate
• Test the full flow: download → analyze → OCR → extraction rows
• Verify text_quality classification logic

Dependencies

• Documents::DownloadJob already has tests (test/jobs/documents/download_job_test.rb) — use as a pattern

[AndreRobitaille]

Status (2026-04-14): partially addressed. AnalyzePdfJob now has test/jobs/documents/analyze_pdf_job_test.rb (1.9k). OcrJob still has no tests. Narrowing the scope of this issue to OCR coverage only.