Parse text-based and scanned PDF resumes.

Text extraction first

For a valid PDF, CVLens inspects the document and uses layout-aware text extraction across the accepted page range. This keeps digitally generated resumes fast and preserves enough structure for the parser to identify sections and dates.

OCR fallback for scanned documents

When embedded PDF text falls below the configured minimum, each accepted page is rendered to an image and processed with Tesseract OCR. The current runtime is configured for English and German text.

OCR is a fallback, not a guarantee of perfect transcription. Low-resolution scans, unusual typography, handwriting, and complex multi-column layouts can still require human review.

Guardrails before model processing

• The uploaded bytes must match the declared PDF format.
• Page count, extraction time, output size, and rendered image dimensions are bounded.
• Extraction warnings remain visible in the parse response.
• Structured output should be validated before it drives candidate decisions.

Review how CV files and parser artifacts are retained.