OCR/document-ingestion library (Node/TypeScript and Python implementations) that converts PDFs, Word/Excel/PowerPoint files, and images to page images, then sends each to a vision-language model (OpenAI, Azure OpenAI, AWS Bedrock, Google Gemini, or Vertex AI) to produce Markdown, preserving tables/charts; supports structured JSON-schema extraction, page selection, concurrent processing, and orientation/edge correction.