[crabl]

Libraries like iText would be SO good with LLM/vision model integration and vice-versa. Huge opportunity to use these tools to generate more training data based from siloed PDFs.