Document parse

Two local engines, picked by input format. Both are deterministic unless an AI model is asked for.

LiteParse - run-llama/liteparse, Apache-2.0. Rust core, PDFium renderer, bundled Tesseract. Command: lit.
AILANG Parse - sunholo-data/ailang-parse, Apache-2.0. Parses Office/ODF XML directly and runs on the AILANG runtime. Command: docparse.

1. Pick the engine

Input	Engine	Command
PDF, text layer or scanned	lit	`uvx --from liteparse lit parse file.pdf --format markdown -o out.md`
PDF where the visual carries meaning (charts, diagrams, handwriting)	docparse	`docparse file.pdf --pdf-backend ai`
Photo or scan image (JPG, PNG, TIFF...) needing plain text	lit	`uvx --from liteparse lit parse photo.png -o out.txt`
Image needing interpretation	docparse	`docparse photo.png` (AI switches on automatically)
DOCX, PPTX, XLSX, ODT, ODP, ODS	docparse	`docparse report.docx`
HTML, Markdown, CSV, TEX, RTF, EPUB, EML, MBOX	docparse	`docparse book.epub`
Convert between formats	docparse	`docparse notes.md --convert report.docx`