cloudgeni-ai/opengeni/document-parsing

Parse local documents

Use the preinstalled anydoc CLI when a task needs the contents of a local office document, ebook, CSV, or text-based PDF. This is read-only extraction; it does not edit the source or replace Opengeni's durable artifact tools.

anydoc report.docx
anydoc slides.pptx -o /tmp/slides.md
anydoc - --format csv < data-without-extension

Rules:

1. Supported inputs: .doc, .docx, .docm, .odt, .rtf, .epub,

.pdf, .ppt, .pps, .pot, .pptx, .pptm, .ppsx, .ppsm, .odp, .xls, .xlsx, .xlsm, .xlsb, .ods, and .csv.

1. Let AnyDoc detect the format from file content. Specify --format only for

signature-less input such as CSV on stdin or a file without a useful name.

1. For large output, use -o and inspect only the relevant sections instead

of streaming the whole document into model context.

1. Preserve the source file. AnyDoc output is a derived reading aid, not a new

canonical document or editable artifact.

1. Scanned/image-only PDFs need OCR and are unsupported. Encrypted, malformed,

or resource-limit failures must be reported rather than guessed around.

1. Never install or download AnyDoc at runtime. If anydoc is unavailable,

report that the selected compute image lacks the parsing runtime.

1. AnyDoc's Markdown does not mark PDF pages. When page numbers matter, get

the count with pdfinfo file.pdf and one page's text with pdftotext -layout -f N -l N file.pdf -.