Snail.Toolkit.Documents 0.1.0

Snail.Toolkit.Documents

Reads PDF, Word, Excel, PowerPoint, HTML, CSV, JSON, Markdown and text into one Markdown text, with its pages, headings and tables — in .NET, with no Office, no LibreOffice and no network.

var reading = await DocumentReader.Every.ReadAsync(File.OpenRead("invoice.pdf"), "invoice.pdf");

if (reading.TryGetDocument(out var document))
    Console.WriteLine(document.Text);
else
    Console.WriteLine($"{reading.Failure}: {reading.Reason}");

What it reads

Text files are read as UTF-8, or by their byte order mark, and as Windows-1251 when they are not valid UTF-8 — the way a Russian Excel saves a CSV.

A document

Document is the whole text as Markdown and where everything in it stands: Parts (pages, sheets, slides, each with where its text starts and ends), Sections (each heading with the stretch of text it opens, to the next heading of its level or above) and Tables. The positions are offsets into Text, taken in the same pass that wrote it, so a page a caller finds in the outline is the page it reads.

A sheet's table keeps where its cells stood: Table.Addresses.At(row, column) is B7, and the Markdown shows it.

## Sheet 1: Invoices

|  | B | C | D |
| --- | --- | --- | --- |
| 2 | Number | Date | Sum |
| 3 | INV-7 | 2026-10-14 | 118000.5 |
| 5 | INV-8 |  | 42 |

Failures are answers

ReadAsync throws only when its caller cancels it. Every other way a file fails — including whatever a format's own library throws on a damaged one — comes back as a Reading with the reason in the reader's words:

The limits hold whatever the size of the file: a workbook of 11 kB can come to 98 million characters, and a zip of a few kilobytes to gigabytes. Each limit is stated with the file that found it in ReadingLimits.

Scanned pages

A PDF page without a text layer is a Part with IsScanned. Its picture is what a model that sees reads:

int[] scans = [.. document.Parts.Where(part => part.IsScanned).Select(part => part.Number)];
IReadOnlyList<ScannedPage> pages = ScannedPages.Pictures(File.OpenRead("scan.pdf"), scans);

Each page comes back as a PNG or a JPEG, or without a picture when no managed decoder opens it (fax and JBIG2 scans) or when it is past ReadingLimits.MostPicturePixels, fifty million pixels.

In a container

services.AddDocumentReading();   // every IFormatReader, and the DocumentReader over them

A program that reads fewer formats registers its own IFormatReaders and a DocumentReader over them.

The public API: docs/reference. Built on Snail.Sdk: the rules, the public API kept in src/PublicAPI.*.txt, validation against the last release, the shared workflows.