Supported Formats

Supported local inputs #

Format	Extensions	Main behavior
PDF	.pdf	Extracts text and page structure; uses local OCR for scanned pages.
Plain text	.txt, .text and compatible plain-text types	Preserves readable text.
Markdown	.md, .markdown and compatible types	Preserves Markdown as authoritative source content.
HTML	.html, .htm	Extracts local HTML structure without loading remote resources.
Rich Text	.rtf	Extracts styled text into Markdown-compatible structure.
Images	.png, .jpg, .jpeg, .tif, .tiff, .heic, .heif	Uses local OCR.
Microsoft Word	.docx	Extracts supported headings, paragraphs, lists, tables, and existing image handling.
Microsoft PowerPoint	.pptx	Extracts slide boundaries, titles, lists, text, and tables.
Microsoft Excel	.xlsx	Extracts sheet boundaries and tables.

The Safari extension also captures web pages, selections, highlights, and multiple tabs from the current window. It can discover and acquire website llms.txt collections under Safari’s website permissions. Local llms.txt import remains offline and accepts an index plus contained relative .md, .markdown, and .txt links.

iPhone and iPad #

From Safari, SourceShelf captures authorized web pages and imports llms.txt research collections. From the Share Sheet, Files, or Import Research…, it accepts:

• PDF, including text PDFs and scanned or mixed PDFs using local Vision OCR only where native text is insufficient;
• images decoded by Apple’s image frameworks, including JPEG, PNG, HEIC/HEIF, and TIFF;
• RTF, Markdown, plain text, and standalone llms.txt;
• AI Reference Pack ZIP, OKF v0.2 Bundle ZIP, and Portable llms.txt Package ZIP exports.

PDF and RTF imports preserve canonical Markdown rather than the original document, matching the Mac conversion policy. An imported image preserves its original bytes in the source’s private assets sidecar, so it remains visible offline; any embedded metadata in those original bytes remains local. Recognized PDF/image text and converted RTF text participate in normal library search.

All conversion and OCR happen on the device. Password-protected PDFs fail cleanly without creating a source. RTFD, Office documents, HTML-file conversion, and folder import remain unsupported on iPhone and iPad. The Share extension deliberately does not register for arbitrary public.data; the explicit in-app picker can safely recognize a mislabeled supported file from its signature or actual decodability.

Not supported directly #

Legacy Office .doc, .ppt, and .xls files are detected but not converted. Save them as .docx, .pptx, or .xlsx first.

Current limitations include advanced spreadsheet formula evaluation, PowerPoint animation/media/speaker notes, newly extracted embedded Office media, Word footnotes and tracked revisions, perfect OCR, and complex scanned layout preservation.

Output formats #

All conversions create Markdown. On Mac, packs can produce AI Reference Pack ZIP, OKF v0.2 Bundle ZIP, Markdown Context Pack, Portable llms.txt Package ZIP, llms.txt Collection Folder, or clipboard Markdown. On iPhone and iPad, packs can produce AI Reference Pack ZIP, OKF v0.2 Bundle ZIP, or Portable llms.txt Package ZIP through the system share sheet.