PDF Librarian

PDFs are now first-class artifacts in myai. Upload one or many, and the platform decomposes each document into navigable, searchable chunks, with full support for scanned, image-only PDFs.

Multi-File Upload Page

A new Uploads page lets you queue multiple PDFs at once. Uploads run sequentially in the background and survive page navigation, so you can keep working while a stack of documents ingests. Progress shows up as a toast, the dialog closes immediately, and the per-file PDF size limit is now 50MB (up from 25MB).

Smart Decomposition

Large PDFs are broken into navigable sub-artifacts with hierarchy and folder structure inferred from the document itself. The pipeline profiles each PDF (table of contents detection, font analysis for header inference, and page-range chunking) then enriches every chunk with a markdown extraction, a Gemini-generated summary, and a vector embedding. Chunks are individually citable, searchable, and shareable, and they render as PDF previews in the artifact viewer.

Scanned and Image-Only PDFs

If a PDF is mostly images (scanned contracts, archived reports, faxes), myai detects it automatically and sends the first 30 pages to Gemini multimodally for structural analysis. The model proposes section breaks with page ranges; if confidence is low, it falls back to fixed page chunking. Either way, you get navigable chunks instead of an opaque blob.

Folder and Search Improvements

Folder navigation now favors drill-down over recursion. list_folders shows immediate children with counts, find_by_folder is shallow by default with an opt-in recursive flag, and semantic search accepts a folder parameter to scope results to a subtree, useful when a single Dimension contains thousands of decomposed pages.

Try It

Open the Uploads page from the sidebar, drag in your PDFs, and watch them appear in the artifacts view as you'd expect, but now with chunk-level depth.