Supported formats
- PDF — best results when the PDF has selectable text. Scanned PDFs need OCR first.
- DOCX — Word documents with proper heading styles produce the cleanest chunking.
How to upload
- Go to Library.
- Click + Upload Book.
- Drag the file in, optionally edit the auto-detected title and author.
- Click Upload.
Processing kicks off in the background. You'll see a progress bar — feel free to close the tab. Typical processing time is 1–3 minutes; large reference books may take up to 10 minutes.
What happens during processing
- Extraction — we read the raw text and table of contents.
- Structure detection — chapters and sections are identified.
- Chunking — content is split into post-sized segments respecting paragraph boundaries.
- Indexing — chunks are stored ready for the scheduler.
Tips for better results
- Use the original publisher PDF when possible — it usually has the cleanest text layer.
- Books with clear chapter headings produce more coherent chunks.
- If chunking comes out weird, try AI Reprocess on the book detail page.