Sources & Content Library · Add Sources
Upload a PDF
Drag-drop a PDF (or click to pick) and Ritsu turns it into Points of Knowledge in under a minute.
2 min read
PDFs are the most common source on Ritsu. Textbooks, papers, lecture slides exported as PDF — they all work. Most uploads finish in under a minute.
Steps
- Open Sources from the left sidebar.
- Click Upload file (or drag the PDF onto the page).
- Wait for processing. A progress bar shows: parse → extract → embed → done.
- When it's done, the source appears in your library with a list of Points of Knowledge auto-extracted (one per chapter or section).
- Click any PoK to open the Learning Chat scoped to it.
What Ritsu does with the PDF
- Parse — extracts text + structure (headings, lists, tables, captions). Fonts and layout are stripped; we keep the meaning.
- OCR if needed — for scanned PDFs (no embedded text), we run OCR. Adds ~30 seconds for a 100-page book. Works in EN, VI, and 30+ other languages.
- Section detection — Ritsu finds chapter / section boundaries from headings + table of contents. Each section becomes one PoK.
- Embed — content is embedded for fast semantic search across the book.
/searchworks across the whole source.
Limits
| Item | Free | Pro |
|---|---|---|
| PDF size | 50 MB | 200 MB |
| Pages per PDF | 500 | 2000 |
| Total storage | 1 GB | 50 GB |
| Files per source | 1 | 1 |
A typical textbook chapter PDF is 5-30 MB / 30-80 pages. A full textbook is 50-200 MB / 400-1000 pages. Both work fine on Pro.
OCR quirks
- Math notation comes through as best-effort. Equations from LaTeX-rendered PDFs are nearly perfect; equations from scanned books may have errors. Run
/explain "the equation on page X"if you suspect garbled OCR. - Two-column layouts read correctly. Tables read top-to-bottom, left-to-right.
- Diagrams + figures are extracted as images and indexed by their captions. You can ask "show me figure 3.2" and Ritsu will surface it.
Tips
- Smaller PDFs make better PoKs. A 50-page chapter produces sharper PoKs than a 800-page tome. If you only need chapters 4-6, export just those pages first.
- Watermarks confuse OCR. Heavily-watermarked PDFs (publisher samples, library copies) sometimes have text mixed with watermark fragments. Workaround: re-export from the source.
- Encrypted PDFs must be unlocked first. We won't process password-protected files.
Trouble
If upload fails, see Upload failed. For a full file-types list, see Supported file types and size limits.
Was this article helpful?