Three answer depths
Every Tutor composer — full workspace and reader panel — now has a depth control under the input. **Quick** keeps the fast, concise answers you already know. **Standard** uses a stronger model, more context from your library, and up to 4,000 tokens of output. **Deep** (Premium) targets long-form analysis with up to 8,000 tokens and prompts tuned for thoroughness.
- Quick — flash-lite, best for follow-ups and flashcards
- Standard — fuller answers without a Premium plan
- Deep — Premium; best for textbooks and detailed writing
Long PDFs: map-reduce summaries
A 200-page file cannot fit in one model call. When you use Deep mode (or `/deep-summary` on a page), Shelf reads indexed sections in order, summarizes each part, then merges them into one structured answer. You will see progress like “Reading section 3 of 18…” while it works.
Use `/notes` for structured notes or `/analyze` for thematic analysis with examples from your file.
Tips
Your depth choice is remembered in the browser. Deep costs more monthly tokens — check Settings for usage. For a single chapter, Standard is often enough; reserve Deep for full-book revision or judgment analysis.
Standard vs Quick in practice
Quick mode is still the default for everyday questions — define a term, check a formula, or queue a follow-up while a reply streams. Standard is the sweet spot when you want a complete explanation with tables and a recap without waiting for a full-book pass.
Switch depth before you send; it applies to that message only in the thread (your choice is remembered for the next question). Slash commands like /explain and /compare respect the depth you picked.
Deep mode and Premium
Deep requires Premium because it runs more model calls and a higher output token ceiling. Free users can still use Standard for noticeably longer answers than Quick.
Token usage scales with depth — a Deep map-reduce summary on a long textbook may consume several times a Quick reply. Check Settings for your monthly Tutor allowance.
Slash commands for long outputs
/deep-summary on an open page triggers chapter-wise synthesis when the file is long. /notes, /analyze, and /mains expand into long-form prompts tuned for revision and exam tracks.
Combine depth with scope: library chat with Standard can pull across a collection; reader Deep focuses on the indexed chunks of the open PDF.
- /deep-summary — full document, section by section
- /notes — revision notes and must-remembers
- /mains — long-form answer skeleton
- /analyze — themes, evidence, gaps
What Deep does not replace
Tutor still grounds answers in indexed text. Scanned PDFs without OCR remain thin until processing finishes. Deep reads more of what is indexed — it does not invent pages you never uploaded.
For the hardest judgment or proof work, pair Deep summaries with your own highlights and planner sessions. The depth toggle makes answers longer and more structured; you still choose what to memorize.