Ask your own textbook
and get your syllabus back.
Add your books and notes — or photograph the pages of a physical textbook. Grout reads them, indexes them locally, and answers questions from that material instead of from whatever a model absorbed during training.
What the module does
A file manager for AI context, organised the way a syllabus is organised.
Subjects, books and chapters
Organise your material the way a syllabus is organised. A subject holds textbooks, notes and other sources; a textbook splits into chapters; a chapter holds its pages.
OCR on photographed pages
Photograph a chapter from a physical book. Grout reads the pages, extracts the text and keeps a thumbnail of each page alongside it.
Semantic search across everything
Your material is embedded into a local vector index, so you can ask a question in your own words and get the passage that answers it — not a keyword match.
Answers grounded in your book
The chapter is passed to the model as context, so answers come from your syllabus rather than from whatever the model absorbed during training.
Memory that persists
The assistant keeps long-term memory across sessions in a local vector store, so it remembers what you have already covered instead of starting cold every time.
Chapters exported as PDFs
Build a clean PDF from a chapter or a whole book you have captured, for revision or for sharing with the class.
Generate from your own material
Summaries, practice questions and explanations built from the chapter in front of you, which is exactly the material that is safe to generate questions from.
The index lives on your disk
Embeddings and the vector store are local files. Your textbooks, your notes and your questions are never uploaded to be indexed by someone else.
Getting your material in
Set it up once per subject at the start of term.
- 1
Create a subject
One per subject you are studying or teaching — Physics, History, whatever the syllabus calls it.
- 2
Add your sources
Drop in PDFs, or photograph the pages of a physical textbook. Notes and handouts go in the same place.
- 3
Split it into chapters
Group pages into chapters so the assistant can be pointed at exactly the part of the book you are working through.
- 4
Let it read and index
Grout runs OCR on image pages and builds a local embedding index over the text. This happens on your machine.
- 5
Ask questions against the chapter
Ask in your own words. The answer is grounded in the chapter you selected, and you can see which pages it came from.
Why grounding in your own book matters
A general model knows a great deal about photosynthesis. It does not know which definition your board expects, which four factors your textbook lists, or which worked example your teacher will assume you have seen. Those are the things exams are actually built on.
Retrieval fixes that. When the chapter is supplied as context, the answer comes back in the vocabulary of your syllabus, using your textbook’s framing — which is what you need for revision, and what makes generated practice questions worth answering.
Two practical consequences
- Generated questions become trustworthy. Questions written from material you supplied can be checked against that material in seconds. Questions written from a model’s memory cannot.
- Chapters cost fewer tokens than you would expect. A chapter is passed as compressed text rather than as one full-resolution image per page, so a ten-page chapter is cheap enough to keep in context on ordinary hardware.
Frequently asked questions
Can I chat with my own textbook using AI?+
Yes. Grout’s AI-Memory module lets you add a PDF textbook, or photograph the pages of a physical one, split it into chapters and then ask questions answered from that material. The text extraction, the embedding index and the model all run on your own computer.
Does this work with a physical textbook?+
Yes — that is the point of the OCR step. Photograph the pages, and Grout reads them, keeps a thumbnail of each page and indexes the extracted text. You do not need a digital copy of the book.
How is this different from pasting a chapter into a chatbot?+
Three differences. Your material is organised persistently by subject, book and chapter rather than pasted per conversation; a vector index lets you search across everything you have added instead of only the chunk you pasted; and none of it is uploaded — the index and the model are both local files on your disk.
Does an AI textbook assistant work offline?+
Yes. The embedding model, the vector index and the language model all run on-device, so once your material is added the whole thing works with the network disconnected.
Will it make things up about my textbook?+
Grounding an answer in a retrieved chapter reduces fabrication substantially compared with asking a model from memory, but it does not eliminate it. Answers show which pages they drew on so you can check. For anything that matters, read the cited page.
How much material can I add?+
The index is a local file, so the practical limit is your disk rather than a plan tier. A full year of textbooks and notes across several subjects is a normal amount.
Put your syllabus behind the assistant
Free download for Windows and macOS. Your books, your notes, your machine.
Download Grout free