Heptabase can work with a book as a PDF or as imported text documents. For long books, use bounded page ranges instead of sending the entire file to AI Agent at once.
Choose a file format
PDF: The best option when you want to preserve page layout, read and highlight the book, or select exact page ranges for AI.
EPUB: Native EPUB reading and parsing are not currently supported. Convert the book to PDF first.
Text documents: If the book is mainly text, you can convert individual chapters to
.md,.txt, or.docx, then choose Import → Text documents at the bottom of the left sidebar.
Each uploaded file can be up to 2GB. To parse a PDF, it must be no larger than 350MB and have no more than 3,000 pages. A PDF may upload successfully but still be unable to parse because it exceeds either per-file limit. If it exceeds 3,000 pages, split it into files within both limits. If it exceeds 350MB, compress it or split it into files within both limits.
Recommended workflow
For a summary or questions about a chapter, follow the page-range workflow below. For a more detailed breakdown into cited whiteboard content, use Break down book as described in the next section.
Open and parse the PDF.
Start a new AI conversation through Ask AI.
Use the page-range context selector to add one chapter or about 20 pages.
Ask AI to create a chapter summary, key concepts, questions, or cards for your whiteboard.
Review the result before processing the next section.
After finishing the sections, start a new conversation with the chapter summaries and ask AI to synthesize the whole book.
About 20 pages is a conservative starting point, not a fixed product limit. If several sections work successfully, you can gradually increase the range.
Create a whiteboard with citations
The Break down book Skill helps turn a book or chapter into detailed whiteboard content backed by citations. Use it when you want to study the structure and ideas in depth, rather than only receive a short summary.
In a desktop or web AI conversation, attach the PDF or chapter you want to work with. For a PDF, select the page range for the first section you want to process.
Type
/break-down-bookin the chat input and choose Break down book.Specify the source and the chapter or page range to process. Start with one section, then review the resulting whiteboard content and citations before continuing.
For example, after attaching the relevant PDF pages:
“Use Break down book on pages 10–20 of this PDF only. Organize the chapter’s main ideas and supporting explanations on a whiteboard with citations. Stop after this section so I can review it.”
This Skill performs AI analysis and creates content, so it uses AI credits when using Heptabase’s AI allowance. It does not bypass context limits or guarantee that a whole book can be processed in one request. If you use your own API key, your provider’s billing and limits apply instead.
AI credits and a low-risk first test
PDF parsing and AI analysis are separate steps. Running Parse extracts text and OCR content and does not use AI credits. Asking AI to identify chapters, create an outline, summarize pages, or create cards does use AI credits. There is no prompt that can automatically divide a book into meaningful chapters without using AI credits.
For a cautious first test, start a new conversation through Ask AI, use the page-range context selector to add only 5–10 pages, and try:
“Please analyze only pages 1–10. Identify the main sections and return a concise outline with their page ranges. Do not create or edit any cards, and do not continue beyond these pages.”
This request still uses AI credits. After it finishes, go to Settings → AI Features → View usage to check the cost before processing the next range.
There is no reliable fixed estimate for a 100-page book. Usage varies with the selected model, text density, images, tables, Skill instructions, attachments, and existing conversation history. Images and graphics do not change the fact that PDF parsing itself is credit-free.
Context limits and token usage
There is no fixed page or word count that guarantees success. Text density, images, tables, Skill instructions, attachments, and existing conversation history all use context.
If you use your own API key, requests use your provider project’s usage, billing, and rate limits and do not consume Heptabase AI credits. See How can I get an API key to use AI in Heptabase?
If the request is too large
Start a fresh conversation and select a smaller page range.
Avoid repeatedly pressing Continue in a conversation that has already reached the context limit.
If smaller page ranges still fail, split the PDF into smaller files.