Skip to content
Yomitomo

AI Reading Companions and Assistants

Configure specialized reading companions and in-depth review assistants to enrich highlights, discussions, and distilled notes.

Yomitomo’s AI capabilities are anchored directly to your source text. Every generated reflection, reply, and review comment references specific paragraphs, selections, or distilled drafts—allowing you to verify claims against original evidence at any time.

Yomitomo assistants

There are three primary ways to interact with AI assistants during reading:

  1. Highlight Reflections: Create a highlight on selected text to receive instant contextual thoughts from an assistant.
  2. Discussion @ Mentions: Type @assistant_name in any highlight discussion thread to invite a specialized persona into the conversation.
  3. Distillation Review: In the distillation window, select review assistants to analyze your draft and generate structured improvement suggestions.

Assistant responses remain strictly grounded in the document context, your notes, and the active discussion—avoiding detached, generic summaries.

Each assistant keeps its own conversation memory for a document, so earlier thoughts, discussions, and Q&A carry over into later replies. When generating thoughts, replies, or reviews, assistants look up what they need within the readable range:

  • The document’s outline and source paragraphs, including full-text search
  • Your historical annotations, questions, and discussion records
  • Reading memory built across previous reading sessions

For EPUB ebooks, a request started from a highlight lets the assistant read earlier chapters and the whole chapter containing the highlight. Questions without a selection stop at your reading progress. Source text lookup, reading memory search, and thread reading all share this same readable boundary, so passages, annotations, and discussions in later chapters are not available to assistants.

Passages and discussions cited in replies appear as links: a passage link moves the reader to the source text (for PDFs, to its page), and a discussion link opens that discussion. Discussions and Reader Q&A show steps such as reading the source text. Closing a window does not stop a reply in progress.

When a reply fails, the assistant shows the cause, such as a missing model, an invalid API key, insufficient balance, rate limiting, or a network failure, with a retry button and a link to Model settings. Reader Q&A can stop generation, start, switch, or clear sessions, and copy individual answers.

Reading companions focus on co-reading, concept clarification, deeper questioning, and structural analysis:

Assistant Persona Key Specialty
June Hartley Marginal Co-reader Clarifying core concepts and adding context
Gideon Frost Root-Cause Reader Examining premises and tracing causality
Maya Brooks Question Mentor Turning vague confusion into precise questions
Marcus Reed Insight Editor Extracting transferable mental models
Iris Chen Concept Translator Explaining terminology and historical context
Daniel Park Structure Guide Dissecting argument flow and paragraph roles

Review assistants specialize in factual evidence, logical rigor, clarity of expression, and actionable outcomes:

Assistant Persona Key Specialty
Arthur Whitfield Evidence Librarian Checking facts and evidence chain validity
Hannah Wells Reader Advocate Safeguarding reader clarity and questions
Julian Cross Final Editor Trimming redundancy and sharpening prose
Simone Carter Logic Reviewer Detecting cognitive gaps and leap of logic
Victor Tan Risk Reviewer Flagging overgeneralizations and biases
Grace Kim Action Calibrator Validating feasibility of concrete steps

You can enable or disable specific assistants and inspect their full prompt configurations on the Assistants page.

Still need help?

GitHub IssuesX
Contents
On this page