WeRead Assistant

Problem It Solves

WeRead keeps search results, shelves, reading history, highlights, notes, and reviews inside the app. When users ask natural-language questions like “how long did I read this book” or “what popular highlights are in this chapter,” there is no direct path to structured API results. weread-skills connects those requests to the Agent API Gateway, using a unified POST request to call WeRead endpoints selected by api_name, instead of hand-writing auth, pagination, field interpretation, and deep-link assembly.

Core Capabilities and Workflow

• Books and shelves: resolve a book title to bookId via /store/search, then inspect details, chapter catalog, and reading progress; shelf count uses books.length + albums.length + (mp 非空 ? 1 : 0) so albums and audio books are not missed.
• Reading and notes: supports reading time, days, preference summaries, personal highlights, thoughts, bookmarks, chapter hot highlights, and book reviews; timestamps are shown as YYYY-MM-DD, and second-based reading time is converted to “X hours Y minutes.”
• Request constraints: requests include Authorization: Bearer $WEREAD_API_KEY, api_name, and skill_version; business parameters stay flat in the body and should not be wrapped in params or data.
• Result continuity: the workflow keeps bookId in context and builds WeRead app deep links using chapterUid and range when available, allowing jumps to books, chapters, or highlight positions.

Boundaries and Caveats

The skill requires a valid WEREAD_API_KEY, and the APIs are bound to a user identity, so it is best suited for personal shelves, reading data, and personal notes. If a response contains upgrade_info, the client should follow the upgrade guidance before continuing. Field explanations should follow the provided documentation rather than guessing from field names. It does not provide offline data or replace all WeRead app capabilities; it mainly connects structured WeRead APIs to conversational queries.