Summary
Your reading queue as something the assistant can file from and read from.
Saving is the easy half. The half that changes how you work is listing with updatedAfter and full content, which turns an archive into a corpus you can ask questions of. The token is a personal access token, so its reach is the whole Reader account.
What it is
A TypeScript server over the Readwise Reader API. It saves documents, lists them with filters and pagination, and updates them in place. HTML content is converted to clean text through r.jina.ai on the way back, so a saved article arrives readable rather than as markup.
What you get
- A URL, or raw HTML, saved to Reader with tags, a location among new, later, shortlist, archive and feed, and a category among article, book, tweet, pdf, email, youtube and podcast (readwise_save_document)
- Documents listed with filtering on location, category and tags (readwise_list_documents)
- Large collections paged through with pageCursor, and narrowed to what changed since a point in time with updatedAfter
- Full document content requested alongside metadata through withFullContent and withHtmlContent
- Documents updated in place, including location and tags (readwise_update_document)
- HTML converted to clean text automatically, so what comes back is fit for a model to read
Requirements
Node with a local build, npm install then npm run build, and a Readwise access token in the server's environment, taken from readwise.io/access_token.