CodeBook
Ask a book and its code together.
Put a PDF book, a GitHub repository and a few documentation pages in one workspace and ask questions across them. Answers come back with page numbers and file:line references you can open in place. The agent decides how to search for each question, and the same sources turn into quizzes, flashcards and summaries.
What you get
- Cited answers
- Every claim points at a page of the book or a file and line range in the code.
- The agent picks how to search
- Vector search, keywords, the table of contents, the call graph or the web, chosen per question and retried when the first pass is weak.
- Code graph
- Callers, dependencies and the path between two functions, built from the repository itself.
- Study tools
- A quiz, flashcards, a mind map or a summary of any source, with the same citations.
- Memory
- What you tell it about yourself carries over between conversations, and you can edit or clear it.
An exchange
Does Express’s next() chain behave like a replication log, where each layer sees the request in order?
- search_sources “router stack next follower log” 42 ms
- read_source lib/router/index.js:139–183 11 ms
Not really. Express walks its router’s stack in order and hands the request to each matching layer, which passes control on with next(); nothing is recorded, so a layer that ends the response simply stops the chain1. A replication log is the opposite: every change is appended durably, so a follower that was down can ask the leader for everything after the last transaction it processed and catch up2. The two share only the idea of ordered processing.