librarian
nonlinear 제작
Search & Research요약
Conversational interface for semantic book search (companion skill for Librarian project)
설명
이 스킬이 하는 일
A conversational interface for searching a personal book library using semantic embeddings. Ask questions in natural language and get back synthesized answers with page-level citations. Runs as a companion to the Librarian project via a three-layer architecture: AI skill, shell wrapper, and Python search engine.
Returns precise citations with page and paragraph numbers from local files instead of summarizing from memory or requiring a web search.
사용 시점
- Finding what a specific author says about a topic across your library
- Locating passages in a single book by concept or keyword
- Getting cited answers from a topic folder of books
- Cross-referencing ideas across multiple texts without opening each file
- Querying a personal research archive by natural language
예시 워크플로
User asks: What does Graeber say about the origins of money?
- 1
Load .library-index.json and .topic-index.json metadata
- 2
Match query against known book titles and topic IDs to infer scope
- 3
Build shell command: ./librarian.sh "origins of money" topic finance 5
- 4
Parse JSON results from Python search engine
- 5
Synthesize answer and format citations with page numbers
Paragraph summarizing Graeber's argument with inline citations like [Debt: The First 5000 Years, p.21, ¶2] and a sources list
요구 사항
- Librarian parent project installed (github.com/nonlinear/librarian)
설치
clawhub install nonlinear/librarianCLI 명령을 실행하거나, 프롬프트를 Claude Code, Cursor, Codex 등 사용하는 코딩 에이전트에 붙여넣으세요.