Searching Your Library

Beaver uses agentic search. That means Beaver actively explores your library by choosing among different search tools, filtering by metadata, and iterating to find relevant information to answer your question. Understanding these search tools helps you formulate your query and understand what Beaver can and can not find.

Search Tools

Beaver has access to four complementary search tools that it can use flexibly depending on your question:

1. Metadata Search

Searches structured bibliographic information stored in Zotero: author names, publication years, titles, and journal names. This allows Beaver to find specific references when you mention authors, titles, or journals, or when Beaver wants to follow up on information discovered in other papers.

How thoroughly do recent papers in Nature describe their methods and data?
Summarize the main arguments in Piketty's work on inequality

2. Related Reference Search (Semantic)

Uses semantic search to find related references based on titles and abstracts. This tool finds papers that are conceptually similar to a query, even if they use different terminology. For example, searching for "crime" might surface papers on incarceration, policing, or violence. This tool allows Beaver to explore a topic area or find papers on related concepts.

What papers do I have on algorithmic bias?
Which references discuss how neighborhoods shape life outcomes?

3. Full-Document Search (Hybrid)

Searches the complete text of all your processed PDFs using a combination of semantic search (finding passages that match the meaning of the query) and keyword search (finding passages containing specific words). Results are then reranked using a specialized model that considers both relevance and context.

How do papers using PSID data measure wealth?
What are common prompt engineering techniques used in research that rely on large language models for data coding?
Contrast and compare all definitions of social capital in my library.

4. Search Within Documents

After identifying relevant papers, Beaver can search within specific documents to locate particular sections or information. This allows for deep exploration of papers without reading them entirely.

How Beaver Uses These Tools

Beaver combines these tools strategically. For a complex question like "What methods have been used to study neighborhood effects on education?", Beaver might:

  1. Start with Related Reference Search to find 10-15 relevant papers on neighborhood effects and education
  2. Use Search Within Documents to locate the methods sections in those papers
  3. Apply Metadata Search to find additional papers by the same authors
  4. Run Full-Document Search to find specific discussions of methodological approaches

This iterative strategy means Beaver can handle complex questions that would require multiple manual searches, synthesizing information across your entire library.

You don't need to tell Beaver which search tool to use—just describe what you're looking for in natural language, and Beaver will choose the appropriate strategy.