Skip to content

Effective Search

Trove uses semantic search, not keyword matching. Write queries as natural language questions or phrases. Trove matches on meaning, not exact words.

GoalUseExample
Find specific contentsearch / trove_search”papers about attention mechanisms in transformers”
Explore a topic broadlydiscover / trove_discover”AI safety”
See what’s newrecent / trove_recent(no query needed)
Read full contentdocument query / trove_get_document(pass a document ID from search results)

Describe what you are actually looking for. Specific queries return higher-relevance results.

Instead ofTry
”AI training""how does RLHF work in large language models"
"startup advice""when should a startup raise a Series A"
"programming""error handling patterns in Rust async code”

Narrow results when you know something about the source.

  • By connector. “Search my Readwise highlights for…”
  • By author. “Find articles by Paul Graham about…”
  • By date. “What did I save last week about…”
  • By content type. “Search my transcripts for…”
  • By tags. “Find everything tagged ‘machine-learning’…“

Search returns snippets, the most relevant chunk from each document. When a snippet looks promising, use trove_get_document (or the document GraphQL query) to read the full text. This two-step pattern keeps search fast while giving you full content on demand.

discover uses a lower similarity threshold than search. Use it when you want:

  • Loosely related content across different sources
  • A broader view of what you have saved on a topic
  • Connections between ideas that a precise search would miss

recent does no semantic matching. It returns documents sorted by index time, newest first. Good for:

  • Checking what was synced today
  • Reviewing recently saved content
  • Keeping up with active connectors

Use after and before to find content from a specific time period. Dates apply to the content’s original creation date, not when it was indexed.

When you know where the content lives, filter by connector name. Searching “within my arXiv papers” is faster and more precise than searching everything.

Every search and discover result includes a relevance score from 0.0 to 1.0.

ScoreInterpretation
0.9+Strong match. Directly answers your query
0.7 to 0.9Good match. Clearly related content
0.5 to 0.7Related but tangential
Below 0.5Weak match. Only loosely connected

If scores are low across all results, rephrase your query or use discover for broader exploration.