Database / LanceDB Interview questions
What is full-text search in LanceDB?
Full-text search (FTS) in LanceDB lets you search a text column for keyword matches using a BM25-based ranking algorithm, complementing vector similarity search with traditional, exact keyword-based retrieval on the same table.
table.create_fts_index("text") results = table.search("machine learning", query_type="fts").limit(5).to_list()
Under the hood, LanceDB's full-text search is built on Tantivy, a Rust-based search engine library, and building an FTS index on a text column lets keyword queries run efficiently without scanning every row's raw text at query time, similar in spirit to how a vector index avoids scanning every vector.
Full-text search matters especially for queries where exact terms matter more than semantic similarity — searching for a specific product code, an exact error message, or a person's name is often better served by keyword matching than by a vector search, which can sometimes surface semantically related but literally different results instead.
More Related questions...
