DuckDB, the in-memory analytical database beloved for its speed and simplicity, now offers full-text search capabilities through its FTS extension. Developer Peter Doherty published a comprehensive guide exploring these features, comparing them against established solutions like Elasticsearch and PostgreSQL. The extension supports stemming, stop word removal, accent normalization, and Okapi BM25 ranking algorithms. For users already working within the DuckDB ecosystem, this provides a convenient way to add search functionality without deploying additional infrastructure.

Doherty tested the FTS extension on a corpus of 13,010 .eml email files, preprocessing them with Python and BeautifulSoup before importing into DuckDB. The setup requires just two commands: INSTALL fts; LOAD fts;. Creating an FTS index involves a simple PRAGMA statement, and queries use the match_bm25 function to rank results by relevance. He notes the feature set is solid for basic use cases but lacks query term highlighting—a capability PostgreSQL provides via ts_headline, which he found himself missing during his experiments.

The author identifies several areas for potential improvement: phrase queries, vector support, and pluggable synonym dictionaries. Despite these gaps, DuckDB's FTS provides a lightweight option for users already embedded in the DuckDB ecosystem, particularly those working with structured data who need basic search without additional tools. The full implementation details and code examples are available on Doherty's website.