Phrase match — "word1 word2"

Wrap words in double quotes to match them in that exact order, next to each other.

Example: "machine learning" — matches only docs where those two words appear together in that order.

Required term — +word

Prefix a word with + to force it into every result. Other (unprefixed) words still influence ranking but aren't mandatory.

Example: +laptop 15 inch gaming — every result must contain "laptop"; size and "gaming" just boost matching docs.

Exclude term — -word

Prefix with - to drop any doc containing that word.

Example: laptop -refurbished — hides all refurbished laptops from the results.

Combine them

All three operators compose freely in one query.

Example: +laptop "13 inch" -refurbished — must contain "laptop" and the exact phrase "13 inch", excludes anything refurbished.

Why this matters in hybrid search

Hybrid (keyword + vector) search can surface semantically-adjacent items that share dimensions or attributes but aren't what you wanted — e.g. searching laptop 1000 GB may also pull in external hard drives or servers with similar storage specs, because the size signal is strong in the embedding. Using +laptop forces the keyword match and cleans those out instantly.