What is vector search?

Vector search is a retrieval method that represents items and queries as embeddings, arrays of numbers produced by a machine learning model, and returns the items whose embeddings are closest to the query’s embedding. Because an embedding model places inputs with similar meaning near each other, a query for “login failure” can find a support ticket titled “can’t sign in to my account” even though the two share no words, which is why vector search is often called semantic search. Large collections are usually searched with an approximate nearest neighbor index, which gives up a small amount of accuracy in exchange for much lower latency.

​

How does vector search work?

A vector search system has an indexing path and a query path, and both must produce vectors in the same embedding space, which usually means the same embedding model.

1. Embed the collection. Each item, or each chunk of a long document, passes through an embedding model that typically outputs one vector.
2. Store and index the vectors. Each vector is saved with an ID and usually some metadata, then organized into an index built for nearest neighbor search.
3. Embed the query into the same embedding space, usually with the same model.
4. Find the nearest neighbors. The system compares the query vector with stored vectors using a distance metric and returns the k closest, called the top k.
5. Use the results. The application displays them, reranks them, or passes them to a language model as context in retrieval-augmented generation (RAG).

​

What role do embeddings play in vector search?

Embeddings are what vector search compares: each item and each query becomes a fixed-length list of numbers, and items with similar meaning end up close together. Vector search only works when stored vectors and query vectors come from the same embedding space, normally the same model or a query and document encoder pair trained together. Vectors from unrelated models, or from two versions of one model, are not comparable. How models produce embeddings, how dimensions and chunking affect them, and what changing models involves are covered in What are vector embeddings? Closeness is computed with a distance or similarity function such as cosine similarity, Euclidean distance, dot product, or Manhattan distance. The right choice is usually the one the embedding model was trained with; see vector distance metrics.

​

What is the difference between exact and approximate nearest neighbor search?

Exact search always returns the true top k, while approximate nearest neighbor (ANN) search uses an index to compare the query with only a small, promising fraction of the vectors, which is much faster at scale but can miss a true neighbor. Both address the k-nearest neighbor (kNN) problem: find the k stored vectors closest to the query. The simplest exact method, brute-force or flat search, compares the query with every stored vector, so its cost grows linearly with the collection. Exact tree indexes such as k-d trees can skip work in low dimensions but approach a full scan at embedding dimensions. ANN indexes differ in how they pick that fraction. Graph-based indexes such as HNSW link each vector to some of its near neighbors and walk that graph toward the query, while IVF (inverted file) indexes cluster the vectors and scan only the clusters closest to the query. What is HNSW? explains how these indexes are built, searched, and tuned.

​

What do recall, latency, and top-k mean in vector search?

Recall measures how closely an ANN search matches exact search. Recall@k is the fraction of the true k nearest neighbors, as found by exact search, that the ANN search returned. A recall of 0.9 at k = 10 means that, on average, nine of the ten true nearest neighbors were found. Most ANN indexes expose a setting that trades recall for latency: search more of the index and you find more true neighbors, but each query takes longer. Top-k is how many results a query asks for. In RAG, k is usually bounded by how much context the language model can use, and many pipelines retrieve a larger k and rerank it down to a smaller set. Recall is measured against exact search over the same vectors, not against what a user considers relevant. A search with perfect recall can still return poor results if the embedding model or the chunking does not capture what the user meant.

​

When does vector search fail?

Vector search is good at paraphrase and fuzzy meaning. It is weak where exact characters or the structure of the data matter:

• Exact identifiers. Order numbers, SKUs, error codes, and email addresses carry little semantic meaning, so a query for one ID can rank a different, similar-looking ID first.
• Names and rare tokens. New product names, internal jargon, and uncommon surnames may be poorly represented in the model’s training data.
• Hard constraints. Dates, prices, ownership, and permissions are conditions to enforce, not similarities to score. “Tickets from last week” is a filter.
• Relationships. “Documents written by people on my team” depends on who is connected to whom, which a text embedding of the document does not encode.

Keyword search with BM25 handles exact terms and rare tokens well, and many systems combine the two in hybrid search. Hard constraints belong in a filter that runs before or during ranking, as described in filtered vector search.

​

How does HelixDB do vector search?

The application computes embeddings; HelixDB stores them as properties on nodes or edges, and a vector index you define over that property enables approximate nearest neighbor search.

• A vector index covers one label and one top-level property, with a fixed dimension and one distance metric: cosine, Euclidean, or Manhattan.
• Search uses approximate nearest neighbor search, and the documentation states over 90% recall.
• Search can run inside an exact candidate set defined by a graph traversal, such as the documents a user can read.
• Within one request, graph traversals, vector search, and text search run in the same ACID transaction over a committed snapshot. HelixDB has no built-in rank fusion, so the application combines vector and BM25 results.

See vector indexes and prefiltered search for working queries.

​

Frequently asked questions

​

Is vector search the same as semantic search?

Mostly. Semantic search names the goal, finding results by meaning, and vector search over embeddings is the most common way to implement it.

​

Do I need an index to run vector search?

Not for small collections. Comparing a query exhaustively against a few thousand vectors is usually fast enough. An ANN index becomes worthwhile when exhaustive search no longer fits your latency budget.

​

Does vector search work for images or audio?

Yes, given an embedding model for that kind of data. Image and audio models produce vectors that are searched the same way as text vectors. Multimodal models map different kinds of input into one embedding space, so a text query can retrieve images.

​

How do you evaluate vector search quality?

Beyond recall@k, build a labeled set of real queries with known good answers and score results with a relevance metric such as precision@k or nDCG. Relevance scores expose problems in the embedding model, chunking, or filters that recall cannot see.

​

Related topics