In this article
Cosine similarity computes the cosine of the angle between two vectors, returning values from -1 (opposite) to 1 (identical). It is the standard similarity metric for embeddings because it is scale-invariant — only direction matters, not magnitude. In vector databases, it is used alongside dot product and Euclidean distance for nearest-neighbour search.
What it means in practice
Cosine Similarity is not just vocabulary; it is a design handle. Use it as a reference point when comparing architecture choices, debugging implementation trade-offs, or explaining system behaviour to another engineer. It helps convert a vague technical conversation into a concrete design question with trade-offs that can be tested.
Why engineers care
- It gives teams a shared name for the behaviour, risk, or architecture choice being discussed.
- It helps separate the goal from the implementation detail, so you can compare alternatives instead of copying a tool pattern blindly.
- It creates a useful checklist for reviews: inputs, outputs, failure modes, ownership, cost, latency, and measurement.
Production watch-outs
Be careful with shallow definitions. The useful meaning usually depends on workload, failure mode, data shape, and who owns the system in production.
Related context
Useful neighbouring concepts: Embedding, Semantic Search, Vector Database.