# Intro to vector embeddings

**Vector embeddings** are numerical representations of data, such as text, images, or events, that capture their meaning in a format that computers can efficiently compare and analyze.

In practice, an embedding is a list of floating-point numbers (a vector) that encodes the semantic relationships between inputs. Similar inputs produce similar vectors.

For example:

| Input                     | Embedding (simplified)        |
|----------------------------|-------------------------------|
| `"error: failed to connect"` | `[0.11, -0.42, 0.87, …]`     |
| `"connection refused"`       | `[0.10, -0.39, 0.85, …]`     |
| `"user logged in"`           | `[-0.72, 0.14, -0.33, …]`    |

The first two messages have similar vectors because they describe related concepts — both are connection errors — while the third one is far apart.

## Why embeddings matter

Embeddings allow you to perform **semantic search**, **clustering**, **anomaly detection**, and **AI-powered analytics** on your data without needing to write explicit rules or string matches.

In Better Stack Warehouse, you can use embeddings to:

* Search events by **meaning, not just keywords**
* Implement [Retrieval Augmented Generation (RAG)](https://betterstack.com/docs/warehouse/use-cases/building-rag/) systems with just two HTTP requests
* Detect outliers or unusual patterns in high-volume data
* Power **AI-driven summaries and recommendations**