Agent Engineering

From Embeddings to Milvus

A model turns meaning into coordinates; Milvus finds the nearest meanings across millions or billions of those coordinates.

The complete data path
1

Raw content

Questions, documents, images, and other business data.

2

Embedding

One model encodes each item as a fixed-length float vector.

3

ANN search

Approximate nearest neighbors trade a little accuracy for much more speed.

4

Business result

Top-K documents go back to the app or language model.

Why keywords are not enough

Different words, similar meaning

“How do I get my money back?” and “refund procedure” share meaning without sharing many tokens. An embedding captures statistical semantics—not verified truth.

Brute force does not scale

Comparing every vector exactly becomes expensive. ANN narrows the candidate set, so evaluate recall and latency together.

Do not reverse the metric
Metric More similar means Useful intuition
L2 Smaller distance Absolute distance in space
IP Larger score Vector magnitude affects the score
COSINE Larger similarity Direction matters; common for text
Hard requirement: inserts and queries need the same embedding model, preprocessing, and dimension. Index and search must use the same metric.
Milvus stores, finds, and manages

Store

Vectors alongside IDs, source, category, time, and other scalar fields.

Find

Top-K vector search plus scalar filters.

Manage

Collections, indexes, loading, and data lifecycle. Milvus does not create embeddings or write the final answer.

Takeaway Embeddings decide how meaning is represented; Milvus decides how it is stored and retrieved quickly and reliably.