If you are learning about LLMs, RAG, or Vector Databases, you will often hear the word embedding.

The simplest definition is:

An embedding turns data into numbers so a computer can compare meaning.

1. What Is an Embedding?

Computers cannot understand meaning the same way humans do.

For example:

"How can I reset my password?"
"Where can I change my password?"

These sentences use different words, but have a similar meaning.

An embedding model converts text into a vector of numbers:

Text
 ↓
Embedding Model
 ↓
[0.21, -0.73, 0.15, 0.88, ...]

The numbers are not random. Together, they represent patterns and relationships in the text.

So you can think of an embedding as a numeric representation of meaning.

2. Similar Meaning → Similar Vectors

This is the most important idea.

For example:

"dog"
"puppy"
"golden retriever"

These concepts are related, so their vectors should generally be close together.

But:

"dog"
"stock market"

would be much less similar.

This allows computers to compare semantic similarity, rather than only matching exact keywords.

3. Embeddings Enable Semantic Search

Imagine you have documents containing:

"Customers can request a refund within 7 days."

A user asks:

"How long do I have to get my money back?"

There may be no exact keyword match, but the meanings are similar.

The system can:

User Question
     ↓
Embedding Model
     ↓
Query Vector
     ↓
Search Vector Database
     ↓
Find Similar Vectors
     ↓
Relevant Documents

This is called semantic search.

Common similarity methods include cosine similarity, dot product, and Euclidean distance.

4. Embeddings + Vector Database + LLM = RAG

An embedding model is not a database and it does not answer questions.

Each component has a different job:

Embedding Model
→ Turn text into vectors

Vector Database
→ Store and search vectors

LLM
→ Generate the final answer

For example, with 20 PDF files:

PDFs
 ↓
Extract Text
 ↓
Split into Chunks
 ↓
Create Embeddings
 ↓
Store in Vector Database
 ↓
User asks a question
 ↓
Search for relevant chunks
 ↓
Give chunks to LLM
 ↓
Answer

This is the basic architecture of RAG (Retrieval-Augmented Generation).

5. What Happens When Documents Change?

In a real application, documents can be updated.

For example:

refund-policy.pdf
     ↓
SHA-256 Hash
     ↓
abc123

Later, the PDF changes:

refund-policy.pdf
     ↓
SHA-256 Hash
     ↓
xyz789

The different hash tells the system that the document has changed.

You can then:

Delete old chunks
      ↓
Extract new text
      ↓
Create new chunks
      ↓
Create new embeddings
      ↓
Store updated data

For a small collection such as 20 PDFs, this simple approach is usually enough.

The Simple Mental Model

Just remember:

Embedding = Turn meaning into numbers.

Vector Database = Store and find similar vectors.

LLM = Use the retrieved information to generate an answer.

Together:

Documents
   ↓
Chunking
   ↓
Embeddings
   ↓
Vector Database
   ↓
Semantic Search
   ↓
Relevant Context
   ↓
LLM
   ↓
Answer