Turning words into numbers
An embedding places each piece of text as a point in a “meaning space”. Similar meaning → nearby points.
What is an embedding?
OpenAI turns a chunk of text into a list of 1536 numbers that capture its meaning. Two texts about the same thing get similar numbers — so they land close together.
Two different OpenAI models
Embeddings — text → numbers (for search)
Chat — writes the answers
The key point: the numbers are used only to search, inside our database. OpenAI never compares them. When writing the answer, only text is sent.
A “meaning map” (simplified to 2-D)
“shipping dates”
“budget”
“hiring”
Your question
the nearest ones are chosen
The database just finds the nearest points to your question. That is the whole “search” step — fast, and entirely on our side.