Vector databases and embeddings
Keep document and query embeddings compatible.
Reviewed: 2026-10-03
Source-based teaching example · Fictional data
Narration: Microsoft Mark · locally synthesized

00:00:00.000 — What you will learn

Keep document and query embeddings compatible.

Visual description:
Keep document and query embeddings compatible.

00:00:05.792 — The idea

Search vectors must come from compatible settings and model versions. Matching vector length alone is not enough. When changing the embedding setup, migrate documents and queries together.

Visual description:
Search vectors must come from compatible settings and model versions. Matching vector length alone is not enough. When changing the embedding setup, migrate documents and queries together.

00:00:20.958 — One useful term

Embedding. A list of numbers that represents an input for tasks such as similarity search.

Visual description:
A list of numbers that represents an input for tasks such as similarity search.

00:00:29.917 — A worked example

The lesson uses two invented two-number vector spaces. Matching pairs score one. Mixing a new query with an old document scores zero, even though both have the same length.

Visual description:
The lesson uses two invented two-number vector spaces. Matching pairs score one. Mixing a new query with an old document scores zero, even though both have the same length.

00:00:44.375 — Your turn to think

Does the same number of dimensions prove two embeddings are compatible? Pause and decide before continuing.

Visual description:
Does the same number of dimensions prove two embeddings are compatible?
Pause the video if you need more time.

00:01:01.917 — Why this answer works

No. Check the model revision and input preparation. These toy numbers explain compatibility, not real model quality.

Visual description:
No. Check the model revision and input preparation. These toy numbers explain compatibility, not real model quality.

00:01:13.417 — Check 1

Record the embedding configuration.

Visual description:
Record the embedding configuration.

00:01:18.583 — Check 2

Reject incompatible queries before search.

Visual description:
Reject incompatible queries before search.

00:01:24.292 — Check 3

Check that deletion reaches indexes and caches.

Visual description:
Check that deletion reaches indexes and caches.

00:01:30.208 — Now try a different case

Open the lesson and try its different practice case. Use every check in the full task. Watching alone does not mark the lesson complete.

Visual description:
Open the lesson
Try the new case
Check and revise
The full task and its requirements are in the linked lesson.