Deep dive · module 89 of 100
Vector databases
Understand how vector databases help with searching through your own documents.
After this lesson
You can explain what a vector database does in a RAG system.
In view · How it connects
Search by similarity
An embedding model converts text into a numerical representation.
The search layer compares the query with stored vectors.
The corresponding text segments are retrieved and checked for usability.
Part 1 of 3 · Text to vector
Your preference for less movement is on. Choose each part yourself.
What is a vector?
A numerical representation of the meaning of a text.
Vector database
Stores these vectors and can quickly find the most similar pieces of text for a question.
In practice
This is the search layer behind many RAG systems. The quality depends on the embeddings and the chunking.
New practice example
From explanation to a usable result.
A vector index searches for matching meaning, but high similarity is not the same as factual accuracy.
Fictional practice situationA fictitious knowledge base contains an old regulation and a new supplement for young people. A search query for 'reserving' mainly finds the old document because it contains the most similar words.
View a possible solution
This is how it can be done
Preserve source and version data with fragments. Also search for conditions and supplements, and check the found passages before you create an answer. If necessary, combine text filters with semantic searches.
Why this works
The index helps find candidates. It does not decide which source is authoritative or up to date.
Make it slightly more difficult
Create two almost identical questions, one for adults and one for young people. What extra source information must accompany the second question?
Extra practice, at your own pace. You do not need to submit anything. This does not count towards an automatic assessment.
Your workplace · learning by doing
Make something. See what changes.
Checking the AI connection… You can already view the written example.
Present your assignment to Ende.
Read the answer, change one thing, and compare.
Previously held
New version
You remain in control
What do you want to improve in your next attempt?
On this device only. AI answers remain only in this page memory. This is self-review, not a grade.
Why this approach works
The index helps find candidates. It does not decide which source is authoritative or up to date.
A step further
Create two almost identical questions, one for adults and one for young people. What extra source information must accompany the second question?
Try & discover · no obligation
Search in a small document bank
Compare small and larger text pieces. Do you also find the exception in the answer?
Found text pieces (maximum three)
- Openingsuren2 keywords found
o'clock. Note the exception: on Friday, the closing is at sixteen o'clock.
- Openingsuren1 keywords found
The fictitious workshop is open Monday to Friday from nine to eighteen
Compare with the full original documents
Openingsuren
The fictitious workshop is open Monday to Friday from nine to eighteen o'clock. Note the exception: on Friday, the closing is at sixteen o'clock. On Saturday and Sunday, the workshop is closed.
Bezoekers
Visitors report to the reception. An appointment is required for a tour. An appointment for a visit does not change the opening hours.
Materiaal
Tools remain in the workshop. Return borrowed equipment on the agreed day. For questions about materials, speak to the manager.
Find the exception yourself: the workshop closes at sixteen o'clock on Friday. Is that information fully present in a found piece? Try 12 words with 4 words overlap.
This example ranks by exact keywords. Real embeddings and vector databases compare numerical representations; these are not performed here. A match is not proof that the entire question has been answered.
This trial works locally with fictitious data and does not count towards your lesson completion. It starts anew with a new lesson or reload.
A prompt to get you started
Briefly explain what a vector database does in a system that provides answers based on own documents.
Apply it
A vector search finds an old passage that strongly resembles the question. A more recent policy version says something else.
Which approach best fits this situation?
- Always choose the highest similarity score as the truth.
- Assume that a vector database only contains current facts.
- Filter for permitted current sources and check which passage supports the answer.
With JavaScript, you can check your approach and receive immediate feedback.
Test your insight
What does a vector database do in a RAG system?
- It quickly finds the most relevant pieces of text for a question.
- It trains the language model.
- It replaces the need for source verification.
Extra practice
Describe in your own words why a vector database is useful when searching through a large number of documents.
It quickly finds the pieces of text that are closest in meaning to the question, instead of just searching on keywords.
