Post Similarity search using vector databases by iocanel · Pull Request #2261 · quarkusio/quarkusio.github.io

iocanel · 2025-03-18T11:45:42Z

This is a post on how to use quarkus with vector databases to implement a similarity search example.

github-actions · 2025-03-18T11:51:52Z

🙈 The PR is closed and the preview is expired.

geoand

Very neat! I've added some comments but I would like @jmartisk to also review

iocanel · 2025-03-19T11:58:55Z

@geoand applied feedback.

gsmet

Nice article, I spotted a few typos here and there, HTH.

gsmet · 2025-03-19T17:15:18Z

+With LLMs becoming increasingly popular we often see them being used even for tasks that are not directly related to text generation. 
+Such case is using LLMs for recommendation systems. In this post we'll see how you can build such a system using https://docs.quarkiverse.io/quarkus-langchain4j/dev/index.html[Quarkus Langchain4j]
+but without using LLMs. More specifically we'll create a simple movie similarity search system using a vector database. The role
+of https://docs.quarkiverse.io/quarkus-langchain4j/dev/index.html[Quarkus Langchain4j] in this store is to abstract the underlying vector database through the `EmbeddingStore` interface.


Wasn't sure what you wanted to write but store looked odd?

Suggested change

of https://docs.quarkiverse.io/quarkus-langchain4j/dev/index.html[Quarkus Langchain4j] in this store is to abstract the underlying vector database through the `EmbeddingStore` interface.

of https://docs.quarkiverse.io/quarkus-langchain4j/dev/index.html[Quarkus Langchain4j] in this story is to abstract the underlying vector database through the `EmbeddingStore` interface.

gsmet · 2025-03-19T17:16:04Z

+but without using LLMs. More specifically we'll create a simple movie similarity search system using a vector database. The role
+of https://docs.quarkiverse.io/quarkus-langchain4j/dev/index.html[Quarkus Langchain4j] in this store is to abstract the underlying vector database through the `EmbeddingStore` interface.
+
+A relevant sample has been recently added to the https://github.com/quarkiverse/quarkus-langchain4j/tree/main/samples/[Quarkus Langchain4j samples].


Could you fix LangChain4j case everywhere?

gsmet · 2025-03-19T17:17:14Z

+    </dependency>
+----
+
+To be able to use these dependencies without needing to specify versions, the bom can be add imported to the `dependencyManagement` of the project:


Suggested change

To be able to use these dependencies without needing to specify versions, the bom can be add imported to the `dependencyManagement` of the project:

To be able to use these dependencies without needing to specify versions, the bom can be added to the `dependencyManagement` of the project:

gsmet · 2025-03-19T17:17:38Z

+</dependency>
+----
+
+To properly use the in process embedding model we need to configure it in the `application.properties` file. 


Suggested change

To properly use the in process embedding model we need to configure it in the `application.properties` file.

To properly use the in-process embedding model we need to configure it in the `application.properties` file.

gsmet · 2025-03-19T17:17:50Z

+----
+
+To properly use the in process embedding model we need to configure it in the `application.properties` file. 
+We also need to configure the pgvector dimension an ensure it's aligned with the dimension of the embedding model.


Suggested change

We also need to configure the pgvector dimension an ensure it's aligned with the dimension of the embedding model.

We also need to configure the pgvector dimension and ensure it's aligned with the dimension of the embedding model.

gsmet · 2025-03-19T17:24:20Z

+}
+----
+
+To use the CSV mapper, we'll need to `jackson` csv dataformat:


Suggested change

To use the CSV mapper, we'll need to `jackson` csv dataformat:

To use the CSV mapper, we'll need to add Jackson's CSV dataformat dependency:

gsmet · 2025-03-19T17:25:04Z

+
+==== Bringing it all together ====
+The only thing that's left is to create a REST endpoint that will allow us to search for similar movies. We could also use a simple UI.
+Let's start with the REST endpoint. It's pretty straight forward. We need to methods one for movie searching and one for searching similar movies.


Suggested change

Let's start with the REST endpoint. It's pretty straight forward. We need to methods one for movie searching and one for searching similar movies.

Let's start with the REST endpoint. It's pretty straightforward. We need two methods, one for searching movies and one for searching similar movies.

gsmet · 2025-03-19T17:25:26Z

+
+The key elements of that page are:
+
+* movie-box: a text filed for entering the movie title


Suggested change

* movie-box: a text filed for entering the movie title

* movie-box: a text field for entering the movie title

gsmet · 2025-03-19T17:25:45Z

+* movie-poster: an image for displaying the movie poster
+* similar-results: an additional unordered list for displaying the similar movies
+
+It's important to remember that the `Movie` entity is using `jackson` to map the CSV columns to the entity fields.


Suggested change

It's important to remember that the `Movie` entity is using `jackson` to map the CSV columns to the entity fields.

It's important to remember that the `Movie` entity is using Jackson to map the CSV columns to the entity fields.

gsmet · 2025-03-19T17:26:05Z

+</html>
+----
+
+I won't go into much detail about the hmtl code as it's outside the scope of this post. 


Suggested change

I won't go into much detail about the hmtl code as it's outside the scope of this post.

I won't go into much detail about the HTML code as it's outside the scope of this post.

iocanel · 2025-03-26T08:38:03Z

@gsmet @geoand: Forgot to mention that I've applied the feedback.

iocanel force-pushed the similarity-search-using-vector-databases branch 2 times, most recently from ea893ff to 5008cb9 Compare March 18, 2025 14:15

iocanel requested a review from geoand March 19, 2025 09:36

geoand reviewed Mar 19, 2025

View reviewed changes

iocanel force-pushed the similarity-search-using-vector-databases branch from b99b1de to 6a4a667 Compare March 19, 2025 11:58

gsmet reviewed Mar 19, 2025

View reviewed changes

post: Similarity search using vector databases

137c925

iocanel force-pushed the similarity-search-using-vector-databases branch from 6a4a667 to 137c925 Compare March 22, 2025 16:59

geoand approved these changes Mar 26, 2025

View reviewed changes

iocanel merged commit 55f2c32 into quarkusio:main Mar 27, 2025
1 check passed

	of https://docs.quarkiverse.io/quarkus-langchain4j/dev/index.html[Quarkus Langchain4j] in this store is to abstract the underlying vector database through the `EmbeddingStore` interface.
	of https://docs.quarkiverse.io/quarkus-langchain4j/dev/index.html[Quarkus Langchain4j] in this story is to abstract the underlying vector database through the `EmbeddingStore` interface.

	To be able to use these dependencies without needing to specify versions, the bom can be add imported to the `dependencyManagement` of the project:
	To be able to use these dependencies without needing to specify versions, the bom can be added to the `dependencyManagement` of the project:

	To properly use the in process embedding model we need to configure it in the `application.properties` file.
	To properly use the in-process embedding model we need to configure it in the `application.properties` file.

	We also need to configure the pgvector dimension an ensure it's aligned with the dimension of the embedding model.
	We also need to configure the pgvector dimension and ensure it's aligned with the dimension of the embedding model.

	To use the CSV mapper, we'll need to `jackson` csv dataformat:
	To use the CSV mapper, we'll need to add Jackson's CSV dataformat dependency:

	Let's start with the REST endpoint. It's pretty straight forward. We need to methods one for movie searching and one for searching similar movies.
	Let's start with the REST endpoint. It's pretty straightforward. We need two methods, one for searching movies and one for searching similar movies.


		The key elements of that page are:

		* movie-box: a text filed for entering the movie title

	* movie-box: a text filed for entering the movie title
	* movie-box: a text field for entering the movie title

	It's important to remember that the `Movie` entity is using `jackson` to map the CSV columns to the entity fields.
	It's important to remember that the `Movie` entity is using Jackson to map the CSV columns to the entity fields.

	I won't go into much detail about the hmtl code as it's outside the scope of this post.
	I won't go into much detail about the HTML code as it's outside the scope of this post.

Conversation

iocanel commented Mar 18, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

github-actions Bot commented Mar 18, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

geoand left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

iocanel commented Mar 19, 2025

Uh oh!

gsmet left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

iocanel commented Mar 26, 2025

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

iocanel commented Mar 18, 2025 •

edited

Loading

github-actions Bot commented Mar 18, 2025 •

edited

Loading