Ep41: Spring AI M7, NODES, RAG + Neo4j, Quarkus, and Intelligent Apps

In this episode, we unpack a busy week of updates, learning, and cool tech! From Spring AI’s milestone 7 release with simplified Pinecone configuration to some tricky wifi, I walk through recent changes and adventures. Plus, NODES 2025 is officially announced, and there’s hints of our upcoming GraphAcademy Java course. I also talk about the first part of my new blog series on Retrieval Augmented Generation and highlight a fantastic article on Neo4j, Quarkus, and intelligent applications.

Episode Transcript

(00:05):
You are listening to the BreaktimeTech Talks podcast, a bite-sized tech
podcast for busy developers where we'llbriefly cover technical topics, new
snippets, and more in short time blocks.
I'm your host, Jennifer Wright, anavid developer and problem solver
with special interest in datalearning and all things technology.

(00:26):
I realized there are a lot ofacronyms in this episode's title.
I feel like maybe I should win somekind of buzzword bingo or something.
Anyway, I just got back from a coupleof events last week, and this week, I
poured time into several other projects,which I will be sharing with you now,
as well as in some upcoming episodes.
I'm really excited about all thelearning and content that I have
scheduled, so let's dive in.

(00:47):
The first thing I noticed lastweek is that Spring AI updated to
a new milestone release, so they'renow on a milestone seven release.
And there were a couple ofshifts here, but really the
changes weren't all that drastic.
The first thing is that the dependencynames of some of the packages shifted
just a little bit, especially formodel and vector store starters.

(01:10):
So just to be aware, if you update tothe M7 Spring AI release, you'll need
to update some of the package names andgroup IDs and such for your dependencies.
This I noticed, especially true forthe Vector Store for Pinecone and
Neo4j, as well as the model that I wasusing, which was at the time OpenAI.
And I noticed that the starter termin the package name just moved earlier

(01:33):
into the package naming, if you will.
This aligns actually much betterwith some of the other Spring
framework dependency name structures.
A lot of their projects are similarlynamed with the dependencies.
And so now this really alignsmuch better with some of the
other Spring dependencies.
But just so you know, thatwill be a breaking change.

(01:54):
You'll have to go in and updatedependency package names and such.
The next thing that I was reallyexcited has been updated is
the Pinecone configuration.
It is way easier now.
I had mentioned a couple episodes backthat it was really a hassle to figure
out because Pinecone shifted the waythat they connect to database instances
on their end, and that made configuringit in a Spring AI application a little

(02:19):
bit tricky to understand where theproject ID and the environment and all
those different pieces were in the, nowwhat's called the host name on Pinecone.
But now Spring AI aligns much betterwith what Pinecone is looking for.
And now you only have to provide the APIkey and the index that you're going to be
searching, and it handles the rest of it.

(02:40):
So much, much easier.
That is gonna be a breaking changethough, if you had the project ID
and the environment configured andwritten in like a bean, for instance.
Those methods for projectID and environment are no
longer there, no longer valid.
That will pop up and say, Hey, youneed to update this, but that really
is the only thing that's impacted.
And again, I feel like this is away better, much more efficient and

(03:04):
streamlined approach and matchesmuch better with what's in Pinecone.
So I was super excited to see this update.
The next thing that I came across,which isn't so much of a, a Spring AI
thing as just something that I foundout that happened to coincide with the
Spring AI milestone seven release isthat I was presenting at some stuff
last week and updated my applicationto the M7 release and realized I was

(03:27):
having trouble connecting to Pinecone.
I thought this was a little odd.
They have a Pinecone IO on the end.
So I thought, well, maybe it'ssomething blocking like .io URLs or
something, but I could access Neo4jjust fine, which Neo4j is also a .io.
So I was like, well, they'renot completely blocking that.
And come to find out, I was working ata customer site for the afternoon and

(03:50):
then presenting there later on, and theyblock gen AI related or AI related URLs.
So a lot of the like access pointsand APIs and things like that are just
completely blocked on their wifi network.
So it was the wifi networkcausing the issue there.
But they're a customer of Neo4j, soI'm sure that's why I was able to
access Neo4j's without any issue.

(04:12):
And then just they blocked Pinecone'sVector store for some odd, strange reason.
I ended up connecting to myhotspot on my cell phone and was
able to have no problem, besidesjust being a little bit slower.
But that worked great.
Now I had the flip thinghappen a day or two later.
I was updating my applications andrunning them and testing them before
I pushed some updates to GitHub.

(04:33):
And I was at the airport doing this,and I had the opposite problem occur
where I could access pinecone.iojust fine, but I could not access
Neo4j's console, the .io again.
So I thought, well, this is really odd.
Maybe they happen to be a Pineconecustomer or something, and not Neo4j.
But an airport wifi, of all things,was blocking Neo4j's cloud database
service and not Pinecone's.

(04:55):
Again, I just updated to using myhotspot for a few minutes to test
it out, make sure everything worked,and then push that up to Github.
No problems after that.
But just something you might wanna beaware of that I initially had a mild
panic attack when I did this the firsttime that it was something in Spring AI
had broken the Pinecone configurationin connection to the database, and I

(05:16):
thought I was gonna have to do somethingelse and rework my demo and so on.
But ended up, it was not Spring at all.
It was actually the wifiblocking those certain sites.
So just to be aware.
If you have problems with that, youmay need to find another connection
network or run something locally.
But it was kind of funny theway this all kind of coincided.
I thought it was a little bit unusual tocome across one website being blocked at

(05:40):
a customer site, for instance, and not theother, and then vice versa at an airport.
Strange things.
But as far as the Spring AI release, theupgrade was very, very seamless, outside
of the dependency and configurationbreaking changes, if you will.
It took me less than 30 minutesto upgrade the entire application.
I was super excited about this.
That is the way to do an upgrade.
If you're gonna introduce somebreaking changes, make them very minor

(06:02):
and relatively small, and then muchfewer things that you have to change.
So the only things I had to update werethe dependencies and the configurations.
Everything else pretty much stayedthe same, so that was awesome.
I really loved that experience.
The next thing I want to chatabout is that NODES 2025 was
announced last week as well.
There's a blog post out about it.
NODES is Neo4j's onlinedeveloper conference.

(06:24):
It is all free, all online, and all dayabout 24 hours, maybe a little over,
of content, technical talks, and code.
There's lots of tracks again this yearfor a variety of different topics,
and of course you'll hear aboutthe latest things going on in the
graph and general technology world.
The CFP is also open, so we arenow accepting abstracts to submit

(06:47):
for conference presentations.
That will be open until June 15th.
If you're interested in speaking aboutsomething graph related, again, we open
this up to the general graph community.
It'll be customers and communityand partners and all sorts
of really interesting things.
People doing cool things withtools and technologies and concepts
and data sets and visualizationsand all sorts of neat stuff.

(07:09):
Whether I present or just attend,I will definitely be there.
I would love to see you there.
This is one of my favoriteevents every year.
There's just so much opportunityfor inspiration and motivation to
really dig into some really cooltopics and things that I might not
have thought of before from all theseamazing people around the world.
And again, speakers are fantastic.

(07:30):
Their content's fantastic.
And no matter whether I speakor just attend, I will be
there and soaking it all in.
So come join me.
The next thing that I worked on isGraphacademy Java Application course.
There's an upcoming one that mycolleague and I are working on.
We have an existing one that'sa full app development course.

(07:50):
If you're not familiar, I guessI'll back up a little bit.
Graphacademy, Neo4j Graphacademyhosts Neo4j related courses
that are free and self-paced ona variety of different topics.
And we have a current Java applicationcourse that builds a full app with
Java and Neo4j, but we wanted to takea step back from this and tease apart

(08:10):
some of the basic foundational conceptsfor working with the Java driver.
So we're looking at how to interactwith Neo4j using just the Java driver
and the connection to the database,how to run queries, how to handle the
results, and really just sticking withkind of the foundational concepts there.
And then use that as a springboardto dive into the fuller

(08:31):
application course at a later date.
I'll talk more aboutthis in a future post.
I don't wanna give too many details away.
There's still a lot of things we'reworking out and, and dealing with.
But this is upcoming.
I hope that will be super excitingfor those of you in the Java
space who just want to get anintroductory start with Neo4j.
Then the last piece that I workedon this week was a new blog post

(08:52):
on the introduction to retrievalaugmented generation, part one.
There will be at least a parttwo, possibly more, but at least
part one and two, and I got thepart one published this week.
I'd had some feedback on some of mypresentations about how helpful it was
to understand some of the foundationalconcepts and terms being tossed around
in the generative AI and AI spaces.

(09:13):
And so I put some of these generalconcepts that I've been presenting
at some of my talks for a whilenow, I put these into a blog format.
This part one covers a little biton just what is Retrieval Augmented
Generation, talks a little bit aboutlarge language models and how they
work, vectors and vector similaritysearch, and, and general topics.
And then there will be a part two blogpost coming up that I hope to cover

(09:36):
vector rag, graphRAG, and agents.
Possibly some other things too, butat least we'll start with those.
I was super excited to get this out, setthe groundwork for some of the things that
are being tossed around and maybe somechallenging concepts that not many of us
understand really and fully very well.
And I really feel like understandingthese foundational concepts helps
build a better structure on top ofthat for putting together really

(10:01):
well architected and design systemsbecause you're understanding the basic
building blocks that go into that.
So this was part one.
It's out on my website now, and again,I hope to have a part two coming soon.
The last piece of content I wantto cover for this week is I came
across an article from a colleagueof mine, Marcus Eisele, who wrote

(10:22):
The Power of Relationships, Neo4jQuarkus, and Intelligent Applications.
I really loved this introductionto so many of these concepts.
It talks a little bit about graphs,which use cases they shine over
things like relational databases.
Then how knowledge graphsoverall can contribute to the
AI and generative AI spaces.
Talks a little bit about what is Quarkusand the benefits for applications,

(10:44):
plus the fact that Quarkus has a greatLangchain4j integration available as well.
And then actually walks throughbuilding a sample recommendations
application using Quarkus and Neo4j topresent, the code here and give you a
springboard for tacking on Langchain4jor Kubernetes or whatever you might be
wanting to plug into past that point.

(11:04):
But I've been wanting to play withQuarkus now for a while, and this is
just another encouragement to bumpit up on the list just a little bit.
I really enjoy this article,and I hope you do as well.
This week has been a busy weekcatching up on things in my backlog.
Spring AI upgraded to a new milestonerelease which was a cinch to migrate.
NODES 2025 was announced last weekas well, along with the CFP, I

(11:25):
published a part one blog post on anintroduction to Retrieval Augmented
Generation, part two forthcoming.
And I worked on an upcomingGraphacademy course for Java and Neo4j.
I'm hoping the next few weekscontinue this trend and a plan to
bring you all along for the ride.
As always, thanks forlistening and happy coding.