Welcome welcome welcome to the deep dive the show that's well your shortcut to
Download transcript (.srt)
0:00
0:04
0:06
0:09
0:14
0:14
0:18
0:20
0:24
0:27
0:32
0:33
0:39
0:40
0:43
0:45
0:48
0:51
0:54
0:57
1:00
1:03
1:05
1:10
1:11
1:13
1:16
1:20
1:22
1:25
1:26
1:30
1:32
1:36
1:37
1:40
1:44
1:45
1:46
1:51
1:51
1:53
1:58
2:00
2:05
2:05
2:09
2:10
2:15
2:17
2:22
2:25
2:29
2:33
2:34
2:38
2:42
2:46
2:48
2:53
2:54
2:58
3:00
3:04
3:06
3:11
3:11
3:14
3:18
3:22
3:27
3:29
3:31
3:36
3:40
3:43
3:46
3:48
3:54
3:56
4:01
4:01
4:06
4:06
4:10
4:11
4:14
4:18
4:19
4:23
4:24
4:29
4:33
4:37
4:38
4:42
4:46
4:49
4:52
4:53
4:57
5:01
5:06
5:09
5:14
5:16
5:20
5:23
5:27
5:29
5:34
5:36
5:40
5:43
5:47
5:48
5:53
5:55
5:59
6:01
6:05
6:06
6:09
6:12
6:16
6:20
6:25
6:25
6:30
6:30
6:34
6:37
6:39
6:45
6:47
6:50
6:54
6:55
6:58
7:02
7:05
7:10
7:11
7:17
7:19
7:23
7:26
7:31
7:33
7:35
7:39
7:40
7:44
7:49
7:51
7:55
7:58
8:02
8:05
8:09
8:11
8:15
8:17
8:22
8:23
8:27
8:30
8:36
8:42
8:43
8:49
8:49
8:53
8:57
8:59
9:03
9:04
9:08
9:12
9:14
9:18
9:22
9:27
9:30
9:34
9:38
9:39
9:42
9:45
9:48
9:53
9:53
9:58
10:02
10:06
10:07
10:11
10:12
10:19
10:20
10:24
10:28
10:32
10:35
10:38
10:42
10:44
10:49
10:49
10:53
10:55
10:58
11:01
11:06
11:07
11:11
11:16
11:21
11:23
11:26
11:31
11:33
11:38
11:39
11:41
11:46
11:47
11:51
11:53
11:57
12:00
12:04
12:06
12:10
12:15
12:16
12:20
12:22
12:27
12:27
12:33
12:35
12:37
12:41
12:42
12:46
12:50
12:55
13:00
13:01
13:06
13:07
13:10
13:11
13:16
13:20
13:23
13:27
13:30
13:35
13:36
13:40
13:44
13:45
13:49
13:52
13:55
13:58
14:00
14:05
14:09
14:11
14:17
14:19
14:21
14:24
14:30
14:31
14:34
14:38
14:41
14:44
14:47
14:47
being genuinely well informed
We cut through the noise to get you straight to the insights that matter most
Before we jump into today's topic though a huge shout out to our supporter safe
server
They're the experts you want for software hosting and they're a real partner in
digital transformation
Basically helping you level up your tech and navigate that whole complex digital
world if you're looking to empower your projects
Definitely check them out at WWW dot safe server dot DE again. That's WWW dot safe
server dot DE
Okay, so let's be real in today's world. We're just swimming maybe drowning in
information, right?
Whether you're trying to find out one perfect pair of shoes online or maybe
pinpoint a specific song in a giant music library
Or digging for that crucial file buried somewhere. We all want the same thing
instant access effortless find exactly what you need right away
No messing around and today we're taking a deep dive into type sense
It's a really powerful open source tool designed specifically to make those search
experiences less frustrating and honestly delightful
That's a great way to put it. Yeah type sense is fundamentally an open source
lightning fast search engine
But it's like built for mere mortals. That's how they put it. The main goal is top
performance
Absolutely, but just as important as ease of use
It's about making sure you don't need you know
A PhD or a huge dedicated team to get fantastic search working or even just to
understand what it can do it kind of
Demystify search deck. I like that for mere mortals
Okay
So our mission for this deep dive unpack what type sense is see how it stacks up
against the big often way more complex
Players out there and figure out why it's getting so much attention for making
search powerful
But also accessible and well efficient for pretty much everyone
We've gathered insights straight from the type sense github repository and their
official website
So we're getting it from the source
Think of this as your fast track to understanding what might be the next big thing
in search
Okay, let's uh, let's make this real
think about those everyday search frustrations, you know the feeling you type
something hit enter and you wait or
Worse you make one tiny typo just one letter off and boom no results found so
annoying
Or maybe you get too much back just a flood of stuff that isn't quite right and you
feel totally lost
We want fast easy precise, but search often feels more like I don't know digging
through a messy attic. Oh, absolutely
It's a common pain point and that's exactly where type sense comes in is this
Elegant solution at its core. It's a fast
typo-tolerant in memory fuzzy search engine
Okay, let's break that down a bit because those terms are key to the no PhD
required idea fuzzy
Just means it's forgiving you misspell something only remember part of a name
Type sense is smart enough to often guess what you meant and still find it
So it understands intent not just the exact letters I type exactly it gets what you
meant not just what you typed
That's huge for user experience and then in memory. This is really interesting
because it's core to its speed
It means the search index the data it's searching is loaded right into the computer's
RAM. It's working memory
Okay, so like keeping the important stuff right on your desk instead of filed away
somewhere precisely imagine looking for a book
Is it faster to grab it off the shelf next to you or go search boxes in the garage
in?
Memory is having it right there on the shelf
Super fast lookups. It's all designed to cut complexity and give you that genuinely
delightful fast search without needing deep technical expertise
That attic versus bookshelf analogy really clicks. Okay, so here's where it gets
like really interesting for a lot of folks
Type sense often gets mentioned as an alternative
Maybe even a better one sometimes compared to big names like elastic search or Algolia
For someone starting out or even a developer looking for a simpler way. What really
sets type sense apart? Why choose it?
Yeah, the differences are pretty stark when you look at actually using them take
elastic search. It's powerful
No doubt hue scale complex queries. It can do a lot, but it's often a really big
piece of software
Setting it up tuning it scaling it that could be complex often needs specialized
teams
Significant ongoing effort so a big investment in just managing the search part,
right?
You have to ask do you want your team focused on managing search infrastructure or
building your actual product?
Type sense flips that it's designed to be lightweight
But still really powerful and scalable and it comes as a single binary one file
seriously
Yeah, pretty much you download one file run it and you're basically ready to go. No
complex installs
No fighting dependencies. It's plug-and-play for search. The focus is really on
Developer happiness a clean well documented API
That's a toolkit for programmers to interact with it and the API is designed to be
straightforward
So it you know, it just works well out of the box less tinkering needed Wow
Okay, one file install sounds amazing for anyone who's struggled with setup before
so what about Algolia then?
That's often a hosted service people know how does type sense compare there,
especially cost and flexibility wise?
That's a really important comparison. Algolia is great performs
Well good UX, but it's proprietary hosted search as a service
The potential issue there is cost as your site or app grows you hit search limits
Indexing limits and suddenly your bill can jump it can get expensive fast
Ah the scaling cost trap exactly type sense being open source gives you the choice
run it on your own servers
Totally free beside your server costs. That's potentially huge savings as you scale
or if you want managed convenience
There's type sense cloud, but its pricing is different. It's based on fixed hourly
costs for the server resources
It's predictable not based on how many records you have or how many searches people
do which can be hard to predict
And budget for with other services. Okay, predictable pricing is a big plus any key
technical differences besides the hosting model
Yes a key one for efficiency type sense can give you sorted results from a single
index
So if you want users to sort products by price or by date or by rating type sense
handles that efficiently with one main data structure
Algolia often needs you to create separate duplicate indices for each sort order
sort by price
That's one index sort by date another index which means more memory usage more
complexity
Right more memory more management overhead now to be fair
Algolia does have some very mature features like built-in personalization and
server-side analytics that types sense is still building out
But for core search functionality and especially for analytics
You can easily use standard web analytics tools with type sense on the client side
for now
For many the simplicity cost and open source benefits outweigh those current
differences
Yeah, it really sounds like ease of use is deeply embedded and the source is
mentioned getting from zero to instant search and like
30 seconds, that's incredible
Plus multiple easy install options Docker native binaries the managed cloud. It
seems built to remove friction
So, okay, it's easy. It's fast. But what about the features?
What specific things does type sense offer that make the search experience itself
better smarter more?
Delightful for the person actually doing the searching right beyond the basics
There's some really user-friendly features first off typo tolerance. This is
massive. It handles typos gracefully out of the box
No special setup needed. Remember the example from their docs search stork and it
finds stark industries, huh?
Okay, that's smart. We've all made those typos. Exactly. It just understands then
the blazing fast speed
We keep saying it but it's engineered for it builds in C plus plus N aiming for
under 50 meters response times
That speed feels good to users. You also get great control with tunable ranking
sorting fasting and filtering
Sounds tucky, but it's about control
Tunable ranking that's decide what's most important in results. Maybe newer items
first or higher rated ones
Sorty is obvious. Let users sort by price date etc and
Faceting and filtering is that like the sidebar options on shopping sites precisely
like filter by brand filter by size filter by color
Yeah, it lets users slice and dice the results easily turns a huge list into
exactly what they want makes finding things intuitive
Okay, that makes perfect sense very user centric definitely and synonyms are great
to handle things like pants versus trousers sneakers versus trainers
Users find stuff even if they use different words then we get into more advanced
territory, but still kept accessible
Vector search and semantic hybrid search. This is cool. It's about understanding
meaning not just key words
so going beyond exact word matching right think searching for comfortable shoes for
walking and
It finds results even if they don't use those exact words because it understands
the concept. It uses AI models to find
Eerily accurate results matching the intent Wow. Okay. That's next level
It is and even more so is conversational search with built-in rag rag stands for
retrieval augmented generation
Think chat GPT but running securely over your own data your company Docs your
knowledge base
Whatever you index wait, so I can ask it questions in normal language exactly
You can ask something like what's our policy on international shipping and instead
of just give you a list of documents where that might be
Mentioned it retrieved the relevant info from your data and then generates a clear
natural language answer
Like our policy states that based on the documents it read that's incredible
It's not just finding it's answering that changes everything for internal knowledge
search or customer support bots
It's a huge leap and there's also geo search perfect for finding things nearby
Like stores or services within a certain distance and finally federated search
This lets you search across different types of information at once like one search
bar could hit your products your blog posts and your help
Docs simultaneously bringing it all together in one place very neat
Okay, those features are compelling but you know proof is in the pudding numbers
talk
What kind of actual performance are we talking about and who's really using this?
Can it handle serious scale?
It definitely handles scale. The benchmarks are pretty telling
Let's take one example a data set of 2.2 million recipes names ingredients the
whole thing
indexing that took only about 900 megabytes of RAM and type sense and on a fairly
standard server for virtual CPUs it handled
104 concurrent searches per second a hundred people searching at the exact same
time
Yep, and the average processing time just 11 milliseconds near instant or a bigger
one
28 million books titles authors categories that used about 14 gigs of RAM same
server
It managed 46 concurrent searches per second average time 28 milliseconds
So even with tens of millions of items it stays incredibly responsive
All right that directly impacts user happiness and can mean lower server costs for
you. Those are solid numbers really backs up the speed claims
What about adoption is it just niche or lots of people using it?
Oh, it's way beyond me now type sense cloud their managed service is serving over
10 billion searches every month
That's massive real-world usage 10 billion. Okay. Yeah, that's serious in the docker
images
How developers often run it downloaded over 12 million times? That shows a huge
active community
It's being used by all sorts of companies. Yeah from you know, scrappy startups
building the next big thing to well-known household names
You can even play with live demos on their site searching 32 million songs or those
28 million books instantly
It really shows it off. Okay impressive scale and adoption
Now that open source aspect is a big draw but licenses
Sometimes they're confusing you mentioned the server is GPL 3.0, but the client
libraries are Apache. Can you break that down?
What does GPL mean for someone using the type sense server?
Yeah, it's a good question and they've put thought into this so the core type sense
server is GPL 3.0
Z the client libraries the bits you use in your app code to talk to the server are
Apache licensed
Let's simplify the GPL part for the server
Think of it like this GPL is designed to ensure that improvements to the core open
source project
Benefit the whole community in the long run if you take the type sense server code
modify it and then distribute that modified version like
Selling it as a service or including it in software you distribute you generally
need to make your modifications available under the same GPL license
Share your improvements back. Okay, so it encourages giving back to the project if
you build upon it and share it like Linux
It's exactly like Linux. It's a common model for foundational open source projects
But here's a key point if you modify the types and server and only use it
internally within your company
Without distributing those changes outside you generally don't have to open source
your internal modifications
Ah, okay. So internal use gets more flexibility that makes sense for businesses,
right?
It protects the open source nature when code is shared but allows private use and
the Apache license for the client libraries
That's much more permissive
It basically means you can easily use those libraries in your own applications even
proprietary closed source ones
Without needing to open source your application code just because you're talking to
typesense so easy integration without strings attached for your own app code
Precisely. It's a smart balance protect the core projects openness with GPL
Enable easy adoption in any kind of app with Apache for the clients that really
sounds like a well-thought-out approach
And if people run into issues or need help, what's the support situation like it's
pretty robust
There's an active public slack community great for general questions getting quick
feedback from other users
Even the core team sometimes hangs out there for bugs or feature requests
Github issues are the way to go standard open source practice and then for
businesses needing guaranteed response times or private support
There are paid support options available with SLA's service level agreements. So
help is there at different levels great
So community help official issue tracking and paid options for businesses covers
all bases
Alright, that brings us towards the end of our deep dive on type sense
I think the big takeaway is this isn't just another search tool. It's genuinely
fast super forgiving with typos
It's open source and really seems built around making developers lives easier while
delivering. Yeah, truly delightful search experiences
No PhD required really sums it up
It makes complex search stuff feel accessible and efficient whether you're just
starting out or scaling up big time a real game changer
It seems absolutely and it makes you think doesn't it if a tool like typesense can
revolutionize something as fundamental as search by focusing on speed
simplicity and well scar features
What other parts of our digital lives our everyday online interactions could be
made significantly better more intuitive
Maybe even delightful it poses a question
Where else can we apply this kind of focus on?
Core user needs and really clever thoughtful engineering to smooth out the friction
in our digital world
Something to ponder as we all keep searching for things online
That's a fantastic thought to end on and once again huge thank you to safe server
for supporting the deep dive your trusted partner for software
Hosting and digital transformation find out how they can help you at
www.safe-server.de one more time
deep dive
deep dive