HomeVideos

Open WebUI Tools: A Comprehensive Guide for Beginners and Advanced Users

Now Playing

Open WebUI Tools: A Comprehensive Guide for Beginners and Advanced Users

Transcript

325 segments

0:00

hello everyone this is Professor pattern

0:01

and in the next couple of videos we'll

0:03

be covering open web UI functions

0:05

pipelines tools and all of those things

0:07

in between we'll start off with this

0:09

video by covering tools now if you are

0:12

an advanced open web UI user the only

0:15

thing that you need to know is that a

0:16

tool is a python script that your large

0:20

language model can execute that's it you

0:23

can skip to the second part which is the

0:25

implementation step if you're a complete

0:27

beginner and had no idea what I just

0:29

said and stick around for the next part

0:32

so let's imagine a scenario where you

0:35

are a boss and you have an

0:38

assistant now if you ask your assistant

0:40

a question something like hey can you

0:42

grab me the transcript from this video

0:45

now your assistant is going to click on

0:47

this video and they'll see that okay

0:49

well this is actually a 10-minute video

0:52

by Professor patterns now I love

0:54

listening to Professor patterns but do I

0:56

have to actually manually type out every

0:58

single word from this video well that's

1:01

one way in which you could do this

1:03

another way is by simply creating a

1:06

python script the goal of the script is

1:08

to first go to the video then it's to

1:11

fetch the text from the video and then

1:13

print out the

1:15

transcript so the next time you ask this

1:17

question like hey can you grab the

1:18

transcript from this video your

1:20

assistant is going to be able to Simply

1:22

execute the script which will fetch the

1:25

video and then print out all of those

1:27

things from the transcript now this

1:29

script that your assist assistant has

1:30

created is what we in open web UI call a

1:33

tool so the goal of this tool which is

1:36

get transcript is to go fetch the video

1:39

grab the transcript and print out the

1:40

transcript similar to this tool we also

1:43

have a bunch of different tools like get

1:44

transcript get stock prices send out

1:47

automatic emails summarize meetings we

1:49

have a tool to convert image to text um

1:52

to create Jura tickets from meeting

1:53

notes all of these are examples of tools

1:56

and the goal here is that any single

1:59

time that you ask a question like hey

2:00

grab me the transcript from this video

2:02

your assistant knows that okay well

2:04

looks like in this case the tool that I

2:05

need to execute is get transcript or if

2:08

your question is something like hey send

2:10

an email to my wife wishing her for

2:12

Valentine's Day your assistant knows

2:14

that the tool to execute is send

2:15

automatic

2:16

emails now in this scenario your

2:19

assistant isn't obviously your assistant

2:22

it's a large language model so for

2:24

example this could be something like

2:25

Chach BT or deep seek or Gro or mistl or

2:29

the Lama 3 Model now what you're doing

2:32

is you're providing all of these llms

2:35

access to these tools so when you ask a

2:38

question like hey send an email to my

2:40

wife wishing her for Valentine's Day the

2:43

large language model is going to be able

2:45

to execute the tool for sending

2:47

automatic emails now here is what it's

2:50

going to look like on the open web UI

2:52

platform the first time you'll say

2:54

something like hey can you transcribe

2:55

this video and Chad gbt 40 is going to

2:58

come back to you and say look look sorry

3:00

I can't really transcribe that video but

3:01

you can try turning on um captions

3:05

that's not really helpful so what we'll

3:07

do is we'll provided access to the tool

3:10

now obviously I'm I didn't write this

3:11

tool by myself we have an entire

3:13

Community which I'm going to talk about

3:15

here in a second but this is a tool or a

3:17

script that I provided and the only

3:20

thing that I'll need to do is to

3:21

activate it which is basically to say

3:23

enable this tool so next time if I was

3:26

to ask a question like hey transcribe

3:27

this video it knows that this tool is

3:30

available so it executes this tool or

3:33

this script and then based on that it's

3:34

now able to actually transcribe this

3:36

video right hello everyone this is

3:38

Professor patterns and in this video you

3:40

get the idea so this is how a tool Works

3:44

a tool is a python script that your

3:46

large language model can execute so what

3:48

does this actually look like on the open

3:50

web UI interface well from your homepage

3:53

you'll select workspace and then select

3:55

tools now here you can see that I have

3:58

this YouTube transcript provider tool

4:00

that I've downloaded from the community

4:02

page and the goal of this tool is to go

4:05

go to the YouTube video grab the

4:07

transcript and then simply print it out

4:08

for me now to make sure that the tool is

4:11

running I need to click on this plus

4:12

icon and then enable it once I do that I

4:15

can ask a question something like grab

4:16

me the transcript from this video and

4:18

you can see here that when the Tool is

4:20

running it should be blinking here so

4:22

when I ask it this question it's going

4:24

to be able to go to this video and then

4:26

based on that it will start printing out

4:28

the transcript

4:30

now if you go over to openweb ui.com I'm

4:33

going to share a link to this in the

4:34

description you can go to the new tools

4:37

section so if I select SE all you can

4:40

see that I have a bunch of different

4:41

tools that are available here there are

4:43

some for email there's one for Reddit

4:46

there's a calculator tool u a transcribe

4:49

tool a wolf from alpha tool a 11 Labs

4:52

tool um let's try one let's say that for

4:54

example I wanted to try the enhanced web

4:56

scrape tool right this looks interesting

4:59

so to actually get this tool it's pretty

5:01

easy all you will do is simply select

5:03

get and then here is going to say import

5:06

to open web UI now you want to make sure

5:08

that open webui is running or it's not

5:10

going to work once I select this it's

5:13

going to automatically show up here on

5:15

open web

5:16

UI now something that you want to be

5:18

careful um about is you want to make

5:21

sure that the tools that you're

5:22

downloading are going to have at least a

5:23

couple of people who have used it in the

5:25

past because um these are going to be

5:27

executable files so some that you might

5:30

want to do is this one's fine cuz it's

5:32

got 7,000 something downloads but just

5:35

as an extra step of precaution I would

5:37

just copy and paste the tool put it into

5:39

chat DBT and just be like hey is this

5:41

safe or something or is this going to

5:42

send my data out somewhere just to be

5:44

sure you never know so here once you do

5:48

this you will be able to see your tool

5:50

available here and I'm going to Simply

5:52

select save and again you'll get this

5:55

warning right so tools have a function

5:56

calling systems that allow code

5:58

execution so I'm going to select confirm

6:02

and let's say that for example a one of

6:05

the tools that you downloaded has an API

6:07

key that is necessary well to do that

6:10

you'll simply select the valves section

6:13

and over here we can see that it is

6:15

asking for an API key but it is optional

6:18

so it allows for a higher rate limit

6:20

when you're scraping but you don't need

6:22

to actually have this so that's great

6:24

there's also another sort of thing for

6:26

citation so the default citation is

6:29

enabled

6:30

let's just say that we disabled citation

6:32

right we can do that over here U do we

6:34

want to disable caching I don't know

6:37

let's just leave it as the default value

6:40

and then I'm going to select save so for

6:42

example if you have other tools that do

6:45

require an API this is where you would

6:46

put that API in so let's go here to the

6:49

new chat and I need to make sure that

6:51

the tool for enhanced web search is

6:54

active and then we'll say something like

6:56

transcribe

6:59

this

7:00

page and when I send this message you'll

7:04

see here that it's going to scrape that

7:05

website and then based on that it's

7:08

going to tell us what this page is about

7:11

and that's pretty much it now all of

7:13

this is fine right like we know that we

7:14

can access tools from the community but

7:16

what if you had a very specific purpose

7:19

right and you wanted to create your own

7:20

tool how do you do that well you don't

7:23

need to be a developer all you'll do is

7:26

just go to the open web UI documentation

7:28

page cop copy this entire whatever

7:31

documentation is available on tools go

7:33

here and paste it next I'm going to copy

7:37

an example tool that works like the

7:39

enhanced web script right we tried that

7:41

we know that this works so I'm going to

7:42

go in here I'm going to say this is an

7:46

example of a tool that

7:48

works and I'm going to paste it and then

7:51

I'm also going to say this is another

7:53

example of a tool that works and then

7:56

I'm going to give it the YouTube

7:57

transcript provider so let's go in here

8:01

and then paste it now I can ask it to do

8:04

whatever I want right so let's just say

8:05

that I want it to create a tool for me

8:08

so I'm going to start typing

8:10

hey can you help create a new tool for

8:14

me that

8:16

fetches the top five posts from the Wall

8:20

Street bets

8:23

subreddit so when I execute this it's

8:26

not only does it have information on the

8:28

tools documentation also has some

8:30

example code from other people's tool

8:32

that does work so here it actually wrote

8:34

this entire script for me and I'd love

8:38

to actually try it out to see if this

8:39

works so I'm going to go here and then

8:42

I'm going to copy this entire script and

8:45

then let's go to settings or workspace

8:49

tools and let's create this new tool so

8:52

let's paste this entire code that we

8:54

were provided and then let's give this a

8:57

tool a name so w Street

9:01

bets and the tool description grabs top

9:05

five posts from

9:07

WSB so first I need to see if this

9:10

actually works I'm going to select

9:12

save and

9:13

confirm so it says code formatted

9:16

successfully let's see if this actually

9:18

works so do I have any vales that I need

9:20

to put in so client ID client secret and

9:22

then Reddit user agent let's leave these

9:26

as default for now because I'm not sure

9:28

um what those things are and let's try

9:30

to activate or run this tool of course I

9:34

put a typo in that U what are the top

9:37

five posts

9:39

today so what we need to do is see if

9:43

this works or not so it says Reddit API

9:45

credentials not configured so it means

9:47

that we need to provide it some

9:49

credentials so if I wanted to I could go

9:52

to api. reddit.com and then go in here

9:55

and then register to get the API but it

9:57

seems like this is a whole thing so

9:59

instead of doing this let me just get a

10:01

different or let me create a different

10:02

tool let's just say that I wanted to

10:04

create a different tool that doesn't use

10:07

an API call anywhere which now it makes

10:09

sense what those valves were which I

10:11

just completely had ignored at the time

10:13

so here it's saying here's a simple tool

10:15

that performs text analysis without

10:17

requiring any API calls so it analyzes

10:20

the text and then provides statistics

10:22

like word count character count most

10:23

common words and some more metrics so

10:27

yeah not the most glamorous tool I will

10:29

say but I mean it's a good use case

10:31

right so for example if you had

10:32

something that you wanted to try out you

10:34

can always go in here and then have it

10:36

kind of write that out for you looks

10:38

like it is actually removing the stop

10:40

words and then building some sort of a

10:41

response so I'm really excited to see if

10:43

this actually works so let's wait for it

10:45

to get

10:48

done and then I'm going to Simply select

10:51

copy now here let's go to workspace

10:54

tools and let's create this tool word

10:57

counter and then I'm going to paste it

10:59

and this

11:01

tool counts words let's hit save and

11:08

confirm only alpha numeric characters

11:10

allowed in the title let's hit save and

11:12

then confirm it says tool um created

11:16

successfully uh there are no valves so

11:18

that is awesome so let's go back here to

11:21

our new chat enable this tool which is

11:24

word counter and say or just type a

11:27

sentence like how many or

11:30

statistics on this

11:34

sentence this is a

11:36

sentence I don't know if that's going to

11:38

work or not but let's try it out this

11:40

analysis complete and then here are the

11:42

detailed statistics so it gave us the

11:44

basic statistics um reading analysis and

11:46

then word frequency so we know that this

11:49

actually works because this is the

11:51

format that the tool is actually working

11:53

in so if you go to workspace select

11:55

tools and then the word counter tool we

11:57

can see here that it's in that same sort

12:00

of format right text analysis results so

12:03

we know that this sort of pipeline Works

12:05

um if you have some use cases please

12:08

feel free to post those in the comment

12:09

section um I'd love to try some out you

12:12

know this is really the workflow you

12:14

don't need to be a developer not to

12:15

create these tools create these

12:17

functions so I encourage you to go out

12:20

maybe try creating some of these tools

12:21

but also be careful because remember

12:23

that these are executable python files

12:26

now that's pretty much it for this video

12:28

thank you all for tuning in hopefully it

12:29

was a little bit helpful and I'll see

12:31

you in the next one goodbye

Interactive Summary

This video introduces tools within Open Web UI, defining them as Python scripts that large language models (LLMs) can execute to perform specific tasks. Using an analogy of an assistant, the speaker explains how LLMs can leverage these tools to handle requests like transcribing videos or sending emails, which they couldn't do natively. The video demonstrates how to find and enable existing tools from the Open Web UI community, while also highlighting the importance of security given that tools are executable files. Furthermore, it illustrates how users can create custom tools with the assistance of an LLM, walking through an example of building a 'Word Counter' tool.

Suggested questions

5 ready-made prompts