NullClaw + Ollama Local Setup with Telegram

*https://www.youtube.com/watch?v=FidBdtZHZ4U
**

пересказ видео

00:00:01 Введение в Nullclaw

  • Nullclaw — полностью автономный ИИ-помощник, созданный на языке системного программирования Zig.
  • В видео показано, как установить Nullclaw локально и интегрировать его с локальными моделями Olama.

00:00:41 Характеристики Nullclaw

  • Nullclaw написан на Zig, что обеспечивает высокую производительность и небольшой размер бинарного файла.
  • Размер бинарного файла — 678 КБ, использование оперативной памяти — около 1 МБ, загрузка менее чем за 2 мс.
  • Входит в семейство OpenClaw.

00:01:35 Установка Zig и сборка проекта

  • Установка Zig с помощью команд, ссылка на GitHub в первом закреплённом комментарии.
  • Клонирование репозитория Nullclaw и сборка проекта с минимальным размером бинарного файла.
  • Решение проблемы с отсутствием SQLite 3 путём установки библиотеки.

00:02:50 Настройка Nullclaw

  • Вводная настройка Nullclaw, выбор модели на основе Olama.
  • Настройка параметров: использование SQL Lite, выбор режима «Supervised», настройка каналов Telegram.

00:05:21 Создание бота в Telegram

  • Создание нового бота в Telegram, получение токена API.
  • Замена токена в конфигурационном файле Nullclaw.

00:07:15 Тестирование и взаимодействие с ботом

  • Запуск демона Nullclaw и тестирование в терминале.
  • Взаимодействие с ботом в Telegram, проверка статуса и работоспособности с помощью команд null claw status и doctor.

00:08:55 Заключение

  • Подчёркивание компактности и эффективности Nullclaw.
  • Призыв к подписке и поддержке канала.

In this video

Chapters
Transcript
Search transcript
Chapter 1: Introduction to null claw
0:01
1 second
This video is about null law. Nulaw is a fully autonomous AI assistant built in
0:08
8 seconds
zik which is a low-level system programming language. In this video, we are going to install null claw locally
0:16
16 seconds
and I will be integrating it with local based models. This is Fad Miraa and I welcome you to the channel. Please like
0:24
24 seconds
the video and subscribe and consider becoming a member. You can see that we have covered almost all of open clause
0:32
32 seconds
angles and all the variants and flavors on the channel already.
0:37
37 seconds
Please also follow me on X if you’re looking for AI updates without any hype.
0:41
41 seconds
Let’s get started with this null law. As I said, this is built 100% in Zigg, which is a low-level systems programming
0:49
49 seconds
language designed for performance and tiny binaries similar in spirit to C, but safer and more modern. And the
0:58
58 seconds
result is remarkable. This null claw is just a 678 kilobyte single binary that
1:04
1 minute, 4 seconds
uses about 1 MB of RAM and boots in under 2 millisecond making it the smallest and fastest member of the
1:12
1 minute, 12 seconds
entire open claw family. Let’s get started. I’m going to use this uh Ubuntu system. Let me quickly show you. Let me clear the screen.
Chapter 2: Installing Zig and repo
1:23
1 minute, 23 seconds
Sorry. This is my GPU card and I am using it for this Olama’s local model and my system is Ubuntu. Let me quickly
1:31
1 minute, 31 seconds
show you the version. This is all I am using.
1:36
1 minute, 36 seconds
Now first up what I need to do I need to install Zigg. These are the commands to install ZIG.
1:43
1 minute, 43 seconds
And don’t worry about all of these commands and configuration. I’m also going to put the link in the first pin comment of this video so that you can
1:50
1 minute, 50 seconds
access it from my GitHub repo. So let me run this.
2:00
2 minutes
Zigg is installed. Now let’s get clone the repo of null claw. And I will drop the link in video’s description.
Chapter 3: Building the binary
2:07
2 minutes, 7 seconds
And that is done. And now let’s build this. If you look at this line, it’s very interesting. It tells the ze compiler to build the binary optimized
2:16
2 minutes, 16 seconds
for the smallest possible file size rather than for speed or debug information. Even though you also have that option and the Z is building.
2:28
2 minutes, 28 seconds
Interesting. Okay. Now you might also get that error and that seems bit scary but if you look closely all it is complaining about is
2:37
2 minutes, 37 seconds
that SQLite 3 is not installed. So what we need to do here we just need to
2:44
2 minutes, 44 seconds
install SQLite dev3 library and that shouldn’t take too long and then we can rerun that command. That’s all there is to it. So you see that this is done.
2:54
2 minutes, 54 seconds
Let’s rerun our zigg build again.
2:59
2 minutes, 59 seconds
And now it is running fine. As you can see here the build is done. Everything is ready.
3:05
3 minutes, 5 seconds
You can also move it the b binary to your bin directory and check out the version.
3:12
3 minutes, 12 seconds
And now you can also run your on boarding. Same we do with open claw. So you see that our null claw is now asking
Chapter 4: Configuration wizard
3:20
3 minutes, 20 seconds
what we are trying to do. I’m just going to select the number seven. And this is good that you know now they are actually
3:26
3 minutes, 26 seconds
asking us our lama based models. Uh you can even just skip it and do it just at your own. All you need to do is to put a config file.
3:37
3 minutes, 37 seconds
So I have just put in seven. It has selected our base local model. Enter API key. You don’t need to enter anything.
3:44
3 minutes, 44 seconds
Just press enter here.
3:46
3 minutes, 46 seconds
And then it is asking for the model name. It is just giving some of the other models which I don’t think so you should be using their very old models.
3:55
3 minutes, 55 seconds
Um so I’m just going to give it my own model name which I showed you already installed.
4:01
4 minutes, 1 second
And for memory and stuff I’m just going to you know go with the default option which the SQL light.
4:08
4 minutes, 8 seconds
I don’t want to configure any tunnel here. I will you know just select the none here because you know I just want
4:16
4 minutes, 16 seconds
to integrate telegram with it without any tunneling.
4:21
4 minutes, 21 seconds
Now this is where your preference comes into play. Do you want the agent to ask you before it does anything or do you
4:28
4 minutes, 28 seconds
just want it to go crazy with full autonomous? I’m just want I’ll go with supervised here. And now it is asking me if I want to configure channels now or not.
4:39
4 minutes, 39 seconds
So as we want to configure Telegram 2.
4:41
4 minutes, 41 seconds
So I’m just going to press Y here. And this is where it asking me to edit the channels in config file after setup.
4:49
4 minutes, 49 seconds
This is good. We will also check it out.
4:52
4 minutes, 52 seconds
And then it is asking me what should be the workspace. Go with default if you don’t want to change it. And everything is now good. And it has also tested it.
5:02
5 minutes, 2 seconds
Uh that what we need to do. You can start chatting with it or you can just check out what sort of configuration file it has created here. So let me
5:11
5 minutes, 11 seconds
quickly show you it to you through my VS code and then we will see what exactly it has generated.
5:21
5 minutes, 21 seconds
And this is our Olama configuration in that config.json. As you can see the model is done and rest of the options
5:29
5 minutes, 29 seconds
are the same which we covered but there are no channels and this is a config file which you should be using and as I
5:36
5 minutes, 36 seconds
said I will give you this config file with all the commands which you can check in the first print command. So rest of it is same all you need to do is
5:45
5 minutes, 45 seconds
to configure your bot in telegram and then give your bot token here. Now let me show you how you can create your own
Chapter 5: Telegram bot setup
5:51
5 minutes, 51 seconds
bot in telegram. I already have done few videos on it. But let’s do it again.
5:57
5 minutes, 57 seconds
So this is a telegram running on my system. Just click on bot further on the left hand side. If you can’t see it, just search for it at the top. And from here just type newbot.
6:09
6 minutes, 9 seconds
It is going to ask me, okay, what should I call it? I’ll say fad null claw uh or maybe null claw bot. Let’s do it full.
6:21
6 minutes, 21 seconds
And then username same and this is a token which it has given me and I’m just going to use it for this
6:30
6 minutes, 30 seconds
video discard it. So don’t worry about copying it or trying it you you know this one just create your own uh because
6:38
6 minutes, 38 seconds
I’ll revoke it and revoking is very simple. You just go to a bot further and just type here /mybot like
6:46
6 minutes, 46 seconds
mybots and then select your bot. This is the bot API token and then just revoke this
6:54
6 minutes, 54 seconds
current token. It will give you another one. So make sure you do that if you’re also using it in public. Anyway, we have got our token now. So I’m just going to
7:04
7 minutes, 4 seconds
go back to my config file and then replace this token here.
7:15
7 minutes, 15 seconds
Save it. And that’s all you needed to do as far as this config file is concerned.
Chapter 6: Running the agent
7:21
7 minutes, 21 seconds
And now in your terminal, you can simply test it by running this command. You can run whatever you like. So it is sending it to Olama. There you go. So our null
7:30
7 minutes, 30 seconds
claw is configured with our Olama based models.
7:34
7 minutes, 34 seconds
And if you want to use telegram first, you need to bring the demon of null claw up. If it is not already up, just run
7:42
7 minutes, 42 seconds
this command. And you can see that the telegram polling thread has started. It is listening here.
7:49
7 minutes, 49 seconds
So just go back to your telegram and then you can select your bot here. Click on start and it is going to there you
7:57
7 minutes, 57 seconds
go. You see nullclaw has responded that hello fad I’m null claw and model is this. And then you can just start
8:05
8 minutes, 5 seconds
talking with it and you can start talking with your null claw like you can ask it does size matter? Well, looking at this 678 kilo
8:13
8 minutes, 13 seconds
binary, I would say not in the way you think. And there you go. Look at the smiley. So here there I see we have had
8:22
8 minutes, 22 seconds
some fun conversation about size matter and you see in the double quotes. So model knows what it is dealing with here.
8:31
8 minutes, 31 seconds
And you can also check the status of null claw by running simply null claw status. It is going to tell you what exactly is the status, which model is
Chapter 7: Diagnostics and conclusion
8:40
8 minutes, 40 seconds
configured and then which channel is configured. As you can see here, you can also you know get out any security
8:48
8 minutes, 48 seconds
related issue or any health checks by running this doctor. So you can see that it is telling you what is screen what is working. Okay. And if there would be any
8:57
8 minutes, 57 seconds
uh you know warning or anything you will also get it here. And this is the warning by the way the curl one which it
9:05
9 minutes, 5 seconds
is telling you about. So look pretty good and impressive but you know I don’t think so it is bad at all. Um that is
9:12
9 minutes, 12 seconds
null claw a 678 kilobyte binary built-in sake running a full AI assistant stack on a you know machine. Um if this is
9:21
9 minutes, 21 seconds
what small looks like I’m not sure we need anything bigger than that.
9:28
9 minutes, 28 seconds
That’s it. Let me know what do you think. Again please like the video and subscribe and consider becoming a member. Please also follow me on X if
9:35
9 minutes, 35 seconds
you’re looking for AI updates without any hype.
Sync to video time

Поделиться: