WEBVTT

NOTE Timing: local ASR word timestamp alignment. Review against audio before publishing.

00:00:00.000 --> 00:00:02.500
In this walkthrough, we'll download a small model

00:00:02.500 --> 00:00:05.840
through pi-kit, start a local server, and get

00:00:05.840 --> 00:00:08.780
our first reply: a welcome message for the

00:00:08.780 --> 00:00:10.880
fictional space station Aster Haven.

00:00:11.542 --> 00:00:14.802
Let's begin in settings. Open local models, then

00:00:14.802 --> 00:00:17.322
search Hugging Face for a small instruction model.

00:00:18.222 --> 00:00:19.962
We'll use this one for a simple greeting,

00:00:20.402 --> 00:00:21.902
so we don't need a large download.

00:00:22.667 --> 00:00:25.947
Before downloading, review the license, file size, and

00:00:25.947 --> 00:00:28.907
estimated memory. If those work for your machine,

00:00:29.447 --> 00:00:32.587
confirm the download. pi-kit downloads the file and

00:00:32.587 --> 00:00:34.227
verifies it before adding the model.

00:00:34.833 --> 00:00:37.713
Now open your model library. The finished download

00:00:37.713 --> 00:00:40.913
appears here. That's our model in place; next

00:00:40.913 --> 00:00:41.733
we'll start the server.

00:00:42.375 --> 00:00:44.535
The engine tools are already installed for this

00:00:44.535 --> 00:00:48.875
walkthrough. In the engines tab, start llama-swap. Wait

00:00:48.875 --> 00:00:51.095
until it's running before heading back to chat.

00:00:51.750 --> 00:00:54.050
In a new conversation, choose the local model.

00:00:54.830 --> 00:00:56.550
Ask it to write a short welcome for

00:00:56.550 --> 00:00:59.450
visitors arriving at Aster Haven. Send the message,

00:01:00.070 --> 00:01:01.930
and here's the model's first local reply.