WEBVTT

1
00:00:00.248 --> 00:00:01.331
This one's the how-to.

2
00:00:01.331 --> 00:00:05.366
After init, you'll want the router to fit your models, your budget and your rules.

3
00:00:05.366 --> 00:00:08.368
All of it is on the dashboard, and every change applies live.

4
00:00:08.772 --> 00:00:11.741
Everything lives in one file, and this page shows it.

5
00:00:11.741 --> 00:00:13.813
Every other page edits the same file.

6
00:00:13.813 --> 00:00:15.325
A change starts as a draft.

7
00:00:15.325 --> 00:00:19.862
Apply swaps it in, with no restart, and apply and save writes it to the file too.

8
00:00:20.115 --> 00:00:23.484
Say you just downloaded a new model in LM Studio.

9
00:00:23.484 --> 00:00:27.334
On the models page, discover asks LM Studio what it has.

10
00:00:27.334 --> 00:00:31.047
It suggests a tier for each model, and what it can do.

11
00:00:31.047 --> 00:00:33.660
Gemma reads images, so it gets vision.

12
00:00:33.660 --> 00:00:35.585
Add it, then apply and save.

13
00:00:35.586 --> 00:00:36.892
Want a paid API as well?

14
00:00:36.892 --> 00:00:37.763
Add a connector.

15
00:00:37.763 --> 00:00:42.389
Any OpenAI compatible endpoint works, like OpenRouter, Groq, or a server of your own.

16
00:00:42.389 --> 00:00:45.219
The key you type is encrypted into the secret store.

17
00:00:45.219 --> 00:00:48.866
The config file only holds a reference to it, never the key itself.

18
00:00:49.119 --> 00:00:52.542
Apply, then discover its models, and add the one you want.

19
00:00:52.542 --> 00:00:54.726
Give it its price per million tokens.

20
00:00:54.726 --> 00:00:57.559
That's how cheapest capable knows what it costs:

21
00:00:57.559 --> 00:01:04.229
local models still go first, then your subscription, and the paid model only when nothing cheaper can do the job.

22
00:01:06.088 --> 00:01:08.666
When prompts land on the wrong tier, don't guess.

23
00:01:08.666 --> 00:01:13.243
The evaluate page runs a labelled set through your configuration, and lists every miss.

24
00:01:13.243 --> 00:01:16.136
This one needed deep reasoning, but nothing matched it.

25
00:01:16.136 --> 00:01:18.766
Add it to the right route, and the set runs again.

26
00:01:18.766 --> 00:01:20.818
One more right, and nothing else broke.

27
00:01:22.946 --> 00:01:24.221
Profiles are policies.

28
00:01:24.221 --> 00:01:28.276
This one is local only, so nothing it covers ever leaves your machine.

29
00:01:28.276 --> 00:01:30.594
A rule picks it for the billing project.

30
00:01:30.594 --> 00:01:35.866
And the compare tool shows what each profile would do with a prompt, before you rely on it.

31
00:01:38.699 --> 00:01:44.900
The usage page shows what you spent, and what your subscription use would have cost at API prices.

32
00:01:44.900 --> 00:01:47.051
To cap the paid API, add a budget.

33
00:01:47.051 --> 00:01:48.570
Here, two dollars a day.

34
00:01:48.570 --> 00:01:52.619
Once it's spent, the router skips that connector until tomorrow.

35
00:01:53.926 --> 00:01:56.223
And when something seems off, run doctor.

36
00:01:56.223 --> 00:02:01.266
It checks the configuration, the embeddings, and every connector, without calling a model.

37
00:02:02.071 --> 00:02:02.932
That's the tour.

38
00:02:02.932 --> 00:02:05.032
Everything you saw is also in the docs.

39
00:02:05.032 --> 00:02:07.831
SemanticRouter.NET is open source, and MIT licensed.