Hey guys, we are back with another

video. This time I will be going through

some free ways to hit API endpoints

using things such as Mistral, Google

API, hugging face, a few different

things. Uh, previously I made a video

about how to use the hugging face free

API backend. Of course, all these do

have limits, but this is enough to at

least get you started if you want to

test some of your MCP servers and

different other things. Uh, I apologize

for my voice. I'm coming off an illness

this past week. So, hopefully at least

this audio is coming through clear

enough for you to get the point and I

should be back to 100% pretty soon. So,

we are back in the home lab. I have my

new Orange Pi 6 Plus running very well

in the Kubernetes cluster. I plan to

also make a video on that. If I jump

over here into Open Web UI, I have quite

a few Open AI API connections set up.

I'm going to go through each of these

and also show you how to get these set

up. So, starting off towards the top

here, this is the Google API, which is

very easy to get set up. If you go into

Google AI studio, get API key, and then

you can generate your own. As you can

see here, I have the free tier. It does

have a limit uh but it's fairly generous

in my testing. So, you can do a lot of

testing with that. Uh coming back, we

also have this local endpoint and I will

post the link to the other video that

talks about this, but essentially this

one is running off of my local machine,

my local laptop. It's just a simple

Python wrapper that exposes the NPU that

I have. So if you have one of these

newer Intel Ultra processors that has

the built-in NPU, you can leverage it

with this application that I built. It

doesn't allow you to have super powerful

models depending on your computer. Uh

but you can at least do some basic

things like I'll show here.

Tell me a story. [clears throat]

And this should come through. It's just

a 1.8 billion model. Nothing crazy. As

you can see, it's going to ramp up my

NPU and then also my GPU is going to

push a little bit here and it comes back

with the response. So depending on how

strong your laptop or desktop is, you

might be able to leverage that as well.

So coming back here, right, you also see

that I have hugging face, I have Gro,

open router, and Mistrol. So all of

these, as you can see, I have them open

here too. They all have free API key

offerings. Again, you won't get a ton of

usage out of this. It's going to limit

you after a certain number of calls, uh,

I believe per day, per month, depending

on each platform. But you can see here I

am able to run a decent amount of

requests even with strong models like

GPTO OSS 12 billion. Uh this is under

grow right and then hugging face. It

also shows you here uh it says 10 cents

which isn't a lot but you can get a few

calls. Uh and I'm not sure how often

this refreshes but um I haven't used it

up too often. Uh and then open router is

also pretty generous. If you come in

here, sign up for an account, create

your API key, you could do the same

thing. Uh, and then Mistral as well,

they have an offering where you can

create your own API key, hook it up to

open web UI and give yourselves a little

bit more testing capability without

having to spend a bunch of money on

something like cloud or whatever else to

actually give yourself that API

capability. Uh, so I hope this is

helpful. Hopefully this gives everyone

some additional ideas to try to extend

some of your testing capability instead

of getting stuck with uh just using a

single API provider such as hugging face

or something else you've heard of. And

yeah, as always, like, comment, and

subscribe and let me know what more you

want to see. Thanks.