every model in Command Code has vision, DeepSeek too!

Love this video format, super informative! Thanks for sharing Ahmad!

Oh great. Made another one. Will try to do more. We have so much to share that’s already shipped and not yet announced.

Vision for every model, that's crazy🔥

😎

Voice 🎙️ mode when? 🥹👉👈

Adding to our desktop soon. Will work in any app.

@AbdullahSh92729 There's a desktop app?

@AbdullahSh92729

🐐

One of the best features of Command Code

GOATED

really cool, it's something that if the model is smart enough it would already found a way to, but it was always a whole process, getting this into the harness makes things way more reliable

Hahaha that way of ending the video. Anyways super great content. Btw Ahmad, is it possible to add a sidebar inside command code harness?

Haha glad you liked it. It’s not. No plans either. Sidebar would look super odd. You can use mods to put stuff in the footer.

Ok. I’m missing something like token usage (in, out, hit)

🐐

seems very intuitive and straightfoward way to help text only models.

nice hair cut

😎

Great thank you Ahmad

I’m glad you like it

may you open source your bench?

That is a plan.

any promotion for group for commandcode?

Was this a pre-recorded video? cuz DSV4F just launched vision.

Dang. Recorded yesterday and scheduled it for today. And DeepSeek decided to mess with me.

wait this is actually great!!!

This is really cool awesome work Ahmad!

This is great, i was thinking about why this is not implemented in command code yet. But glad to know it, it’s there now.

What do you mean? We are the ones who came up with this. They suggest nowhere else. How could you be thinking about this not being here when nobody had this before. That makes no sense lol 😂

While using command code with deep seek model, i gave it a image to look into a issue in this, then got a response its text only model. Then this thought came in my mind that why can’t this transfer it to a vision model to extract the text? I hope it makes sense now!

"basically routing the request to another model"

very nice, similar to LookAt up @AmpCode guys, and honestly this is very helplful extend this to also support video: checking few frames per video maybe to understand what's going on