dave can render notes now

Reply to this note

Please Login to reply.

Discussion

can dave suggest something like 'mostr notes people on nostr would like?'

could be an interesting DVM or feed of some sort, maybe

yes we have a relay index so i can get dave to query mostr notes

Okay but can notedeck render pictures yet? 😂 (jk, what youre doing is absolutely amazing)

of course it can…

Ah, Ive only tried out the public binary that just shows a link. Excited to see how far its come since

Although some pictures load? But the screenshot in this note just loads a link

yeah we’ll have a new binary release soon, but its also trivial to compile and run

Probably will tonight! Been planning on it, but havent gotten around to it. I have a feeling its gonna be worth getting Purple

Ahahahahah

how do I access Dave?

#asknostr

TH!S!

Its in heavy development at the moment. If you are ambitious you can download notedeck and compile it, with OPENAI_API_KEY set. You won’t have to do this when its released obviously

guessing this works with openrouter? trying now!

should in theory, I haven’t tested or have the ui for switching providers yet

so which model does it default to when you're using openai?

4o i think but its configurable in code

ok i feel so close lol

i ran it with the flag of the openai api key, and changed the base url to the openrouter api route, but clicking the block on the left doesn't do anything lol

NVM U CLICK THE DAMUS LOGO

ok yeah, so far not working haha

are you setting OPENAI_API_KEY in env?

yep

"2025-04-18T22:45:46.388962Z ERROR notedeck_dave: failed to get token: stream failed: Invalid status code: 401 Unauthorized"

even though it's literally the right key??

i'm running;

export OPENAI_API_KEY="[redacted]" which i literally just copied the key from from OR

I fixed this. you can set these to get openrouter working:

DAVE_API_KEY=$OPENROUTER_API_KEY

DAVE_ENDPOINT=https://openrouter.ai/api/v1

DAVE_MODEL="google/gemini-2.0-flash-001"

thank you!!

I will test with open router, but i haven’t used that before

cheers

it should theoretically work, not sure why it's not

openai compatible an all

You wont have to because it will be fully local? That was my understanding of the feature, although I understand why youd use an API for testing

it supports both, openai works better for testing because its not as dumb as the models that can fit on my 8gb vram gpu

Makes sense. As long as thats optional, thats all that matters

Is this available on purple/ios?

Who’s Dave?

damus nostr assistant with anonymous backend +

nwc pay per token

Can we use dave for free or it costs zaps..?

It’s free if you use ollama, but if you are using the anonymous lmzap backend it will be pay per token. Or you could use your own openai/claude/gemini/openrouter account (once i’ve tested all these)