programming

[quasi furnace]

Dlss is multiple features, upscaling would decrease vram usage, I think frame gen does, not sure what else tho

[mighty thorn]

Did someone say 3060 12gb?

[unkempt citrus]

If you're just gaming 8gb 3060 is fine

I do have a 12gb 3060 myself

[quasi furnace]

5050

[nocturne olive]

Yes I did say 3060 12GB
One of the people I work with has one

Work with as in NeuroSynth

[quasi furnace]

If neuro were to be trained on mass amounts of player replays instead of autos would her osu gameplay be more human ?

[nocturne olive]

You can't exactly train an osu bot like that and expect it to work as efficiently
Reinforcement learning is just the most efficient for that

[quasi furnace]

I see

and would it be possible to create something similar locally ran?

[nocturne olive]

Also machine learning makes the average of its training anyway so it'd average out the human inaccuracies

[nocturne olive]

Well Neuro's osu bot was made on a 1080Ti if I remember correctly
Could be wrong

[quasi furnace]

wth

[unkempt citrus]

It's not an llm, it has much lower requirements

[nocturne olive]

Vedal used a 1080Ti until he eventually just decided to get a 4090

[quasi furnace]

are there any documentations on other other osu/target detection and acquisition bots

[mighty thorn]

I have 3

[nocturne olive]

That's insane

Same VRAM as my 2 GPUs combined as a raw amount

Though of course not as efficient to use

And nowhere near the bandwidth

[quasi furnace]

my 750 ti has 2gb 💔

[nocturne olive]

This is my system

[quasi furnace]

sick

what psu lol

[nocturne olive]

Don't worry about the 20GB VRAM used it's just an LLM for Xoda's school project

[nocturne olive]

It used to be a 1000W PSU but I was having instability so I changed it but that didn't help but now it's on a 1200W PSU

[autumn ore]

lule

[keen ravine]

other servers I'm in seems to be looking at implementing OCR to combat the new tactics by the looks of it.

[unkempt citrus]

OCR is so overkill

It's the same set of images everytime

A hash will do it

[keen ravine]

i was going to think of doing the image hashing but given how many samples i looked at, each looks same but they look slightly different for some reason.

[unkempt citrus]

There's ways to account for that too

But that could be jpeg compression

[keen ravine]

could be. i found a message in another server that had a png extension instead.

[unkempt citrus]

You could also use a similarity score too

[keen ravine]

yeah i guess so.

[rigid snow]

heuristic to trigger -> check images for known hashes -> if not, ocr -> if bad, cache for step 2 and ban

[unkempt citrus]

Pretty much

Honestly the heuristic will catch most of them

[autumn ore]

holy mply how many languages died vedal use lol

[unkempt citrus]

4 images posted in multiple channels really quickly?

[rigid snow]

just messages with links/images posted from one user in multiple channels in a quick enough succession

4 images too specific

what if they make it 5

[unkempt citrus]

True

But also its always the same 5 images

[keen ravine]

i just tend to take the first image and ignore the rest

[real sierra]

c# experts

im running into a common hurdle of asyncs and idk what the C# workaround is

[autumn ore]

hb!

[real sierra]

im overloading a synchronous function but i need to call an asynchronous function from it

how do

offload to thread pool?

[autumn ore]

vedal names folders like me lol

[real sierra]

        NetworkIdentity newCharacter = Task.Run(async () => await SpawnCharacter()).Result;

wait

you can't call any unity functions from outside the main thread?

so you can't call any of them from an async context?

[frozen hollow]

Isn’t that what coroutines are for

It’s been a while sorry

[real sierra]

the core issue im having is that

i need to create objects on the server

but the client is the one that triggers that

[frozen hollow]

IIRC when you async stuff you can’t access a lot of things you think you should be able to

Properties of objects and such and at least when I did it it didn’t tell you it was going to fail

[real sierra]

but calls to the server with this library can only return void or tasks

[warped narwhal]

does unity not support RPCs for this? I have a feeling it would be better to have some kind of event get sent to the server, and the server calls the function

[real sierra]

so if you need the return value, it must be run in an async context

[real sierra]

im using purrnet https://purrnet.gitbook.io/docs

sits on top of unity networking

documentation is pretty terrible though

[warped narwhal]

https://purrnet.gitbook.io/docs/systems-and-modules/remote-procedure-call-rpc/generic-rpc looks like there are RPCs

[real sierra]

yes there are

im using them

my issue is not doing the client-to-server call

[fleet jungle]

probably i18n

[real sierra]

my issue is that the server apparently cannot create the object in an RPC

protected override void OnSpawned(bool asServer)
    {
        base.OnSpawned(asServer);
        if (asServer) return;
        Debug.Log("Spawned Character");
        NetworkIdentity newCharacter = Task.Run(async () => await SpawnCharacter()).Result;
        bindCamera.characterSprite = newCharacter.gameObject;
    }

    [ServerRpc]
    private Task<NetworkIdentity> SpawnCharacter()
    {
        return Task.FromResult(Instantiate(networkIdentity,  transform.position, transform.rotation));
    }

[fleet jungle]

No way that number is actually just counting the amount of letters in that one stupidly long word 🤦

[warped narwhal]

yup, same way that kubernetes is called k8s cause there is 8 letters inbetween

[opaque wharf]

Flatbuffer

[fleet jungle]

Ohh I hate this...

[warped narwhal]

happy birthday

-# you too shiro 🎂

[real sierra]

not the serialization issue

separate issue

[opaque wharf]

Thanks a lot!

[hard delta]

i learned this like twice already
and forgot both times

[opaque wharf]

Yeah, I mean using flatbuffer solves a ton pf headache by abstracting the rpc away

[real sierra]

im actually really lost as to why this doesnt work

if i cant call unity functions from RPCs then this is gonna become way more complicated

i'll need to have a main thread handler that pulls from a queue

and have RPCs populate the queue

[fleet jungle]

Well, at least my curiousity has been quenched about that 8.

[real sierra]

and then some structured request format

[frozen hollow]

Kubern8es

[fleet jungle]

Thought it was some weird German thing I wasn't understanding

oh

[frozen hollow]

No no I’m joking

[fleet jungle]

That's actually clever

oh but the e would have to be consumed by 8

[frozen hollow]

Kubern80s

[fleet jungle]

That... works... (?)

[real sierra]

oh wait

i think im being stupid

[opaque wharf]

Just do a hack and perform the RPC on the main thread

[real sierra]

nah i found the issue i think

[tender river]

happy birthday

[opaque wharf]

Thanks chay!

[opaque wharf]

Now I understand how you feel being a forever birthday boi lmao

[tender river]

and shiro

happy opposite of that

[real sierra]

no nevermind

idk what the issue is

or well i do but

i dont know of an easy fix

this must be on the docs somewhere right

no way nobody has ever had to make an object from an rpc

[opaque wharf]

That is what protocol buffer, flatbuffer, and capnproto solves

[real sierra]

i would still need a receiving server

that isnt solving any of the issues i already have

[autumn ore]

does anyone have the firmware for the lava lamp? if it exists

[real sierra]

@rare bridge

[idle dune]

I'm getting really sick of people doomposting about PC parts.

Most of these people are going to run Minecraft at 1080p and complaining they can't get a RTX 5090

[opaque wharf]

Well, yeah. The server still needs to be developed lol. But, you get a nice reflection and code generation for serialization and deserialization to transfer over the wire

Protocol buffer and capnproto even proides the rpc transport so you don't have to think about it too. But I prefer flatbuffer with something like zeroMQ (or nanomessage next gen)

[hard delta]

the problem isn't just the high end stuff tho it's that everything is going up

[idle dune]

What did I say that was filtered?

[real sierra]

if i get to the point of making a whole receiving server then my current solution already works

[rare bridge]

bro thinks i have lavalamp firmware

[real sierra]

but ideally i can just write it to work in an RPC

[real sierra]

nah, i just figured if you didn't have it then nobody did

[rare bridge]

true

[rare bridge]

what you thinking of doing

[azure lynx]

you know what would be a funny solution to not having the API for the lava lamp?
Make a library which watches the stream and clips just that corner and samples it to get the color, then sends that to control your off-brand lava lamp

[real sierra]

so it seems like the way im supposed to do spawning in this library is to spawn on the client and then have the server validate the spawn

[opaque wharf]

That's a good thing no? Otherwise people could spawn anywhere where they shouldn't have access to

[real sierra]

its good, just not the paradigm i expected

i just rewrote half the client side spawn testing code in expectation of spawning on the server and replicating to clients

now i have to undo it all

[opaque wharf]

[real sierra]

oh well

at the end of the day, it's easier this way

[opaque wharf]

I think having the client request a spawn is also a good thing because you can check for connection issue?

[real sierra]

still not nearly enough documentation on this feature

https://purrnet.gitbook.io/docs/systems-and-modules/network-identity/spawning-and-despawning/client-spawning-validation

this is all i get

[autumn ore]

workign on a app that links to home assistant and lets you choose lights you have set up in there, then during stream, the lights change to neuros lavalamp, i started doign tis project as i asked on reddit and it seems a lot of people would love it https://www.reddit.com/r/NeuroSama/comments/1qciy6i/how_would_you_guys_like_a_software_that_changes/ the api would be completely MY SIDE so, im going to somehow make the appication get lava lamp status from a api that i make on my side, and only have the clients get that api. The server would also connect to a phone app, where you can edit the lights that change (i have a google play license and i can use web clip configs for ios) so im looking for a way to actually get the lava lamp color, and im thinking if i use the web requests that the lamp gets to the wifi on my private server, themn have that extend to a api, users will have both a api for lava lamp and a app that lets them automatically change all their lights to it

[rare bridge]

interesting project. i did think about getting strip lights on my desk to compliment the lavalamp + act as a general way to light up my desk when i've got my light off. unfortunatly currently i don't know how to get the lavalamp color that neuro has set. i know it's possible to control the lamp itself via an API (and i already do this, my lavalamp is in home assistant), but i can't get the color set by neuro unfortunatly

but what you're trying to do here is get the whole "neuro chooses color" thing, i don't really know how that works

[opaque wharf]

By connecting to the same server that the lamp connects to

[rare bridge]

that'd be tuya's servers

[opaque wharf]

And either masquerading as a lamp or do a funny mitm

[rare bridge]

hello yes i am a lavalamp with 8 compute cores and 16gb of ram pls trust

[rare bramble]

I mean surely you can read some wire output to get the color and just route that to another light

thought I dont have the lamp so I dont know how much you would need to pull it apart to do that

[opaque wharf]

HW modification is usually not preferred when possible

[unkempt citrus]

You can just take the lamo apart

the wires are expossed

[opaque wharf]

...what?

[opaque wharf]

So neuro is connected to tuya? Nice to know

[unkempt citrus]

Well not exposed but visible

[rare bridge]

most likely vedal just has an API he can fire requests to which change the lavalamp color

[unkempt citrus]

I thought it would've been done by a pubsub server

[rare bridge]

i think tuya is the pubsub server most likely

[unkempt citrus]

Having a universal seed would be my preferred option but you'll need a pubsub anyway

[opaque wharf]

Yeah, I have a developer account on their platform and they do offer pubsub iirc

[autumn ore]

we can use a pc hotspot and wireshark to listen to the requests the lamp is making maybe?

who here has a lavalamp

[opaque wharf]

That is the funny mitm thing. But if the firmware generator is any competent, they would do tls with cert pinning

[fleet jungle]

Well, this is embarrassing. I for some reason thought win95 could run on less than 125MB storage, but trying to test install it now just proved me wrong.

[autumn ore]

i mean we can try?

we will need someone weith a lavalamp

[opaque wharf]

Yes, sure. I just don't have a lavalamp myself so I can't experiment much

[unkempt citrus]

The LEDs are in the base not the glass so you can just splice the wires

[real sierra]

disclaimer

[autumn ore]

well this guy has one

[opaque wharf]

I'd love to have neuro control my light too

[sick owl]

Tip for any UK bros here, if you're sick of Imgur region blocking you

https://greasyfork.org/en/scripts/551401-imgur-proxy/code

Ez proxy script I just found to bypass that

[real sierra]

vedal ai and juniper are not responsible for and damage or harm that may come from use of the lavalamp outside of what is described in the manual

[rare bridge]

^^^\

[real sierra]

please tinker responsibly and at your own risk

[rare bridge]

bigger disclaimer don't try to modify the lavalamp pls i don't want you burning your house down or murdering your lavalamp

[unkempt citrus]

Side note, pretty sure the universal adaptor it shiped with don't conform to electrical stnardards

[sick owl]

Redirects through duckduckgo

[real sierra]

this

[autumn ore]

you down to help us try get the tuya api?

[unkempt citrus]

Nope

[opaque wharf]

Yeah, even when I know what I'm doing, I still don't like it when I do HW mod

[unkempt citrus]

Pssh wheres you sense of fun

<- guy who licks batteries

[fleet jungle]

lol

[sick owl]

Obviously as with most community tampermonkey scripts if you do install it make sure to disable automatic updates

[fleet jungle]

I was just recalling how that was something I did whilst fooling around with electronics

Managed to zap myself a couple of times witha transformer and a 1.5v battery

[autumn ore]

Doesn’t have anything sensitive in it so I think I’m ok to post this here

[opaque wharf]

Here is the API
https://developer.tuya.com/en/docs/iot/subscribe?id=Kbwtw7fhhjabw

[autumn ore]

what is this for?

[warped bolt]

(Wait you're supposed to do that? )

[autumn ore]

holon, i know wher id and secret, but not the url

i know the id and secert

but not the url

[unkempt citrus]

The url might be hardcoded so never logged

[autumn ore]

the secret is in plaintext in the apk

[sick owl]

Stops a formerly benign script from being updated with malicious code

Just good practice

[warped bolt]

Ah yeah fair

[sick owl]

Obviously you have to check code before approving updates anyway but most just hit okay without a second thought, and there's usually no reason to update userscripts unless they break regardless

[opaque wharf]

Vedal sent the message to the broker and then the device receive that. That is the API to use on vedal side. The embedded firmware is generated by tuya

[autumn ore]

not the full key

but its there

i mean i cropped it out

thats what i mean but not full key

[warped bolt]

icic

[autumn ore]

i belive that the url is in the js

whoever is good at reading it, please look through it

[unkempt citrus]

18MB

[tacit sage]

I finally figured out how to make my own LLM Model o: I even have it in a discord server im so proud of myself

[warped bolt]

Ayy congrats

[tacit sage]

I named her jackie XD

[frozen hollow]

Jackie!

[tacit sage]

I only have the capabilities of using Ollama with my current pc I think but I want to upgrade her eventually

[autumn ore]

would vedal put the tuya servers in us or eu?

i foud the url

[tacit sage]

im trying to figure out how to get her to respond without having to @ her tho

[rare bridge]

gonna tell you right now

[autumn ore]

would vedal put the tuya servers in us or eu

[rare bridge]

going via the app will be a dead end

[tacit sage]

Really?

Whats a better way

[rare bridge]

the app doesn't know the color of the lavalamp when it's in swarm sync mode

[tacit sage]

XD

I wanted to like make her own discord

[sage crag]

https://tenor.com/view/30000-usd-gif-13674633028161831364

[autumn ore]

vedal hosts the servers on eu i see

[sage crag]

rare cloudburst

[tacit sage]

Also I have no idea lol

[rare bridge]

rare pull from neurocord gamba

[sage crag]

https://cdn.discordapp.com/attachments/1421861060507602944/1425936142754123927/neuroCatModeOfOfnnfOOf.webp

[frozen hollow]

Good question I don’t know (yet)

[tacit sage]

I mean I suppose I can post my code if you wanna take a look at it to see if im making any mistakes

do I screenshot it or just like post entire thing

[frozen hollow]

Oh nah I am doing something similar and genuinely don’t know either

[tacit sage]

Damn XD

[frozen hollow]

Just have to do more research

[tacit sage]

what if we made them look at every message as if it were a command prompt

like saying their name triggered a command prompt that allowed them to see it as if their being @'d

[frozen hollow]

I think you use a classifier system in the middle on every message to determine when to actually prompt the chatbot to look

But I’ve not tried to get it to work yet

[lost hill]

This is just null prefix, isn't it? Discord disallows this because listening to all messages becomes a mess.

[tacit sage]

Hmm then we can specify for them to look at every message that specifically says their name?

[azure lynx]

can you make it check if you are replying to its message?

[rare bridge]

you can enable getting message content

[lost hill]

Just make their name a prefix, I suppose. It's the same as Sapphire's s?, except just a name.

If you are making a Discord bot, it must be at the start of the message though

[rare bridge]

there is no limit on requiring a prefix for any sort of interaction if you've enabled message content intents for discord bots

[lost hill]

I just don't know how they are pulling this off. I assume they are going with standard Discord bots

[tacit sage]

yea she's able to be in discord with just a standard discord bot and I hooked my python code to discord so she kept her personality

[lost hill]

I love that the first result about message content fetching on StackOverflow is just null prefix command_prefix=""

[jagged turtle]

hey cloudburst can I borrow the first half of your name

[sage crag]

https://media.discordapp.net/stickers/1432138279679033354.png?size=160&name=YouShouldBeFrightened

[tacit sage]

Damn it

Jackie crashed

[sage crag]

spoilers

[tacit sage]

Oh it wasnt that, that crashed her I was trying to tweak how she responds to my discord ID and it crashed her XD

[lost hill]

Yeah, message content will be the way then. It's much easier to get the full string and extract data from that then bully Discord API when they explicitly advise against null prefixes

[burnt marsh]

lava lamp and esp32 ordered

neuro lava lamp is soon

[autumn ore]

we are working on a api to get color of the lavalamp

[burnt marsh]

oh rlly? i was just gonna see if vedal would make it public, but havent gotten a response

[autumn ore]

i dotn think he is

[burnt marsh]

yeah, me neither, but it doesnt hurt to ask

[wary mauve]

I assume it'll be the same voice, just trained off Ashleys voice and yeah just nneed to find something expressive

and one i can train off of

[tacit sage]

Fwaaaaaaaa

I think I have phantom bugs

Sigh I think im gonna have to start jackie over again

got too ahead of myself in trying to put her into discord I think

[wary mauve]

whos jackie

[fleet jungle]

jackie is great! (who tho?)

[tacit sage]

my LLM I started and tried to integrate into discord so I could speak to her outside of console

[wary mauve]

what're you sturggling on in the discord side, i did this a couple days ago

using discord py

[tacit sage]

I was trying to get her to respond to me without having to @ her like just having a normal conversation without @ing me or me having to her

and it ended up giving me an error I just reverted her before the error but this kinda sucked

python ended up crashing entirely

[wary mauve]

oh so you're using text, thought you meant voice calls

but still

[tacit sage]

yeaa I dont have her in voice calls yet

I just made her today basically

also omg its already almost 4pm wtf

[wary mauve]

I mean the best way I can say is take every chat message that comes in, and send it off to the LLM, and if you want rules like "Only reply if contextually the conversation involves you" then you can write a strict system prompt to do so, only issue is you're prompting EVERY message, but you can always use a keyword to get her to know when to do an LLM call, you can even pass in the past x messages as short term memory sort of

[tacit sage]

ooh you mean put the message in her memory?

I didnt think about making a short term memory

I only made a long term

I thought the short term would just come automatically for some reason

[wary mauve]

Put the previous messages in the system prompt, and say

Past Conversation:
{messages_here}

then youve given them context of whats been said

no no, LLMs have no context of previous requests

you have to feed it what you want to get the result you desire

[tacit sage]

oooh

im learning so much more about LLMS now

[wary mauve]

But you can make it so you don't need to @ her by sending EVERY message incoming to the LLM, but like I said thats a request per message, openrouter that i use is pretty cheap so I don't see it as a big issue if you're just messing around and learning so

[fervent jolt]

Wdym llm have no context of previous request? You could just save the session and then append every new user request?

[wary mauve]

guess im spreading misinfo, didnt realise sessions were a thing, my bad :)

[fervent jolt]

That's how ai chatbot wrappers do, they save chat session and just append every new user message

[wary mauve]

I'm fairly new to LLMs / AI as a whole too, so take what i sain with a grain of salt

[wary mauve]

does the provider do the session handling?

[fervent jolt]

No, you need to configure that yourself. Create a mono script for it to save chat session

[tacit sage]

Well. Jackie is something im creating for a long term project. I kinda wanted her to be like a "Super fan" Of Evil and Neuro XD

[fervent jolt]

Then send those chat session with new appended user message to the ai provider

[wary mauve]

any docs so I can see what you mean more? I thought you meant that the LLM provider would return a session id you can reuuse

[fervent jolt]

That's for caching

[wary mauve]

mono script? like unity mono?

bit lost my b

[fervent jolt]

Any language script just one file (mono) to handle the chat session is enough.

What's the use case? I don't know what you guys talking about, i only read the part where you mention llm doesnt know previous message

[wary mauve]

I was explaining a simple way to give past conversations back from discord, which would be giving the system prompt the previous messages snippet so the next run it has that context

which sounds like what you're explaining but in a stronger way

[tacit sage]

I feel like this is making sense but idk how i'd get this into my code

[wary mauve]

but i assume instead of system prompt you ca just send it in the messages=[] part of the openai api ( my reference)

[fervent jolt]

I see, so you need exactly a chat session in order to retain context

[tacit sage]

Do I just make another folder specifically for short term context?

[fervent jolt]

Basically it will be like this:
Chat Session:
SYSTEM PROMPT
USER PROMPT
AI Response
Save the session then append new
USER PROMPT 2
AI Response

It will then back and forth

This way ai retain contrxt of all chat session

[tacit sage]

oooh so I gotta string it together then?

[wary mauve]

Well when you make the request to the LLM you can send a "system" prompt, which tells the LLM, Hey I'm giving these instructions to your system to follow, alongside your "user" request.

In that system prompt you want to send, you just make it say

Here is the previous x messages of this conversation:
Username: message
BotName: Message
Username: Message

[wary mauve]

You got an example, I assume a chat session is a specific thing? sorry bit sluggish rn and also new to LLMs too

I know chat completions

[fervent jolt]

Chat session is basically a chat history

[wary mauve]

ohh yeah so a session id here is what you're on about

[real sierra]

[tacit sage]

why not just give it an entirely seperate session id for it to reference each time it needs to remember something and have each session in its own numbered folder?

[fervent jolt]

This is exactly how i want this channel to be

[real sierra]

so i figured out how to get the client to request server for spawn auth but there's no way to automatically grant the client ownership over something they spawn

only the server can grant that ownership

[tacit sage]

oh

[real sierra]

but i can't do that without the network id of the thing that was spawned

which doesnt exist at the time of spawn validation

[opaque sigil]

[wary mauve]

Oh you're talking about PERSISTANT memories, I just meant per session, but persistent memories can easily be a JSON file, but obviously you limit what you send otherwise too much noise, but this is just "chat" history which is a form of memory

[tacit sage]

yessssss I want it to remember specific things across all sessions so like for example how I can shape its personality with the system prompt can I also make it remember everything through just the JSON file?

[fervent jolt]

You need tools for it, and store the memory in RAG

[tacit sage]

RAG?

[fervent jolt]

Tools to fetch memory

Create a tool similar to mcp but local

[tacit sage]

oooh I see

[fervent jolt]

Yea that RAG

Sort of a database

[wary mauve]

You'd want a RAG database if you're trying to pull relevant memories based on the input you give it, instead of giving the LLM all your messages

look into chromadb if you're using python, pretty easy to use

but read up about using RAG for memories

[tacit sage]

ahhhh hahaha ok that makes so much sense. im gonna read up more on rag for memories thank you so much

one day little jackie I promise you'll be watching neuro and evil with me in no time lol

[wary mauve]

https://github.com/kimjammer/Neuro/blob/master/memories/readme.md

Here's where I learnt about it, this readme, this is a specific implementation of one way to do memories which I use but good little read

[tacit sage]

you guys have been massive help thank you again

[fervent jolt]

You want to recreate your own neuro?

[tacit sage]

pretty much idk if I wanna have her be a streamer but I am basically trying to create my own neuro

[fervent jolt]

Goodluck

[tacit sage]

just also trying not to copy vedal directly as well

wouldnt want him mad lol

[wary mauve]

all good, im in the middle of rewriting my "neuro" like bot, got a v1 prototype, goal is for it to just hang out really and be background entertainment for discord friends for like 2 weeks until i get bored

[rigid snow]

[wary mauve]

imo being copied is flattery

[fervent jolt]

Btw that repo is already old, llm right now are exponential so you might need to find new repo for sources

[tacit sage]

trueee but if I do start having jackie be a little streamer I probably wouldnt use the exact prompt she has rn XD its pretty much just evil

[wary mauve]

what do you mean exponential?

and yeah its old but its still a decent resource to start from, its not 1:1 ofc but thats not the idea

[fervent jolt]

Its old compare to exponential rate of how llm development is going right now

[wary mauve]

Yeah not sure how filtering works, I understand its most likely another model trained on being sfw

[wary mauve]

oh yeah for sure, but its a good basic ground of where to begin imo, good little skeleton

for something basic

[tacit sage]

Im not worried too much about filter atm I wont worry about it unless I decide I want jackie to be a streamer XD

[wary mauve]

yeah do small things at a time

[tacit sage]

Exactly I havent even gotten her designed yet cuz idk how commited I am to her being something I can see outside a text box XD

[fervent jolt]

Will you be fully api based or local model?

[tacit sage]

Api

I also want her to play video games with me so I want that full API

idk about minecraft levels yet though I was thinking like some visual novels as a start when that happens

honestly if I can get her to play a danganronpa game at the end of the day ill be satisfied with her game playing capabilities XD

[tidal vapor]

so I'm playing a game called stormworks where you can make microcontrollers where you use blocks or lua code to control different stuff like logic, math, video, audio, etc.
i decided it was a great idea to fetch code I dont understand at a time where I am way too tired to fix it and of course I have to resort to barbaric means to accomodate this peculiar situation

so, gaze upon my blocks and weep. all this is doing is taking a bunch of numbers in a composite signal that are 3 spaces apart and putting them in another composite signal but 1 space apart since I dont need all the other values in between. it also quickly pulses the information instead of sending it for like a whole second.

i put this here solely to pain your eyes.

[tacit sage]

tbh that looks both painful and beautiful too me XD

[tidal vapor]

i am displeased to announce it probably broke something

or maybe it was broken in the first place and this fixed it

[tacit sage]

XD

[fiery anchor]

how about both

[tidal vapor]

oh yeah something's definitely broken its for a radar graph and it now says its detecting the same reflection for like half of the fov

[autumn ore]

fah its frida time

[tidal vapor]

now it does nothing new lmao

gonna make a new simpler thing tomorrow anyway just wanted to try a graph. stupid idea cause im making tech that it already automates for me

by simulating the reversal of the automation

[amber fractal]

Making a circuit for my code is definitely on my bucket list, I need to actually finish said code first before I attempt any circuit making.

[tidal vapor]

one of the sentences of all time thats for sure

[amber fractal]

Yeah, it sure is

[tidal vapor]

programming is a great way to learn to take precautions and have ways to check yourself and your work to know where you went wrong
on the other hand, it's 1:30AM and I am very impatient

[amber fractal]

The classic two sides of every programmer

[tidal vapor]

hmm... spend 20 mins setting up some proper way to debug or spend 4 hours fixing a bug that was actually just you being an idiot and forgetting to connect something

this applies to anything in that vein actually. the amount of times I have spent hours trying to fix things in Stormworks only to forget to put a constant "1" on the pump clutch(if i dont it doesnt work) is insane for the amount of time ive had it (like 3 weeks)

[tacit sage]

I wonder how many times Neuro has crashed before vedal got it right

I need to take my ADHD medicine this is getting hard to focus XD

[azure lynx]

no idea, but it's been 000 days since somebody tell vedal that there's a problem with my AI

[tacit sage]

?? neuro's been live after the subathon?

[azure lynx]

that's what the sign says in her room.

[tacit sage]

neuro and vedal?

oh lol

AH SHE REMEMBERED ME IN PROMPT

It wasnt entirely killed lets goooo XD

[opaque wharf]

We do a little trolling. So the server pre generate the id, then the data is populated by the client. After it has been authorized, grant it to the client

[tacit sage]

THE TOOL IT EXIST AND SHE USE IT

wait

wha

oh no nvm

[real sierra]

server can't pre-generate the id

the id originates from a component, which can only exist attached to an object

it's part of the prefab that gets spawned

its ok i found a sketchy workaround ish

the client simply sends an RPC to the server to ask for ownership

[opaque wharf]

I mean the server generate the object itself, id included. The client populate the data after the request to spawn is completed, then authorize

[real sierra]

the server cant actually do that many checks on whether the ownership is supposed to be granted if i do it this way

[tacit sage]

user_input = input("You: ")
File "<frozen codecs>", line 322, in decode
KeyboardInterrupt
what does it mean keyboard interrupted

[real sierra]

but for something like a character, the server knows where the spawn locations are and if you already have a character

so im sure its fine

[opaque wharf]

It's fiiine

[real sierra]

https://gyazo.com/5d5babc2eeacc725eff61d8289cd0d5f.gif

i added a little status readout on the menu

it looks so official

[burnt marsh]

is the source released anywhere? how does it work?

[autumn ore]

no unfortunaltely

it connects to tuya

[burnt marsh]

ah ok ok

what info does tuya provide? does it just look at the stream and find the color? or is it more advanced

[opaque wharf]

You may also want to check out tuya convert to see other device existing firmware packet capture

[autumn ore]

idk, vedal has a api on tuya that has the color, we are just finding it

[opaque wharf]

Tuya is a white label company. They provide all in one solution from hw manufacturing to infrastructure for IoT device

[burnt marsh]

yeah, i know about that api, just unable to find it

[opaque wharf]

They even have a mobile app generator that you could just plop your brand and ship it to google play / app store

[autumn ore]

i going to use frida to grab all the keys

[burnt marsh]

well i wait to see your results! would love to use this in my project

[autumn ore]

im working on a application to change all of your lights to the lava lamp

[burnt marsh]

neuro controls your ceiling light lol

if it doesnt work out i will release an api that just takes the colour from the stream, but that might be slow and wouldnt work on every stream, such as the 3d ones

[simple sky]

My roommate is considering trying out Linux

[burnt marsh]

[simple sky]

Because Mint is based on Ubuntu, does it do that thing where if you use apt get, it defaults to snap packages?

[opaque wharf]

Hmmm, Neuro API that lets Neuro control stuff for her viewer (or at least subscriber) would be wack

[burnt marsh]

thats a thing, just she can control only lava lamps because the api is closed source

[opaque wharf]

Yeah, what I meant is that an API that the swarm could use

[real sierra]

all the issues with player input were because of a ! that wasn't supposed to be there

[azure lynx]

would need to be (a) opt-in, and (b) something that didn't really matter. with those hurdles to development, i don't think anyone would bother.

[opaque wharf]

Of course it is opt in. You should register yourself after all to get the token for authenticating. A public API would overload the message broker real quick

[azure lynx]

you could just make something where she only has to visit page A or page B and have that control something. i think she has limited webpage reading access now?

[opaque wharf]

And for point B, since this is an API, I assume someone knows what they are doing when giving neuro control over it

[burnt marsh]

heres the repo of my api, nothing done so just being posted for visibility, any contributions are also appreciated

https://github.com/alyfrillnya/NeuroSync

[real sierra]

why are unity colliders so STUPID

not make any sense

[obsidian mantle]

whats up there

[opaque wharf]

The physics or the geometry?

[real sierra]

the basic function of them

they just dont collide

or detect

or do anything

[obsidian mantle]

using wrong functions?

[opaque wharf]

Oh so the physics. Yeah, unity physics tends to do that

[obsidian mantle]

there is also 2d collider and normal collider

and if you use normal collider functions on 2d collider it doesnt work

[real sierra]

it was working until i tightened the network rules 🥲

[opaque wharf]

They are numerically unstable

[real sierra]

yes im using 2d colliders for my 2d game

[obsidian mantle]

smart

[opaque wharf]

So one wrong optimization can indeed eff shit up

[real sierra]

collisions are just not happening now

🥀

at all

period

[obsidian mantle]

how does it even work? who is processing it, server or clients?

[real sierra]

server should be

[obsidian mantle]

is server side visualized somehow

[real sierra]

not really

host is both client and server

[opaque wharf]

You mean the executable can perform as both client and server when needed?

[real sierra]

[opaque wharf]

Noice

[real sierra]

tho i wont be distributing the host version

at least not unless i open source or something

[opaque wharf]

[burnt marsh]

okay yall i suck at coding and am very new, can anyone tell me the best way to analyze a twitch stream and grab a color hex from a certain area [neuro's lava lamp] i think ill do it in python but if thats not the best language then id accept other ideas

[opaque wharf]

Easiest way is just perform screen capture every now and then

Code bullet on youtube will show you how to NOT do that properly

[obsidian mantle]

will it look bad because colors are changing slowly

[real sierra]

a video stream is a super complicated state machine

getting image data from that is not trivial

even worse if live

[burnt marsh]

yeah, that seemed like the best way, but if the lamp updates quickly and fluidly then it might look bad

[azure lynx]

but that requires the video stays in the same place. better would be to stream the video from discord and sample frames from the stream. once you have a frame you can use opencv probably to analyze it.

[burnt marsh]

yeah... i know this isnt the best way but its kinda the only way

unless tutel changes something

[opaque wharf]

If you want to grab the stream directly, you're in for a ride. You need to know how twitch performs the auth, then you need to make the client to send the right request, and then parse those, and only then you get a frame

Unless twitch provides an API

[obsidian mantle]

isnt the app available to everyone

[burnt marsh]

swarm sync?

[azure lynx]

you don't need to sign in to watch

[obsidian mantle]

yes

[burnt marsh]

yes but i dont have a lava lamp, i missed the preorders so i want to make my own

[obsidian mantle]

imagine ad pops up and your lavalamp takes random colors because of that

[opaque wharf]

But I bet they still have some measure to protect the stream itself

[azure lynx]

you need a web browser.

[burnt marsh]

id figure out how to prevent against that ig if i dont give up because this doesnt look very feasible

[sick owl]

Okay so update on my path tracing saga, I've found it can actually look amazing with one big caveat

[sick owl]

I have to disable frame gen

[opaque wharf]

Is the video stream paused when it is not in focus?

[sick owl]

Turns out the big thing that was causing the image instability I was complaining about the other day was frame gen doubling the persistence of existing artefacts from dlss performance and ray reconstruction

[real sierra]

omg

im gonna lose it

the physics debugger only works for 3D colliders and rigidbodies, not 2D

[sick owl]

When running at 60fps these artefacts are unnoticeable but they appear as fizzling and wobbling with frame gen on

Of course what that then means is I have to deal with sample and hold motion blur if I want a good image while using path tracing

[azure lynx]

and if you can do it with a web browser, you can do it with python code. it doesn't have to be a real web browser. just something which can make requests and read responses with some processing to choose what to ask for and how.

so you don't get ads or "not in focus". you only need like a 144p stream to check the color

and you only really need it every few seconds probably

[burnt marsh]

if you are subbed to vedal thats a non issue

[obsidian mantle]

yeah you have to be subbed or the source where you take the video from has to be subbed

[burnt marsh]

okay this seems actually possible now

[obsidian mantle]

then he moves the lavalamp or it disappear and it stops working this solution will never be good enough

[opaque wharf]

That's... what I'm saying? You need to figure out how twitch make the request. They could also have some DRM for the stream itself hence why I said some form of authentication

[obsidian mantle]

i think someone tried to revers engineer lavalamp sync

was it quack

[opaque wharf]

If you want to emulate a browser anyway, it's best to just use a browser automation like selenium or pupeteer

[burnt marsh]

maybe if the swarm sync app makes calls to the tuya api you could get those with wireshark? seems like the best implementation but also the hardest

[opaque wharf]

But then the question become, does the videostream get paused if the window is not in focus?

Sure audiostream may still work, but the video? I do not know because I never do it before. I can deduce that audio still works for obvious reason that I can still hear the audio even when I switch to another program

[azure lynx]

if there is no real window, that's up to you. it only pauses because the window sends an "i lost the focus" message.

[burnt marsh]

id assume so because if im coding and watching stream the video still plays

[opaque wharf]

Is the lava lamp connected to the internet directly or does it need to connect to your phone via bluetooth at stream?

[obsidian mantle]

https://discord.com/channels/574720535888396288/1350944219115683871

maybe this can help

[quick condor]

Welp I am about to end up giving myself (what I consider) "Vedal brain rot" AKA request timing data logging to reduce perceived latency across my project stack

[burnt marsh]

i think when it connects to your phone it connects to the internet, so that wouldnt work, but if the app for some reason makes calls seperate from the lamp it could work, though i dont know why it would

[opaque wharf]

If the lava lamp is connected directly to wifi, and only use the phone to enter the wifi credential, then the firmware is the client and you cannot capture the packet easily

[burnt marsh]

yes thats my point, it wouldnt be for the light, which is why calls are unlikely, so its an if

[obsidian mantle]

or is it something different

it looks like it takes stream

and some colors from it

[burnt marsh]

yeah it does, i might base my implementation off that

[obsidian mantle]

i guess its not so big of a project for it to have issues because of it

[burnt marsh]

[opaque wharf]

The firmware performs some crypto handshake with the built in id when doing OTA. So you can't just capture it easily (see the tuya convert project). But I don't know if they do the same for the pub/sub connection, I assume it would be the same (at least mtls + certificate pinning)

[quick condor]

I wouldn't trust gpt with anything code related

[burnt marsh]

i think its a good tool as long as you know what its actually writing and if its reliable

[quick condor]

It was ok when 5.0 first came out but I legit have to check codex line for line because it will edit files without reading... Claude is far better at just working

[burnt marsh]

oof. yeah thats not great...

[real sierra]

guess i'll die then

[obsidian mantle]

what does it mean

oh i get it

can you calculate collides manually

do you have all collider dimensions and positions and

oh you want physics

[opaque wharf]

How tf could they mess it up that badly lmao

[obsidian mantle]

what if you make distance very close to 0

maybe it skips collision moment and just clips through

[opaque wharf]

Make the distance -Inf for funsies. Or NaN

[real sierra]

what a shitass engine

[obsidian mantle]

figured it out?

[real sierra]

if i knew nothing would work i would've gone with rpgmaker

[real sierra]

no

doesnt make sense to me at all

[quick condor]

Me every time gpt goes schizo because my project violates most of its guardrails but doesn't do anything technically wrong

[obsidian mantle]

what exactly did you change in networking rules

you said it worked before

[opaque wharf]

Godot

[obsidian mantle]

does rpgmaker support multiplayer

[real sierra]

let me try undoing that change and see if it fixes itself

[opaque wharf]

If it doesn't, either you are schizo or unity is gaslighting you

[real sierra]

still broken

[obsidian mantle]

idk what those rules are but maybe it started filtering something important for collisions

[real sierra]

even with the least strict rules

[obsidian mantle]

then its not the rules that broke it

[quick condor]

I love programs gaslighting you

[real sierra]

where the client can do anything

[opaque wharf]

So which is it?

[unkempt citrus]

Neuro lamp doesn't accurately show the irl colour on screen to be clear

Theres some varriation

it'll still work but won't be 100% accurate is all

[burnt marsh]

ughhhhhhh why is this so hard tutel please just release itttt

[opaque wharf]

It is not trivial to implement pub/sub for a large audience

Nor is it cheap

[obsidian mantle]

you probably need real lavalamp to reverse engineer it and its still not guaranteed to give results

[burnt marsh]

2$ a month for api access. seems doable

[unkempt citrus]

I still think splicing the wires is the easiest way

¯_(ツ)_/¯

[quick condor]

Same

[opaque wharf]

Yep, hence my idea to tie it to twitch subs since that is tied to discord already

[obsidian mantle]

you can probably sniff whats lavalamp doing and connect your pc pretending its lavalamp

but i wouldnt know details

[quick condor]

That's my plan for pixmob rfid fobs for an event

[burnt marsh]

what is that way, sorry if its a stupid question or joke because im dense as tungsten

[stray dragon]

if you have lavalamp it's actually quite simple to get the colors from it

[opaque wharf]

No, if they use mtls with certificate pinning, they would refuse to connect

That is mitm

[burnt marsh]

thats the issue though. i dont

[stray dragon]

oh

[burnt marsh]

i wouldnt need to do this if i had a lava lamp

[obsidian mantle]

how

[opaque wharf]

Splicing the wire is opening up the lava lamp and get the rgb signal from the wire directly

[obsidian mantle]

i dont own it but im curious

[burnt marsh]

oh, yeah. dont have one. sadly

[unkempt citrus]

Its an LED so theres 4 wires, 3 for each of the colours and 1 ground

so you can just hook up to the three and read the votlage to figure out the colour

[opaque wharf]

Have you checked it?

[azure lynx]

i remember they had a feature that was very important: turn it off if it's been on for 10 hours.

[unkempt citrus]

Unless Vedal used a completely proprietary LED thats how it'll work

[quick condor]

Yup

[unkempt citrus]

Not impossible, but unlikely

[opaque wharf]

It can be common cathode or common anode

Or it could be one of those WS2818

[unkempt citrus]

Thats true I guess, but the proceedure is roughly the same

[stray dragon]

you can use tinytuya in python script on pc to ping lamp and ask for current state info (this includes light color and such)

tinytuya requires some info to setup, but the only thing i had trouble with was the local key, which can be obtained through the debug output in the swarm app

you can be connected in the swarm app while also having tinytuya access as far as i know

[burnt marsh]

unless some wonderful person wants to do that and make it public i have to figure out another way

[opaque wharf]

The problem is, does neuro still able to control the color when connected to the local network?

[stray dragon]

that i did not specifically test for, but i believe it's possible

probably requires you to not set anything from the tinytuya side

[opaque wharf]

Oh, you're right

tinytuya does not act as a tuya server, but an api gateway

[stray dragon]

uhhh no not really, it acts over local wifi network

api calls are only to find the device as far as i know

i do think you can have it use api calls instead

[opaque wharf]

No, your device (the lava lamp) does not connect to tinytuya

tinytuya instead make a request to the lava lamp. The lava lamp itself is still connected to the tuya infrastructure

[stray dragon]

let me test this real quick by unplugging the wifi usb

[opaque wharf]

So Neuro, theoretically can still control the color since that comes from the tuya pub/sub, and you can poll for the current color via tinytuya through your local network. The lava lamp act as a server

[burnt marsh]

okay, for anyone with a lavalamp, what happens to the color duing 3d streams? ditto with when evil is on

[stray dragon]

wait i don't have one anymore?

oh right motherboard switch

[obsidian mantle]

but you need real lavalamp to establish connection first... can this be bypassed

[opaque wharf]

Wdym? A lava lamp?

[stray dragon]

wifi usb

[opaque wharf]

Nope

[stray dragon]

but i'm confused now because i don't think i have the wifi capabilities passed through

[obsidian mantle]

and you cannot sniff whatever lavalamp is sending to the server and just copy it?

or the key id is only issue

[fleet jungle]

@mossy fossil YES, my friend! You can, in fact, run win 9x on DOSBox-x.

[stray dragon]

key id is solved issue

[obsidian mantle]

can you reuse the key for multiple connections

[opaque wharf]

The point of tinytuya is to act as an api gateway. That is why tinytuya need the IP, device ID and local key of the lava lamp. They make the request on the local network (with the lava lamp acting as a server from tinytuya POV)

[stray dragon]

the local key is unique to the tuya device (in this case, the lava lamp)

[stray dragon]

ah that fits

i see now

anyways

problem: i need to buy a router to deal with college network bullshit so i can actually use a smart lamp i got
question: what do i look for in a router, and are they good/cheap if i buy used

[obsidian mantle]

is this 100% the case here

[quick condor]

Don't ask me, I use a nsa6650 and a ubiqui ap

[obsidian mantle]

i dont understand how its mitm if you just make unique connection to the server

[stray dragon]

ew expensive

but why expensive

i do not understand

[amber fractal]

Doing a homelab and lazy I imagine

[quick condor]

Ding Ding Ding! Can't be bothered to set something up manually

[unkempt citrus]

What does an IoT even need out of a router

wifi range?

[opaque wharf]

Just my deduction due to minimal amount of resources online

[opaque wharf]

Are you talking about tinytuya or having a tuya server emulated on the local network?

[stray dragon]

the issue: i am on a college network, and some smart devices i want to set up connect to the network and then don't expose their ips, so i cannot automate them via script
i cannot obtain the ips because i have no control over the college network
i could maybe phone the IT department and ask them to look up the IP associated with the specific MAC address of the device, but the ips for these kinds of things tend to change, as i've observed with the tuya bulbs, and phoning them every few weeks sounds like a hassle for both parties here

[stray dragon]

above message is in response to this

[opaque wharf]

Do an arp-scan

[stray dragon]

i tried :)

[opaque wharf]

Dangit

[unkempt citrus]

If the IT team is remotely comptent, I'd expect they'd lock down the network to prevent new routers from being added

[stray dragon]

wait seriously

[burnt marsh]

yall are the questions i ask annoying and bothersome? sorry i just have to get this off of me because i tend to overthink about things and am not the most educated in this matter... sorry if this is weird

[opaque wharf]

How would one block a router from joining a network? Genuinely curious as I don't study networking much

[unkempt citrus]

You would just prevent new unapproved ethernet devices from being assigned an IP

[opaque wharf]

Nah, technical questions are fine here

[opaque wharf]

And how would the admin knows it is a router?

[burnt marsh]

oki. just sometimes feel like im being stupid and asking useless questions

[unkempt citrus]

I would expect them to just block ethernet connections as a whole

Whitelist some MAC addresses

and just block

[quick condor]

I use to manage college networks. Comptent is not a word that I would use to describe their teams

[unkempt citrus]

Thats fair

[opaque wharf]

I see, yes that is plausible. I was thinking one of those router that connects to other network via wifi thing (bridge)

[amber fractal]

Just do what some of the american internet companies are doing and just have a wireless upstream

[opaque wharf]

Wtf

[quick condor]

Fixed wireless is better than dsl or satellite if available

[stray dragon]

i have an ethernet connection from my PC to the wall right now, and i have a way to register MAC addresses on the network so they'll work

[unkempt citrus]

Then you might be fine

[quick condor]

I work with several wisps

[unkempt citrus]

you could also just call up IT to check

[opaque wharf]

Oh, P2P style? But then again, bad weather could mean no internet too

[stray dragon]

i suppose

[unkempt citrus]

On my own network, I can see that the other router we have plugged in is identified differently as it has a permanent assigned IP

Identified during the inital communication protocol if anything

oTHER HAVE a leased time

[quick condor]

Same goes for sat, fixed wireless is less susceptible though

[amber fractal]

I wish I knew myself, but if you only need like 100mbps connection at home then it is fine for those people.

[opaque wharf]

That's true I suppose. But nothing beats physical wire

[knotty obsidian]

Hey,can somebody explain why the heck does the phone have a slowed internet connection at night ? Even though I have full data.

[opaque wharf]

Maybe congestion due to how many people are using it at the same time

So you, your neighbour, their kids, and everyone else hit the same cell tower at the same time

[knotty obsidian]

Oh.

[amber fractal]

everyone just scrolling through reels

[quick condor]

100% agree but fiber to a fixed antenna is better than dsl or nothing at all

[knotty obsidian]

i live in India.

[opaque wharf]

The 4K VOD watcher:

[knotty obsidian]

So,does that affect the connection ?

[stray dragon]

don't the full connection bars only mean that you're not dropping any packets sent between phone and the big antenna?

[amber fractal]

that's just the avarage netflix enjoyer

[knotty obsidian]

Idk

[stray dragon]

signal strength and congestion are different

[knotty obsidian]

I know man.

[opaque wharf]

I don't know anything about that. My theory is congestion because if they have full data and good signal at daytime, the only other culprit is too many people is using jt

[stray dragon]

if you're right by antenna it'll show full bars no matter how slowly the antenna is handling your connection

probably congestion

[amber fractal]

The amount of resources the internet towers have isn't infinite, so everyone trying to do stuff at once will be too much for it eventually

[stray dragon]

people all using phone more at night sounds likely unless this is at 3am or something

[unkempt citrus]

Peak vs offpeak sounds about right to mne

[amber fractal]

Most likely watching all forms of video content

[burnt marsh]

(mods dont airstrike me ping was turned off)

[quick condor]

Example of people getting home from work and using the internet

[unkempt citrus]

If ping is turned off he won't see it unfortuantely

[burnt marsh]

it doesnt show the little thing?

[unkempt citrus]

Nope

Best you could do is a silent ping

which still sends the red without a notification noise

done by attaching @silent to the start of the message

[burnt marsh]

ah

[unkempt citrus]

like this

[burnt marsh]

@silent @unkempt citrus worked?

oop

[unkempt citrus]

Nope

[burnt marsh]

@silent test

[unkempt citrus]

nada

[burnt marsh]

[unkempt citrus]

What version of discord are you using

[burnt marsh]

latest, vanilla

[unkempt citrus]

strange

Are you on english?

[burnt marsh]

one more try, lemme restart it

[unkempt citrus]

as your langauge or differnet

[burnt marsh]

yes

english

[unkempt citrus]

[opaque wharf]

tinytuya is your best bet

[burnt marsh]

how would that integrate with neuro

[opaque wharf]

You don't directly interact with her. She change the lava lamp color, and you can get those color

But you can't send anything to her

[burnt marsh]

yes, like how would i get the color

[burnt marsh]

test

ooh yes

[unkempt citrus]

There you go

[burnt marsh]

ok ill send it to vedal again

[opaque wharf]

tinytuya. Technically you just ask the lava lamp "hey, what is your current colour?"

https://raw.githubusercontent.com/jasonacox/tinytuya/master/docs/TinyTuya-diagram.svg

[burnt marsh]

any updates? would love to use this, maybe could be paid or integrated with a twitch sub

[opaque wharf]

Better not lol

[unkempt citrus]

https://tenor.com/view/rip-bozo-gif-22294771

[burnt marsh]

silent ping

[burnt marsh]

what i do

[unkempt citrus]

Ping vedal

[burnt marsh]

i asked mods and they said its fine with silent ping

[unkempt citrus]

I'm just messing around

you'll be fine

[burnt marsh]

okii srry

[opaque wharf]

Have you seen this diagram?

[burnt marsh]

really? you don't need any key?

[burnt marsh]

yes

[opaque wharf]

You do, but the process to get the key is not that hard

[burnt marsh]

enlighten me

[autumn ore]

anyoen down to help me get the api?

[opaque wharf]

You make a tuya developer account, pair the lava lamp with the smart home app (not the swarm sync), and get the local key there. It's in the tinytuya repo

https://github.com/jasonacox/tinytuya

[autumn ore]

we still need tohe api for the color of the lava lamp

[burnt marsh]

okay, without a lava lamp though, i cant really get it. also, what proot said, also, i don't know what id do to make my esp32 to make those calls

[unkempt citrus]

Esp32 can do internet calls simply enough

Its designed for IoT after all

[burnt marsh]

oh, do i need a special library for tinytuya?

sorry if this is an obvious question, im new

[unkempt citrus]

https://github.com/FrBerger83/EspTuya

It'll be easiere with onee

[burnt marsh]

ooh that repo looks helpful

[fleet jungle]

Just saw "microsoft java"... Wow.

[autumn ore]

who wants to help me find the api

[burnt marsh]

wish i knew how, best i can do is wish you luck in finding it!

[autumn ore]

anyone have a lavalamp device id i can use?

[opaque wharf]

You make the call to tinytuya. Or you can see how tinytuya make the call to the local device

But yeah, you cannot have neuro control anything without owning a lava lamp first

[burnt marsh]

yeah...

[opaque wharf]

Unless someone is generous enough to share their API and provide it. But that would violate tuya ToS real quick I think lol

[sick owl]

In conclusion having now played a few hours with the new settings sample and hold was a mistake, frame gen sucks, DLSS 4 performance mode is black magic at 4k and path tracing with RR looks really nice on it if you don't throw frame gen on top

[autumn ore]

i just realised i can use proxy man on my phone lol

[patent shard]

you need like 80 base frame rate for frame gen to be potentially good

[burnt marsh]

yeah, read tos, doesn't explicitly say it but implies it

[patent shard]

way too much like screen tearing below that
but some people aren't sensitive to that...

[autumn ore]

the app doesnt like having a vpn on

[stray dragon]

found a router to buy, really hoping there aren't any more surprises with this damn smart bulb

[sick owl]

Screen tearing is a non-issue if you have variable refresh rate

My problem with it was that the generated frames were drastically exacerbating DLSS artifact persistence and temporal accumulation time

Artefacts are shown for twice as many frames and so where they're normally unnoticeable, they become very visible, and when ray reconstruction is accumulating it takes double the frames to do so and the presentation is more unstable since frame generation will be bridging between different levels of ray accumulation

[real sierra]

[patent shard]

well, it makes me think of screen tearing
it's basically just parts of the screen breaking up

[real sierra]

is it just over?

[jagged turtle]

"from a mailing list" uh huh

[real sierra]

do i give up on the project?

colliders are pretty mandatory

if unity colliders are just not working anymore then like..?

idk what to do?

[patent shard]

I realize this is the programming channel and I should be more technical

[faint sandal]

jimbo jambo space continuum quantum particles

[unkempt citrus]

steal someone else's project with colliders and just mod it into what you want

[autumn ore]

got the requests in the app

[obsidian mantle]

there must be a way

when did it break

you said it worked before then stopped

something broke it

[real sierra]

i might have an idea as to what broke it

[obsidian mantle]

trace it only you know what you did

[real sierra]

no i still have no idea

no wait maybe i do

i cant decide

no actually i might know what it is

[obsidian mantle]

make new empty project make colliders from scratch

[real sierra]

currently how its working is

[obsidian mantle]

if works - compare

[real sierra]

im sure an empty project will work

i dont actually think unity colliders are broke

i think it is a networking thing

[obsidian mantle]

is it only broke with multiplayer or if you try to detect collisions locally it works?

[real sierra]

currently the movement is done on the client and then replicated to the server through this NetworkTransform component

https://purrnet.gitbook.io/docs/plug-n-play-components/network-transform

the movement does replicate as expected

and is visible to other clients

however, im not sure if movement through this component actually invokes the physics engine to reassess the colliders

[opaque wharf]

Godot is whispering into your ear

[obsidian mantle]

can you make colliders visible somehow

for everyone

[real sierra]

i can look at them in unity

and its very obvious that they are intersecting

i also confirmed this with code

[autumn ore]

vedal please drop us the lavalamp api

[obsidian mantle]

who says this - server or client?

[real sierra]

server

[opaque wharf]

Maybe during the transform the geometry isn't translated very well

But then again, the SDF said negative

[real sierra]

my suspicion is that it's more of a "teleport" than a movement

[obsidian mantle]

why is distance negative

[real sierra]

distance is the shortest distance between the points of either collider

[obsidian mantle]

oh

[real sierra]

it's negative because they are overlapping

[opaque wharf]

That could be it. The physics engine struggle because maybe they use a differential somewhere

The discontinuity from teleport broke that

BUT THE SDF

[obsidian mantle]

how do you get "can collide"

[real sierra]

https://docs.unity3d.com/6000.3/Documentation/ScriptReference/Collider2D.CanContact.html

[autumn ore]

love that embed

[fast torrent]

Should I use Unity or Unreal?

[real sierra]

unreal

[fast torrent]

Alrighty then

[real sierra]

(source: i use unity)

[stable bough]

C# and unity
UE is trash with TAA imo

[fast torrent]

Oh

[obsidian mantle]

are you sure unreal doesnt have other even more bizzare problems

[stable bough]

ifu want blury gross games, use UE

[real sierra]

how is this a bizarre problem

this is colliders not colliding

[autumn ore]

I CANTQAJSFKAFAISF ASRFFFIND THE FUCKCING APIIIIIIIIIII\

[real sierra]

this is like

a fundamental feature

[obsidian mantle]

but it worked before

[opaque wharf]

How do you check is colliding?

[real sierra]

i knowwwwwwwwwwwwww

[fast torrent]

Is unreal that bad?

[real sierra]

https://docs.unity3d.com/6000.3/Documentation/ScriptReference/Collider2D.IsTouching.html

[stable bough]

TRASH, thanks why AAA games looks blurry and trashy

C# is easier to code right ?

[fast torrent]

Oh yeah the blur, I hate that too

[stable bough]

also linus torwards hates C++

[opaque wharf]

It is important to understand that checking whether colliders are touching or not is performed against the last physics system update; that is the state of touching colliders at that time. If you have just added a new Collider2D or have moved a Collider2D but a physics update has not yet taken place then the colliders will not be shown as touching. This function returns the same collision results as the physics collision or trigger callbacks.

Does the physics run?

[real sierra]

thats under the hood

no clue

[obsidian mantle]

filter

[real sierra]

its supposed to do that on its own

[obsidian mantle]

do you do it with no filter

[opaque wharf]

That is still an assumption. Remember screeps time running forward?

[autumn ore]

VEDAL PLEASE HELP ME

[real sierra]

i use the other signature

[obsidian mantle]

[real sierra]

please scroll up

about half a page

[autumn ore]

vedalllllllllllll

ive benn at my computer for 8 hours straight

[obsidian mantle]

but this is different

[autumn ore]

trying to find this goddamn api

[obsidian mantle]

one is (is colliding) other is (can collide)

one of them is lying

[real sierra]

??

it can collide but isn't

[autumn ore]

one can collide with the other

[real sierra]

neither are lying

[obsidian mantle]

are you checking if it collides with noFilter

to exclude filters

[autumn ore]

im tempted to ping vedal atp

[real sierra]

im not using filters

period

[obsidian mantle]

[autumn ore]

ill jsut work on the app part for now ig

[real sierra]

[obsidian mantle]

okok no filter

what if that library is designed for 3d

and uses some 3d functions

[opaque wharf]

What about checking the overlap? Does it shows up?

If it does shows up, then it would be hilarious lol

[real sierra]

if i have still somehow not convinced you they are overlapping

[obsidian mantle]

https://docs.unity3d.com/6000.3/Documentation/ScriptReference/Physics2D.GetContacts.html

[opaque wharf]

No, I mean from this
https://docs.unity3d.com/6000.3/Documentation/ScriptReference/Collider2D.Overlap.html

[real sierra]

https://gyazo.com/d8d2277aa646fdf99db83e663798cde2.gif

green circle is the collider in question

[obsidian mantle]

same thing if you come from outside?

[real sierra]

no collision

no matter what i do

[opaque wharf]

If the collider shows that it is overlapping with the collider that you check, then that would be funny

[real sierra]

going to try moving from the serverside instead of synchronizing it through NetworkTransform

[opaque wharf]

But if it is not colliding yet overlapping, then maybe it was the physics

[glad path]

hear me out: logarithms should be formatted like c++ type casts (at least for ascii notation)

[opaque wharf]

Why not log5(x)?

As in logbase(number)

[real sierra]

i

i found the problem

[glad path]

less clear

[real sierra]

its working now

perfectly, in fact

[glad path]

log<base>(number) is very clear

log<5>(x)

[real sierra]

i cant believe it was something this stupid

[opaque wharf]

Nah, log5 is cleaner

[real sierra]

it turns out

you need to use Physics3D instead of Physics2D

[opaque wharf]

Lmao

[real sierra]

even for a 2D scene with only 2D colliders

[glad path]

log<5> x?

[obsidian mantle]

[opaque wharf]

Why tf lol

[real sierra]

why

i dont know

[obsidian mantle]

why did it work before

[real sierra]

i have no idea

[obsidian mantle]

or were you not using this library yet when it worked

[opaque wharf]

Nope, log5(x)

[real sierra]

since the host is both server and client, i used to just load the scene locally, which would load it for the server as well

i switched to doing it over the network because it seemed a bit less hacky and better practice for if i was tightening networking rules

[obsidian mantle]

oh so it wasnt checked over actual network

[glad path]

log<5>(x) better and this is #programming so it makes more sense to use syntax from a programming language

[real sierra]

and only then did i have to add the physicsMode parameter

i picked 2D because of course! it's a no-brainer!

[obsidian mantle]

[real sierra]

[opaque wharf]

If we were coding then yeah, log<5>(x) is better

But for everyday use just log5(x) is good enough

[real sierra]

https://gyazo.com/bbf5474d3d4e3c65f1f7fe5ad5bbe654.gif

[glad path]

log5(x) could be confused with log(5) * x

would be hard to do

but

[opaque wharf]

Well, x is just the number. So log5(10) for example

[glad path]

hence log(5) * 10