№17317161[Quote]
Hello, I made some vichan hider thing with jeet tier vibecoding. can you guys test it for me and report issues?
https://dpaste.com/9WHZA8WQN (this is the userscript)
https://violentmonkey.github.io/Features:
Thread hiding
Image hashbanning
Hashban removal
Automatic image hashing
Vichan API support
Built-in MD5 hashing
Catalog support
Thread support
Live thread update support
Per-board thread hiding
Image hash caching
Separate hidden-content controls
Separate hashban controls
Hashban import/export
Aggressive/similarity hash matching
Compact control panel
Adjustable panel transparency
Vichan icon
Settings panel
Hideable UI
Persistent settings
Temporary status feedback
Local-only storage
№17317181[Quote]
SOMEBODY ADD THIS TO THE 'KI
№17317182[Quote]
>local only storage
marge?
№17317186[Quote]
nigga really needed to make this because salt fucked the site up so hard
gem
№17317188[Quote]
stole my session cookies award
№17317214[Quote]
>>17317161 (OP)oooh this is really gemmy
>>17317189 №17317230[Quote]
can you just upload this as a .txt? dpaste glows
№17317274[Quote]
>>17317251just checked it for malicious code, it's clean
№17317277[Quote]
>>17317262it's not good code it's a mess of gpt and claude. If anyone who actually is good at this thing wants to fix it up they would be a legend
№17317313[Quote]
>>17317297it is pretty late for me
THOUGH since i am a euromutt, so I probably will forget to make it gegggggg
№17317317[Quote]
Shift + Click image → Hide thread/post
Ctrl + Click image → Hashban image
Ctrl + Click image (Mac: Cmd + Click) → Hashban image
Alt + Click image → Remove hashban
Ctrl + Alt + H → Show/hide the Vichan Hider panel
№17317320[Quote]
Add this to the 'ki nowwwwwwwwwwwwww
№17317365[Quote]
I don't think similar hashing works that good
№17317376[Quote]
works
№17317398[Quote]
>>17317161 (OP)Put it on Github
№17317400[Quote]
>>17317262okay i checked the code. There seems to be a lot of query selectors for whatever reason, like a copious amount of them geg.
<redditThe similar image check works by downscaling the image and making it grayscale. I hecking forgor what exact resolution but that might lead to falsepositives.
<plebbitOther than that, good enough i guess if it works for regular hashbans. Could be rewritten to be less 'cado sized doe
№17317402[Quote]
>>17317384was testing on the /plier/ board and it wasn't working aswell as I'd like
№17317418[Quote]
>>17317402you wanted to block all pliers or what?
№17317425[Quote]
>>17317418no just for the purpose of the test I am a fan of 'pliers actually
№17317432[Quote]
>>17317425I doubt the similarity extends over different variants, its just to prevent if they changed some pixels to change to hash ig.
№17317447[Quote]
>>17317432from the code
/*
*
* SIMILARITY HASH
*
* This is deliberately separate from MD5.
*
* It creates a 64-bit dHash from a small grayscale
* representation of the image.
*
* Similar images can therefore have similar hashes even
* after resizing/recompression.
*
*/
№17317454[Quote]
>>17317447oh I see so it compresses the image or something
№17317455[Quote]
>>17317432someone can easily modify an image and but it through all kinds of filters or move it before spamming. Thinking this is a bit beyond the scope of things but the idea of the aggressive toggle is to be like an emergency option
№17317466[Quote]
>>17317447 const canvas =
document.createElement(
'canvas'
);
canvas.width = 9;
canvas.height = 8;
<redditthe compressed image is 9 by 8 for whatever fucking reason. That can easily lead to falsepositives
№17317490[Quote]
>>17317466I think it would be smarter to do pixel wise comparision or some kind of error metric with images you want filtered.
№17317503[Quote]
>>17317466Apparently a 9×8 input is actually the standard dHash construction but idk
№17317507[Quote]
>>17317490idk, pretty sure imageDNA and other anti 'p systems do the same shit with the compression and creating some sort of lower resolution hash to block and detect all possible variants of the image in question.
№17317521[Quote]
>>17317507yeah whatever ig its correct then
№17317548[Quote]
>>17317521ok before i go to sleep
/*
* Similarity threshold.
*
* 0 = identical
* 64 = completely different
*
* 10 is deliberately conservative.
*/
const SIMILARITY_THRESHOLD =
10;
found this in your code
№17317555[Quote]
>>17317548you can tweak (or make the ai jeet add this to UI) this value to make the similarity hash more strict or not
№17317572[Quote]
>>17317555I'm gonna try and implement this