If you want to explore the embedding approach though, I’d recommend CatBoost as it does something similar with internal kNN matching.
It uses decision trees which should have better performance than a basic kNN, can accept other features as well, and is still lightweight to run
semisol
npub122…cgrkj
2026-04-23 20:52:08 UTC
in reply to nevent1q…e4zc
Author Public Key
npub12262qa4uhw7u8gdwlgmntqtv7aye8vdcmvszkqwgs0zchel6mz7s6cgrkjPublished at
2026-04-23 20:52:08 UTCEvent JSON
{
"id": "e75d8e29df7d5426cfcac4f1a67535c99514d81b2704b4933182b218fd66c267",
"pubkey": "52b4a076bcbbbdc3a1aefa3735816cf74993b1b8db202b01c883c58be7fad8bd",
"created_at": 1776977528,
"kind": 1,
"tags": [
[
"e",
"bb2b59a5ec8683b22037fe5353dc7511e026ae150c669c4fda770da727da2aee",
"wss://relay.primal.net/",
"root",
"5cdbde0a550fc046a38d75e9fc238094f96b0165b8297c4fa69134ae4ec80024"
],
[
"e",
"73654644dbb32a525757d2d20258431f3437af9301e8403fc097138e75e0a9ad",
"wss://nos.lol",
"reply",
"52b4a076bcbbbdc3a1aefa3735816cf74993b1b8db202b01c883c58be7fad8bd"
],
[
"p",
"5cdbde0a550fc046a38d75e9fc238094f96b0165b8297c4fa69134ae4ec80024"
]
],
"content": "If you want to explore the embedding approach though, I’d recommend CatBoost as it does something similar with internal kNN matching.\n\nIt uses decision trees which should have better performance than a basic kNN, can accept other features as well, and is still lightweight to run",
"sig": "52d1c933e7e5d665b3d50be3db5cee4ef8caba4ea75de6b80cdc5e96102109c967c1c12e0937684adf2b217439eb197176d48edfbcd38670378e15c8c2767719"
}