karpathy

LLM Wiki

A pattern for building personal knowledge bases using LLMs.

This is an idea file, it is designed to be copy pasted to your own LLM Agent (e.g. OpenAI Codex, Claude Code, OpenCode / Pi, or etc.). Its goal is to communicate the high level idea, but your agent will build out the specifics in collaboration with you.

The core idea

Most people's experience with LLMs and documents looks like RAG: you upload a collection of files, the LLM retrieves relevant chunks at query time, and generates an answer. This works, but the LLM is rediscovering knowledge from scratch on every question. There's no accumulation. Ask a subtle question that requires synthesizing five documents, and the LLM has to find and piece together the relevant fragments every time. Nothing is built up. NotebookLM, ChatGPT file uploads, and most RAG systems work this way.

carlosedp

// Run with scala-cli httpserver.scala
// Save logback.xml file into ./resources dir
// Before generating native image binary, run first as above and make a real access to the URL
// so the native-image-agent can generate the metadata in ./resources dir.
// Then generate native image binary with: scala-cli package --native-image httpserver.scala
//> using scala "3.3.0"
//> using lib "dev.zio::zio:2.0.15"
//> using lib "dev.zio::zio-http:3.0.0-RC2"
//> using lib "dev.zio::zio-logging-slf4j2::2.1.13"
//> using lib "ch.qos.logback:logback-classic:1.4.8"

djspiewak

package sillio

import cats.syntax.all._

import scala.annotation.tailrec
import scala.concurrent.ExecutionContext
import scala.util.control.NonFatal

import java.util.concurrent.atomic.{AtomicBoolean, AtomicReference}

blagerweij

import java.io.ByteArrayInputStream;
import java.io.OutputStream;
import java.util.ArrayList;
import java.util.List;

import com.amazonaws.services.s3.AmazonS3;
import com.amazonaws.services.s3.model.AbortMultipartUploadRequest;
import com.amazonaws.services.s3.model.CannedAccessControlList;
import com.amazonaws.services.s3.model.CompleteMultipartUploadRequest;
import com.amazonaws.services.s3.model.InitiateMultipartUploadRequest;

matfournier

/** An effect for infinitely repeating an action, use with care
  * as there is no way to kill the process w/o shutting down the
  * server in this implementation */
trait RepeatEval[F[_]] {
  def repeat[A](fa: F[A], duration: FiniteDuration): F[Unit]
}

object RepeatEval {
  def apply[F[_]](implicit ev: RepeatEval[F]): RepeatEval[F] = ev

Daenyth

import cats.Eq
import cats.effect.{ContextShift, IO, Timer}
import org.scalactic.Prettifier
import org.scalactic.source.Position
import org.scalatest.exceptions.TestFailedException
import org.scalatest.{Assertion, AsyncTestSuite}

import scala.concurrent.Future
import scala.concurrent.duration._
import scala.reflect.ClassTag

mpilquist

// Alternative to sealed abstract case class pattern for Scala 2.12.2+
// Benefits:
//  - 1 final class instead of 1 sealed class + anonymous subclass
//  - portable to Scala 3 regardless of opaque types
//  - less boilerplate

final case class Angle private (toDegrees: Int) {
  // Define our own `copy` method to suppress synthetic one
  // Add private to prevent it from being used
  def copy(degrees: Int = toDegrees): Angle = Angle.fromDegrees(degrees)

Daenyth

import cats.effect.Sync
import io.chrisdavenport.log4cats.Logger
import cats._
import cats.implicits._
import scala.concurrent.duration._

class RetryExample[F[_]](implicit F: Sync[F], log: Logger[F]) {
  case class ApiError(msg: String) extends Exception(msg)
  
  private def getTempFromInternet: EitherT[F, ApiError, Float] = ???

Daenyth

import java.io.OutputStream
import java.util.concurrent.Executors

import cats.effect.{Async, Effect, IO, Timer}
import cats.implicits._
import fs2.async.mutable.Queue
import fs2.{Chunk, Stream}

import scala.annotation.tailrec
import scala.concurrent.{ExecutionContext, SyncVar}

yashdeeph709

Prerequisite : Public Key Setup is required for using this setup 
1. First open ~/.ssh/config file and add these lines filling the appropriate details.
This will setup a netcat tunnel through jump server which will forward all your traffic to mail machine.
 ```
 Host <HOSTNAME>
           User <USERNAME>
           HostName <HOSTNAME>
           ProxyCommand ssh <JUMPSERVER-USER>@<JUMPSERVER-ADDR> nc %h %p 2> /dev/null
 ```
2. ssh -v -D :port: :username:@:hostname: