retroryan

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.

retroryan

import zhttp.service.{ChannelFactory, Client, EventLoopGroup}
import zio.{Tag, ZEnvironment, ZIO, ZLayer}

trait NumService:
  val get: ZIO[Any, Throwable, Int]

final case class NumServiceLive(c: ChannelFactory, e: EventLoopGroup) extends NumService:
  val url = "https://random-num-x5ht4amjia-uc.a.run.app/"
  val get: ZIO[Any, Throwable, Int] =
    val request =

retroryan

#create class
CreateClass({ name: "ledger" })

#create index
CreateIndex(
    {
      name: "UNIQUE_ENTRY_CONSTRAINT",
      source: Class("ledger"),
      terms: [{ field: ["data", "clientId"] }],
      values: [{ field: ["data", "counter"] }],

retroryan

#Create a Class
CreateClass({ name: "orders" })

#Create an Index
CreateIndex(
    {
      name: "orders_index",
      source: Class("orders"),
      values: [{ field: ["data", "orderId"], reverse:true }, { field: ["ref"] }],
      serialized: true

retroryan

15kX9REWJS9v3NmT1JeeUN6NnovQk9WnWz

retroryan

Keybase proof

I hereby claim:

• I am retroryan on github.
• I am binaryknight (https://keybase.io/binaryknight) on keybase.
• I have a public key ASD4qP7q58L0A88jrHOloskDr5ixa3CT77MWpbiEtxO5Cwo

To claim this, I am signing this object:

retroryan

override val supervisorStrategy: SupervisorStrategy = {
    val decider: SupervisorStrategy.Decider = {
      case Guest.CaffeineException =>
        SupervisorStrategy.Stop
      case Waiter.FrustratedException(coffee, guest) =>
        barista.tell(Barista.PrepareCoffee(coffee, guest), sender())
        SupervisorStrategy.Restart
    }
    OneForOneStrategy()(decider orElse super.supervisorStrategy.decider)
  }

retroryan

sealed abstract class QueryType[T] {
  val data: T
}

case class IntQueryType(data: Int) extends QueryType[Int] with Ordered[IntQueryType] {
  def compare(that: IntQueryType): Int = this.data - that.data
}

case class StringQueryType(data: String) extends QueryType[String] with Ordered[StringQueryType]  {
  def compare(that:StringQueryType):Int = this.data.compareTo(that.data)

retroryan

import org.apache.spark.streaming.kafka.KafkaUtils
import org.apache.spark.streaming.Time
import kafka.serializer.StringDecoder
import org.joda.time.DateTime
import org.apache.spark.sql.SaveMode
import sqlContext.implicits._

 val ratingsStream = KafkaUtils.createDirectStream[String, String, StringDecoder, StringDecoder](ssc, kafkaParams, topics)

   val msgs = ratingsStream.transform {

retroryan

import scala.annotation.tailrec
def atoi(chList:List[String]):Int = {

  @tailrec
  def atoiAccumulator(chList: List[String], accumulator: Int): Int = chList match {
    case Nil => accumulator

    case head :: tail =>
      val tensMult = scala.math.pow(10, tail.length).toInt
      val nxtAccumulator = (head.toInt * tensMult) + accumulator