mutolisp

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.

mutolisp

SELECT 
	id, ST_BUFFER(geom_centroid, 0.0083333) as geom_buff
FROM 
(SELECT 
   1 as id, 
   ST_SRID(ST_GeomFromText('POLYGON((1 1,2 2,2 0,1 1))', 4326)) as srid,
   st_centroid(ST_GeomFromText('POLYGON((1 1,2 2,2 0,1 1))', 4326)) as geom_centroid,
   ST_GeomFromText('POLYGON((1 1,2 2,2 0,1 1))', 4326) as geom)
AS s;
	      

mutolisp

DROP TABLE IF EXISTS test;
CREATE TABLE test (
	id serial,
	long double precision,
	lat double precision,
	country varchar,
	species varchar,
	geom geometry(POINT, 4326),
	CONSTRAINT pk_test PRIMARY KEY (id)
);

mutolisp

library(sp)
library(gstat)
library(raster)

graycol <- gray.colors(4, 0.55, 0.95)
# use IDW to generate zn.idw

data(meuse)
coordinates(meuse) <- c('x', 'y')
data(meuse.grid)

mutolisp

mutolisp

# install sp package before use
# read coordinates csv
twTransform <- function(input = file.choose(), x='x', y='y' ,from = 'wgs84', to= 'twd97'){
  require(sp)
  coords <- read.csv(input, sep = ',')
  coords <- cbind(coords[x], coords[y])
  colnames(coords) <- c('x', 'y')
  pts <- coordinates(coords)
  # projection
  wgs84 <- CRS('+proj=longlat +ellps=WGS84')

mutolisp

library(data.table)
library(jsonlite)
library(readxl)
library(curl)
# Get your API from GCP
API_KEY <- '&key=$YOUR_API_KEY'
URL <- 'https://maps.googleapis.com/maps/api/geocode/json?address='
# Data source:
# # Elementary high school: https://data.gov.tw/dataset/6087 
# # Junior high school: https://data.gov.tw/dataset/6088

mutolisp

 SELECT 
  time_bucket('1 day', timestamp) AS period, 
  summit, 
  direction, 
  avg(temperature), 
  max(temperature), 
  min(temperature) 
 FROM 
  temperature_hyper 
 WHERE

mutolisp

# install required packages
requiredPkgs <- c("BiocManager", "seqinr", "data.table", "devtools")
for ( i in 1:length(requiredPkgs)){
  if (!requireNamespace(requiredPkgs[i], quietly = TRUE))
    install.packages(requiredPkgs[i])
}
BiocManager::install("Biostrings")

devtools::install_github(repo = 'mutolisp/seqParser')

mutolisp