Erik Nomitch eriknomitch

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.

aider-commit

A simple wrapper for using aider to create commits with automatically generated messages using AI. Includes a few extra checks and outputs.

Installation

Set up aider
Add the function to your ~/.zshrc or ~/.bashrc.

Usage

ChatGPT Resources

Context

ChatGPT appeared like an explosion on all my social media timelines in early December 2022. While I keep up with machine learning as an industry, I wasn't focused so much on this particular corner, and all the screenshots seemed like they came out of nowhere. What was this model? How did the chat prompting work? What was the context of OpenAI doing this work and collecting my prompts for training data?

I decided to do a quick investigation. Here's all the information I've found so far. I'm aggregating and synthesizing it as I go, so it's currently changing pretty frequently.

	import pyaudio
	import audioop

	p = pyaudio.PyAudio()

	stream = p.open(format=pyaudio.paInt16,
	channels=1,
	rate=44100,
	input=True,
	frames_per_buffer=1024)

	"""
	Toy demonstration of chain-of-thought and consensus prompting using OpenAI API.

	© Riley Goodside 2022
	"""

	import os
	import re
	from statistics import mode

	# Script for converting a HF Diffusers saved pipeline to a Stable Diffusion checkpoint.
	# Only converts the UNet, VAE, and Text Encoder.
	# Does not convert optimizer state or any other thing.
	# Written by jachiam

	import argparse
	import os.path as osp

	import torch

	#!/bin/zsh

	# Add the following to your ~/.gitconfig
	#
	# [alias]
	# up = !zsh /path-to-this-script/git-up.sh
	#
	# Then you can call
	#
	# `git up`

	import { useState, useEffect, useCallback } from 'react';

	export const useVisbilityState = () => {
	const [visibilityState, setVisibilityState] = useState(null);

	const handleVisbilityChange = useCallback(() => {
	setVisibilityState(document.visibilityState);
	}, [setVisibilityState]);

	useEffect(() => {

	#!/usr/bin/env python3
	# Based on code from these stackoverflow answers:
	# https://askubuntu.com/questions/60837/record-a-programs-output-with-pulseaudio/910879#910879
	import re
	import subprocess
	import sys
	import os
	import signal
	from time import sleep

	import six
	from peewee import Model, FieldAccessor
	from playhouse.postgres_ext import BinaryJSONField


	class GForeignKeyAccessor(FieldAccessor):
	def get_rel_instance(self, instance):
	value = instance.__data__.get(self.name)
	model_name = value['model']
	rel_model = self.field.allowed_types[model_name]