Pike yxlwfds

LLM Wiki v2

A pattern for building personal knowledge bases using LLMs. Extended with lessons from building agentmemory, a persistent memory engine for AI coding agents.

This builds on Andrej Karpathy's original LLM Wiki idea file. Everything in the original still applies. This document adds what we learned running the pattern in production: what breaks at scale, what's missing, and what separates a wiki that stays useful from one that rots.

What the original gets right

The core insight is correct: stop re-deriving, start compiling. RAG retrieves and forgets. A wiki accumulates and compounds. The three-layer architecture (raw sources, wiki, schema) works. The operations (ingest, query, lint) cover the basics. If you haven't read the original, start there.

	"""
	The most atomic way to train and run inference for a GPT in pure, dependency-free Python.
	This file is the complete algorithm.
	Everything else is just efficiency.

	@karpathy
	"""

	import os # os.path.exists
	import math # math.log, math.exp

	from datasets import load_dataset
	from sentence_transformers import (
	SentenceTransformerTrainer,
	SentenceTransformerTrainingArguments,
	)

	from pylate import losses, models, utils

	def main():
	# As ReasonIR do not re-upload the BRIGHT data, we need to load it from the original source

	import queue
	import random
	from functools import partial
	from time import sleep, perf_counter

	import trio

	CONCURRENCY_LIMIT = 8
	limiter = trio.CapacityLimiter(CONCURRENCY_LIMIT)

	from datetime import timedelta, datetime
	from typing import Type, Optional, Dict, Any, Tuple
	from uuid import uuid4
	from abc import ABC, abstractmethod

	from fastapi import FastAPI, Request, Depends, HTTPException, Response
	from fastapi.security.api_key import APIKeyBase, APIKey, APIKeyIn
	from base64 import b64encode, b64decode
	from itsdangerous import TimestampSigner
	from itsdangerous.exc import BadTimeSignature, SignatureExpired

	import inspect
	import random
	import time
	from typing import Dict, List, Tuple

	from rich.console import Console, RenderGroup
	from rich.live import Live
	from rich.panel import Panel
	from rich.progress import Progress
	from rich.syntax import Syntax

	from contextlib import contextmanager
	import threading
	from thread import get_ident
	from sqlalchemy import *
	from sqlalchemy.orm import *

	from sqlalchemy.ext.declarative import declarative_base

	from flask import Flask
	from flask_sqlalchemy import SQLAlchemy

	#!/bin/bash

	# file descriptor
	# Change 400000 to increase or decrease number of file descriptor

	echo "* soft nofile 400000" >> /etc/security/limits.conf
	echo "* hard nofile 400000" >> /etc/security/limits.conf

	# Changing kernal parameters (modify value if required)

	app.directive('datepickerLocaldate', ['$parse', function ($parse) {
	var directive = {
	restrict: 'A',
	require: ['ngModel'],
	link: link
	};
	return directive;

	function link(scope, element, attr, ctrls) {
	var ngModelController = ctrls[0];

	license: mit
	border: yes
	height: 640