Mikalai Karelin karelin

LLM Wiki v2

A pattern for building personal knowledge bases using LLMs. Extended with lessons from building agentmemory 10K Stars ⭐️, 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.

Currently, Working on AKBP: Agent Knowledge Base Protocol based on my findings, a protocol for creating, updating, retrieving, and sharing durable knowledge across AI agents.

What the original gets right

name	wiki
description	Compile personal data (journals, notes, messages, whatever) into a personal knowledge wiki. Ingest any data format, absorb entries into wiki articles, query, cleanup, and expand.
argument-hint	ingest \| absorb [date-range] \| query <question> \| cleanup \| breakdown \| status

Personal Knowledge Wiki

You are a writer compiling a personal knowledge wiki from someone's personal data. Not a filing clerk. A writer. Your job is to read entries, understand what they mean, and write articles that capture understanding. The wiki is a map of a mind.

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.

CLAUDE.md

This file provides guidance to Claude Code (claude.ai/code) when working with code in this repository.

What This Is

Technical and architectural audit of a system. This is an evolving documentation vault, not a codebase.

Language

Anti-hype LLM reading list

Goals: Add links that are reasonable and good explanations of how stuff works. No hype and no vendor content if possible. Practical first-hand accounts of models in prod eagerly sought.

Foundational Concepts

	0. Install "Obsidian Advanced URI" plugin and (optionally) "Local Images Plus" to save images from clipped pages locally
	1. Create bookmark with from any url
	2. Edit address to:
	```
	javascript:(function()%7B%2F%20Optional%20vault%20name%20%2F%0Aconst%20vault%20%3D%20%22%22%3B%0A%0A%2F%20Optional%20folder%20name%20such%20as%20%22Clippings%2F%22%20%2F%0Aconst%20folder%20%3D%20%22%2B%2F%D0%A1%D0%BE%D1%85%D1%80%D0%B0%D0%BD%D0%B5%D0%BD%D0%BD%D0%BE%D0%B5%2F%22%3B%0A%0A%2F%20Optional%20tags%20%20%2F%0Alet%20tags%20%3D%20%22clippings%22%3B%0A%0A%0A%0A%2F%0A%20%0A%20%20Embed%20turndown%406.0.0%20to%20work%20around%20content%20security%20policies%20breaking%20import%20statements%0A%20%0A%20*%2F%0A%0Afunction%20extend%20(destination)%20%7B%0A%20%20for%20(var%20i%20%3D%201%3B%20i%20%3C%20arguments.length%3B%20i%2B%2B)%20%7B%0A%20%20%20%20var%20source%20%3D%20arguments%5Bi%5D%3B%0A%20%20%20%20for%20(var%20key%20in%20source)%20%7B%0A%20%20%20%20%20%20if%20(source.hasOwnProperty(key))%20destination%5Bkey%5D%20%3D

	"""Module for handling authentication, interactions with Firebase and JWT cookies.

	This solution is refactored from the ‘streamlit_authenticator’ package . It leverages JSON
	Web Token (JWT) cookies to maintain the user’s login state across browser sessions. For the
	backend, It uses Google’s Firebase Admin Python SDK. This solution ensures that the content
	of the page and user settings panel are only displayed if the user is authenticated. Similarly,
	the login page can only be accessed if the user is not authenticated. Upon registration, the
	user is sent a verification link to their e-mail address.
	Important - to make this app run, put the following variables in your secrets.toml file:
	COOKIE_KEY - a random string key for your passwordless reauthentication

	import streamlit as st
	import spacy
	from spacy import displacy
	import pandas as pd

	from scispacy.umls_linking import UmlsEntityLinker
	from scispacy.abbreviation import AbbreviationDetector


	SPACY_MODEL_NAMES = ["en_core_sci_sm", "en_core_sci_md", "en_core_sci_lg"]

	// Basic Routes libraries
	const express = require('express');
	const app = express();
	const bodyParser = require('body-parser');
	const cors = require('cors');
	const axios = require('axios');

	//Logging
	const pino = require('pino')("./logs/info.log");;
	const expressPino = require("express-pino-logger")({