Ramunas ramunasnognys

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.

Meta Prompting

In the Generative AI Age your ability to generate prompts is your ability to generate results.

Guide

Claude 3.5 Sonnet and o1 series models are recommended for meta prompting.

Replace {{user-input}} with your own input to generate prompts.

Use mp_*.txt as example user-inputs to see how to generate high quality prompts.

Setting Up MCP Servers on Windows

A step-by-step guide to setting up Model Context Protocol (MCP) servers for Claude Desktop on Windows.

Prerequisites

Install Node.js (v18.x or later)
- Download from: https://nodejs.org/
- Verify installation by opening Command Prompt (CMD) and running:
```
node --version
npm --version
```

Directory structure for JavaScript/Node Projects

While the following structure is not an absolute requirement or enforced by the tools, it is a recommendation based on what the JavaScript and in particular Node community at large have been following by convention.

Beyond a suggested structure, no tooling recommendations, or sub-module structure is outlined here.

Directories

lib/ is intended for code that can run as-is
src/ is intended for code that needs to be manipulated before it can be used

Debug Horizontal Scroll

(function (d) {
    var w = d.documentElement.offsetWidth,
        t = d.createTreeWalker(d.body, NodeFilter.SHOW_ELEMENT),
        b;
    while (t.nextNode()) {
        b = t.currentNode.getBoundingClientRect();
 if (b.right &gt; w || b.left &lt; 0) {

	import argparse
	import random
	import sys

	from transformers import AutoModelForCausalLM, AutoTokenizer, DynamicCache
	import torch

	parser = argparse.ArgumentParser()
	parser.add_argument("question", type=str)
	parser.add_argument(


	<TEMPLATE>
	<INSTRUCTIONS>
	Use the <CODEBASE> code as reference, and convert the high-level <TASK> into a set of very detailed step-by-step instructions that an AI coding agent can complete.
	Only includes steps an AI coding agent can take. Do not include testing or any other work a human would do to confirm the task has been completed.
	ALWAYS have the agent run a build when it is complete. Be specific and decisive about what the agent should do.
	Do not include any additional meta instructions to the user. Use markdown formatting.
	</INSTRUCTIONS>
	<TASK>

	# .cursorrules

	Components & Naming

	- Use functional components with `"use client"` if needed.
	- Name in PascalCase under `src/components/`.
	- Keep them small, typed with interfaces.
	- Use Tailwind for common UI components like textarea, button, etc. Never use radix or shadcn.

	Prisma

	<artifacts_info>
	The assistant can create and reference artifacts during conversations. Artifacts are for substantial, self-contained content that users might modify or reuse, displayed in a separate UI window for clarity.

	# Good artifacts are...
	- Substantial content (>15 lines)
	- Content that the user is likely to modify, iterate on, or take ownership of
	- Self-contained, complex content that can be understood on its own, without context from the conversation
	- Content intended for eventual use outside the conversation (e.g., reports, emails, presentations)
	- Content likely to be referenced or reused multiple times

	// Replace SmartObject’s Content and Save as JPG
	// 2017, use it at your own risk
	// Via @Circle B: https://graphicdesign.stackexchange.com/questions/92796/replacing-a-smart-object-in-bulk-with-photoshops-variable-data-or-scripts/93359
	// JPG code from here: https://forums.adobe.com/thread/737789

	#target photoshop
	if (app.documents.length > 0) {
	var myDocument = app.activeDocument;
	var theName = myDocument.name.match(/(.*)\.[^\.]+$/)[1];
	var thePath = myDocument.path;