Martin Czygan miku

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.

title

AI Agents - A Comprehensive Database of The Best AI Agents

AI Agents: A Comprehensive Database of The Best AI Agents

Primo Search

Browser automation tool to search FU Berlin's Primo library catalog and capture results.

Setup

uv venv .venv
uv pip install playwright --python .venv/bin/python
.venv/bin/playwright install chromium

Which GGUF is right for me? (Opinionated)

Good question! I am collecting human data on how quantization affects outputs. See here for more information: ggml-org/llama.cpp#5962

In the meantime, use the largest that fully fits in your GPU. If you can comfortably fit Q4_K_S, try using a model with more parameters.

llama.cpp feature matrix

See the wiki upstream: https://github.com/ggerganov/llama.cpp/wiki/Feature-matrix

Universal Workflow for Approaching a Machine Learning Problem

Define the Problem and Assemble a Dataset :

What is your input data?
What are you trying to predict?
What type of problem is it - Supervised? Unsupervised? Self-Supervised? Reinforcement Learning?
Be aware of the hypotheses that you are making at this stage:

	# th30z@u1310:[Desktop]$ psql -h localhost -p 55432
	# Password:
	# psql (9.1.10, server 0.0.0)
	# WARNING: psql version 9.1, server version 0.0.
	# Some psql features might not work.
	# Type "help" for help.
	#
	# th30z=> select foo;
	# a \| b
	# ---+---

	#!/bin/bash
	set -eu -o pipefail

	token="${GITHUB_TOKEN:?Set GITHUB_TOKEN before running this script}"

	clone_or_pull() {
	local page="$1"
	local tmpfile
	tmpfile=$(mktemp)
	trap "rm -f '$tmpfile'" RETURN

	// radioscript allows to capture redio stream (this is a mostly complete script, taken out of an not yet published project)
	package main

	import (
	"bufio"
	"crypto/sha1"
	"errors"
	"flag"
	"fmt"
	"io"

	#EXTM3U
	#EXTINF:-1,ARD
	https://daserste-live.ard-mcdn.de/daserste/live/hls/de/master.m3u8
	#EXTINF:-1,ARD ONE
	https://mcdn-one.ard.de/ardone/hls/master.m3u8
	#EXTINF:-1,ARD Alpha
	https://mcdn.br.de/br/fs/ard_alpha/hls/de/master.m3u8
	#EXTINF:-1,ARD Tagesschau
	https://tagesschau.akamaized.net/hls/live/2020115/tagesschau/tagesschau_1/master.m3u8
	#EXTINF:-1,ZDF

	""" To use: install Ollama, clone OpenVoice, run this script in the OpenVoice directory
	brew install portaudio
	brew install git-lfs
	git lfs install

	git clone https://github.com/myshell-ai/OpenVoice
	cd OpenVoice
	git clone https://huggingface.co/myshell-ai/OpenVoice
	cp -r OpenVoice/* .