Sylwester Lachiewicz slachiewicz

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.

Why did you do this?

Relax, I only have one Sunday to work on idea, literally my weekend project. So I tried Deepseek to see if it can help. Surprisingly, it works and it saves me another weekend...

What is your setup?

Just chat.deepseek.com (cost = free) with prompts adapted from this gist.

Does it work in one-shot or I have to prompt it multiple times?

Script

The below rules and steps are mostly automated by simply running the Maven4Migration.java jbang script in the plugin directory.

Rules

The rules are quite simple

the plugin can only depend on org.apache.maven:maven-api-* artifacts (no any other maven related artifacts) at compile time
avoid plexus artifacts at compile time
mojo annotations are now in org.apache.maven.api.plugin.annotations package

This page has been moved to https://github.com/int128/hello-envoy

Envoy TCP Proxy

This is an example of Envoy TCP Proxy from localhost:10000 to www.google.com:80. See also https://www.envoyproxy.io/docs/envoy/latest/intro/arch_overview/listeners/tcp_proxy.

To run a proxy:

Overview

Internet connection and DNS routing are broken from WSL2 instances, when some VPNs are active.

The root cause seems to be that WSL2 and the VPN use the same IP address block, and the VPN routing clobbers WSL2's network routing.

This problem is tracked in multiple microsoft/WSL issues including, but not limited to:

	///usr/bin/env jbang "$0" "$@" ; exit $?
	//DEPS com.github.javaparser:javaparser-core:3.25.7
	//DEPS org.apache.commons:commons-lang3:3.14.0
	//DEPS info.picocli:picocli:4.7.5
	//DEPS org.dom4j:dom4j:2.1.4
	//JAVA 17

	import com.github.javaparser.JavaParser;
	import com.github.javaparser.ast.CompilationUnit;
	import com.github.javaparser.ast.ImportDeclaration;

	from uuid import uuid4
	from argparse import ArgumentParser
	from datetime import datetime, timedelta

	# avro stuff
	from avro.datafile import DataFileWriter
	from avro.io import DatumWriter, BinaryEncoder

	# kafka stuff
	from confluent_kafka import Producer, Consumer

	# Clone llama.cpp
	git clone https://github.com/ggerganov/llama.cpp.git
	cd llama.cpp

	# Build it
	make clean
	LLAMA_METAL=1 make

	# Download model
	export MODEL=llama-2-13b-chat.ggmlv3.q4_0.bin

	<!DOCTYPE NETSCAPE-Bookmark-file-1>
	<!-- This is an automatically generated file.
	It will be read and overwritten.
	DO NOT EDIT! -->
	<META HTTP-EQUIV="Content-Type" CONTENT="text/html; charset=UTF-8">
	<TITLE>Bookmarks</TITLE>
	<H1>Bookmarks</H1>
	<DL><p>
	<DT><H3 ADD_DATE="1604897638" LAST_MODIFIED="0" PERSONAL_TOOLBAR_FOLDER="true">Bookmarks bar</H3>
	<DL><p>