alexlib

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.

alexlib

Below are the generic details and resources of the machine I used:

• Operating System: Linux Mint 22.3 - MATE 64-bit (safe-graphics / boot with 'nomodeset') - Linux Kernel: 6.14.0-37-generic
• GPU: not relevant / not used
• CPU: 2 cores – Intel® Core™2 Duo CPU E8400 @ 3.00GHz × 2
• RAM: 8 GB (DDR2)

Given my recent personal good experience in running inference with CPU-only (no GPU) on local models up to 2B-4B parameters with more than acceptable speed and performance on really limited hardware and very little RAM, I decided to share the instructions and commands I used to compile llama.cpp on my machine and the arguments I currently use to launch llama.

alexlib

FROM ubuntu:20.04 as builder

FROM python:3.10.7

# Create a user to run the application
RUN apt-get update && \
      apt-get -y install sudo

RUN useradd -ms /bin/bash docker && echo "docker:docker" | chpasswd && adduser docker sudo

alexlib

alexlib

alexlib

Cython example of exposing C-computed arrays in Python without data copies

The goal of this example is to show how an existing C codebase for numerical computing (here c_code.c) can be wrapped in Cython to be exposed in Python.

The meat of the example is that the data is allocated in C, but exposed in Python without a copy using the PyArray_SimpleNewFromData numpy

alexlib

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
% \embedvideo{<poster or text>}{<video file (MP4+H264)>}
% \embedvideo*{...}{...}                     % auto-play
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

\usepackage[bigfiles]{pdfbase}
\ExplSyntaxOn
\NewDocumentCommand\embedvideo{smm}{
  \group_begin:
  \leavevmode

alexlib

import napari
import numpy as np
import matplotlib.pyplot as pl
import h5py
from matplotlib.backends.backend_qt5agg import FigureCanvasQTAgg as FigureCanvas
from matplotlib.figure import Figure
#from qtpy.QtWidgets import QApplication, QMainWindow, QPushButton, QMenu, QVBoxLayout, QSizePolicy, QMessageBox, QWidget, QPushButton
import pyqtgraph as pg
from pyqtgraph.Qt import QtGui, QtCore
import sys

alexlib

import cv2
import numpy as np

def minMaxFilter(img, filterSize, minContrast):
    """
    After R.Adrian, J.Westerweel, "Particle image velocimetry", Cambridge
    university press, 2011. See ch.6.1.2, p.248-250.
    Parameters: img (cv2.imread) - image to be filtered
                filterSize (nd.array) - a 1x2 numpy array of the filter height
                                        and width correspondingly

alexlib

def customDeen(frame0,
               frame1,
               algorithmType = 'basic',
               cliplimit = 2.0,
               tileGridSize = (4, 4)
              ):
    """
    This is the custom implementation of the image processing alogrithm for PIV
    after Deen, Willems, Annaland, Kuipers, Lammertink, Kemperman, Wessling,
    Meer, "On image preprocessing for PIV of single and two phase flows over