gsamaras

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.

gsamaras

There are many Git workflows out there, I heavily suggest also reading the atlassian.com [Git Workflow][article] article as there is more detail then presented here.

The two prevailing workflows are [Gitflow][gitflow] and [feature branches][feature]. IMHO, being more of a subscriber to continuous integration, I feel that the feature branch workflow is better suited.

When using Bash in the command line, it leaves a bit to be desired when it comes to awareness of state. I would suggest following these instructions on [setting up GIT Bash autocompletion][git-auto].

Basic branching

When working with a centralized workflow the concepts are simple, master represented the official history and is always deployable. With each now scope of work, aka feature, the developer is to create a new branch. For clarity, make sure to use descriptive names like transaction-fail-message or github-oauth for your branches.