Guillaume Desforges GuillaumeDesforges

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.

marp	true
theme	default

Marp diagram plugin example

Powered by kroki.io

	To whom it may concern,

	I wanted to follow up on my previous emails.

	My understanding after consulting with others is that what we do in the nixpkgs derivation for CUDA does not preclude binary caching & redistribution, as we only modify the library metadata such as the dynamic section (i.e. DT_RUNPATH, setting interpreter, setting RPATH). As I understand, object code refers to machine code that is executed by the processor, and thus my understanding is that we leave the object code untouched. I previously shared these post-distribution patches on 2/3/2020 for your review.

	Furthermore, my understanding under Section 2.3 is that we are ok to redistribute the SDK in full as long as this redistribution only happens under Linux.

	Thus, we plan on proceeding with setting up a binary cache for distributing CUDA and packages requiring CUDA using nixpkgs.

	-- show running queries (pre 9.2)
	SELECT procpid, age(clock_timestamp(), query_start), usename, current_query
	FROM pg_stat_activity
	WHERE current_query != '<IDLE>' AND current_query NOT ILIKE '%pg_stat_activity%'
	ORDER BY query_start desc;

	-- show running queries (9.2)
	SELECT pid, age(clock_timestamp(), query_start), usename, query
	FROM pg_stat_activity
	WHERE query != '<IDLE>' AND query NOT ILIKE '%pg_stat_activity%'