Show HN: IA-SQL – Postgres compiles your documents into a wiki with an LLM
IA-SQL is a new tool that transforms PostgreSQL into a self-compiling knowledge base. It utilizes an external LLM to convert raw documents into a cross-referenced Markdown wiki while maintaining consistency through auditing. The system is designed to improve knowledge accumulation over time by processing documents asynchronously.
- ▪IA-SQL implements Andrej Karpathy's 'LLM Wiki' pattern within PostgreSQL.
- ▪The tool compiles documents into a Markdown wiki and audits for inconsistencies.
- ▪It operates asynchronously, allowing normal queries to continue while processing new information.
Opening excerpt (first ~120 words) tap to expand
IA-SQL Turn PostgreSQL into a self-compiling knowledge base. An in-database implementation of Andrej Karpathy's "LLM Wiki" pattern: you INSERT raw documents, and a background worker uses an external LLM to compile them into a maintained, cross-referenced Markdown wiki — then keeps auditing that wiki against the sources for hallucinations. Status: 0.1 — working proof of concept. Built and tested on PostgreSQL 17. Español aquí → 🌐 Live demo: https://iasql.dev.feres.cl 📖 New here? Start with the tutorial by user profile. The idea Most "chat with your documents" systems use RAG: at query time they retrieve a few text chunks and ask the model to improvise an answer. The model re-discovers your domain from scratch on every question and never accumulates understanding.
…
Excerpt limited to ~120 words for fair-use compliance. The full article is at GitHub.