Documentation

Board Game Librarian Wiki

How the system works, what it can and can't do, and how to integrate it. Pick a section below.

How It Works

From rulebook to answer — the full pipeline explained.

How Board Game Librarian Answers Your Questions

A plain-language guide to how the system understands questions, searches rulebooks, handles any language, and decides how confident it is in an answer.

Read →

From PDF to answer: how rulebooks become knowledge

When you ask "Can I interrupt an attack with a reaction?" the system doesn't search a PDF. It searches a compressed, enriched, vectorised representation of that PDF — one that was built the first time

Read →

Ten languages, one answer

Board Game Librarian answers rules questions in the language you ask them in. No language selector, no profile setting, no dropdown. Type in Polish, get Polish back. Type in Japanese, Japanese comes o

Read →

The Question & Answer Pipeline

How questions flow through the two-tier pipeline -- game detection, vector search, and when I escalate to Tier 2.

Read →

How I Read Rulebooks

When you ask about a game for the first time, I process the rulebook on demand -- chunking, embedding, and storing it in the vector database.

Read →

How I Generate Answers

How I turn raw rulebook chunks into a cited answer -- prompt templates, AI synthesis, and the citation system.

Read →

Technical Deep Dives

Semantic search, prompt architecture, and the limits of what the system can read.

How semantic search works: finding meaning, not keywords

You're mid-game and you need to know what happens when two units occupy the same hex. You type "same hex two units" into a search box. A keyword engine looks for those exact words. If the rulebook say

Read →

Inside the prompt architecture

Answering a board game rules question isn't a single model call. It's a pipeline: classify, retrieve, route, synthesise. Each stage has a specific job, and they're designed to be independent — swap an

Read →

What the system can't see: limits of PDF extraction

The system works by reading text. Rulebooks communicate through text, images, diagrams, colour, spatial layout, and visual examples. That gap between what the system reads and what a rulebook actually

Read →

System Architecture

All 20 services, how they connect, and the full technology stack behind Board Game Librarian.

Read →

Vector Search & Embeddings

How HNSW indexing, 768-dimensional embeddings, and cosine similarity combine to find the right rulebook passages.

Read →

The RAG Pipeline in Detail

RAG explained step by step: query expansion, vector retrieval, context assembly, and AI synthesis.

Read →

The Two-Tier Answer System

How auto-escalation works, what triggers Tier 2, and why most questions resolve in under 7 seconds.

Read →

10 Languages, One Pipeline

How I detect the question language and answer in the same language, even though rulebooks and forum threads are in English.

Read →

Autonomous Quality Optimisation

How the autonomous quality loop works: nightly LLM-as-judge evaluation, YAML prompt proposals, test battery validation, and safe auto-deployment of improvements.

Read →

For Publishers

Partner portal, widget integration, and embedding the assistant on your site.

The Project

Where it started, how it grew, and where it's going.

Other Pages