Project · 2025

OSRS RAG

Retrieval-augmented Q&A system over the Old School RuneScape wiki. Scraper, indexer, and query pipeline as separable scripts.

Role

Solo

Stack

Python, Embeddings, Vector search

Tags

rag, applied-llm, osrs

OSRS has one of the densest community knowledge bases on the internet, which makes it a good corpus to learn retrieval on.

The system is split into a handful of small scripts that each do one job:

• scraper.py pulls down the wiki.
• indexer.py chunks and embeds it.
• query.py is the front door for asking it questions.
• update.py re-syncs when the wiki changes.

Most of the learning was in the boring parts — chunk-size tuning, when reranking actually helps, and how badly an LLM can hallucinate when it confidently mistakes one quest for another. The biggest takeaway: most “good RAG” is chunking, not embeddings.

← All projects