Terraria Assistant (RAG + Phi-3-mini)

A question-answering assistant for the game Terraria. It answers questions about crafting, items, classes, and bosses, grounding every answer in the Terraria Wiki using Retrieval-Augmented Generation (RAG).

How it works

  1. Your question is matched against chunks of the Terraria Wiki using semantic search (sentence-transformers + FAISS).
  2. The most relevant wiki text is passed to Phi-3-mini, which answers based only on that context. This reduces hallucination.

The repo also includes optional LoRA adapters fine-tuned on Terraria Q&A pairs (in adapters/), for use with Apple's MLX on Apple Silicon.

Requirements

  • Python 3.11
  • Apple Silicon Mac (for MLX)

Setup

pip install -r requirements.txt

Usage

python rag_chat.py "How do I craft a Molten Pickaxe?"

Example output: You need 20 Hellstone Bars, crafted at an Iron Anvil or Lead Anvil.

Strengths and limits

  • Strong: crafting recipes, item stats, boss mechanics.
  • Weaker: specific gear recommendations may occasionally be inaccurate.
  • Only answers Terraria questions; off-topic questions are declined.
  • Not an official source; verify important details on the wiki.

Data source

Data collected from Terraria Wiki (https://terraria.wiki.gg), licensed under CC BY-NC-SA. Non-commercial use only. Fan-made, not affiliated with Re-Logic.

Downloads last month

-

Downloads are not tracked for this model. How to track
MLX
Hardware compatibility
Log In to add your hardware

Quantized

Inference Providers NEW
This model isn't deployed by any Inference Provider. ๐Ÿ™‹ Ask for provider support

Model tree for imperialdiva/terraria-assistant

Adapter
(862)
this model