Godot Llama 1.0.0 Tools 4.5 Community

Submitted by user oceancucumber; MIT; 2026-02-22

Godot Llama brings llama.cpp into Godot 4.x as a native GDExtension so you can run local GGUF models directly in your game or tool.

Included API classes:

LlamaModel for model loading, tokenization/detokenization, vocab, and metadata
LlamaContext for context creation, generation, streaming signals, cancellation, and perf stats
LlamaSampler for configurable sampling behavior
LlamaAsyncWorker for non-blocking generation workflows
Features:

Synchronous and token-streaming text generation
Common generation controls (temperature, top_k, top_p, min_p, penalties, stop sequences, seed, max tokens)
Context state helpers (save/load state to memory or file, clear KV cache)
Demo scene included at demo.tscn
Requirements:

Godot 4.5
Local GGUF model files
Platform binaries in addons/godot_llama/bin/ (provided by release artifacts)
Quick start:

Copy addons/godot_llama/ into your Godot project.
Open res://addons/godot_llama/demo/demo.tscn.
Select a .gguf model, create context, and generate.