Example Project

The plugin ships with an example project containing Blueprint examples for chat completion, streaming, multimodal vision, and server management.

Download

A Blueprint-only project is available for each supported Unreal Engine version. Please download the one that matches your engine. All links point to Google Drive.

Requirements

The GenAI Llama plugin must be installed in your project from the Fab store before opening these projects. For HTTP provider examples, you need a local inference server running (e.g., Ollama, LM Studio, or any OpenAI-compatible server).

Engine Version	Blueprint-Only Project
Unreal Engine 5.7	GenAILlamaExample_UE5.7_BP.zip
Unreal Engine 5.6	GenAILlamaExample_UE5.6_BP.zip
Unreal Engine 5.5	GenAILlamaExample_UE5.5_BP.zip
Unreal Engine 5.4	GenAILlamaExample_UE5.4_BP.zip
Unreal Engine 5.3	GenAILlamaExample_UE5.3_BP.zip
Unreal Engine 5.2	GenAILlamaExample_UE5.2_BP.zip
Unreal Engine 5.1	GenAILlamaExample_UE5.1_BP.zip

What’s Inside

The example project demonstrates all core features of the plugin:

Chat Completion (HTTP)

A Blueprint actor that sends a chat request to a local server (Ollama by default) and prints the AI response. Demonstrates configuring connection settings, model selection, and system prompts.

Streaming Chat (HTTP)

A streaming example that displays the AI response token-by-token in real-time, demonstrating the typewriter effect with the On Event delegate.

Multimodal Vision (HTTP)

An example using a vision model (e.g., llava) that sends a UTexture2D to the AI and asks it to describe the image. Shows both the ImagesAsTextures and Images (Base64) approaches.

Server Health Check

Demonstrates using the Check Server Health node to verify a local server is running before sending requests.

List Available Models

Shows how to query a server for all available models and display them in a dropdown.

Provider Switching

An example demonstrating switching between Ollama and OpenAI Compatible providers at runtime with different base URLs.

Quick Start

Install a local server (e.g., Ollama) if using HTTP providers.
Pull a model:
```
ollama pull llama3
```
Open the example project in Unreal Engine.
Press Play — the example actors will send requests and display responses.

For the Embedded example specifically

Follow the Embedded Inference Setup Guide to drop the pinned b8802 shared libraries into Plugins/GenAILlama/ThirdParty/LlamaCpp/lib/<Platform>/.
Download a tiny GGUF (e.g. Qwen2.5-0.5B-Instruct Q2_K, ~280 MB) into the example project’s Content/Models/ folder.
On the Load Embedded Model node, set GPU Layers = 0 unless you have a CUDA/Vulkan/Metal backend DLL in place. Non-zero values on CPU-only setups will fail the load with a backend error.

Tip: Check the Output Log (Window > Developer Tools > Output Log) to see full request and response details. Set LogGenAILlama=Verbose for detailed streaming output. Enable Project Settings > Plugins > GenAI Llama > Debug > Enable Extended Logging to see llama.cpp’s own startup diagnostics and model-load output under [llama.cpp].