MemoriaForge: Local AI in C++ Without the Complexity

A lightweight llama.cpp wrapper for building local AI applications in just a few lines of code.

Building Local AI Applications in C++ Has Never Been Easier

Language models are no longer exclusive to Python. Thanks to llama.cpp, it is possible to run GGUF models locally on virtually any computer. However, integrating llama.cpp directly into a project usually involves handling context, conversation history, state persistence, and generation parameters.

MemoriaForge removes that complexity.

Its goal is to provide a simple API so you can focus on developing your application rather than dealing with the internal details of inference.

What Is MemoriaForge?

MemoriaForge is a C++ library built on top of llama.cpp that allows you to integrate GGUF models through a simple and modern interface.

With just a few lines of code, you can:

• Load GGUF models
• Maintain conversations with memory
• Inject custom context
• Save and restore complete sessions
• Configure generation parameters
• Get responses from local models

All without interacting directly with the low-level llama.cpp API.

Installation

The easiest way is to use the precompiled libraries available in the Releases section.

Download the package corresponding to your operating system and add:

libMemoriaForge.so

or on Windows:

MemoriaForge.dll
libMemoriaForge.a

to your project.

Then link the library as you would any other C++ dependency.

Your First Program

#include <iostream>
#include "MemoriaForge.h"

int main() {
    MemoriaForge::LLMSession llm("models/Qwen3-0.6B-Q4_K_M.gguf");

    std::cout << llm.chat("What is a neural network?") << std::endl;

    return 0;
}

That's it.

You now have an application capable of interacting with a language model running locally.

If you want to compile it from the command line:

c++ main.cpp -lMemoriaForge -o ChatSimple

Persistent Memory

One of MemoriaForge's main features is the ability to save and restore the complete state of a conversation.

Save:

llm.save_state("chat.bin");

Restore:

llm.load_state("chat.bin");

This allows you to continue a conversation even after closing the application.

It also eliminates the need to manually rebuild the context on every execution.

Injecting Knowledge

You can add additional information to the model's context before generating responses.

llm.context_injector("You are an assistant specialized in Linux.");

or load information from files:

llm.context_injector(document);

Ideal for:

• Personal assistants
• Documentation systems
• Educational tools
• Enterprise software
• Video game NPCs

Configuring Generation

Sampling parameters can be modified dynamically.

llm.set_sampling_params(
    0.2f,  // min_p
    0.8f,  // temperature
    42     // seed
);

This allows you to adjust the balance between creativity, accuracy, and reproducibility.

Why Use MemoriaForge?

Because it allows you to leverage the full power of llama.cpp without working directly with its API.

Less Code

MemoriaForge::LLMSession llm(model);
std::string response = llm.chat("Hello");

Built-In Persistence

Save and restore complete conversations with a single function call.

Simple API

Designed for C++ developers who want fast results.

Cross-Platform

Compatible with Linux and Windows.

Use Cases

• Local chatbots
• Desktop assistants
• Intelligent NPCs
• Documentation systems
• Educational tools
• Enterprise applications
• Offline AI software

Open Source Project

MemoriaForge is free software released under the GPL-3.0 license.

Source code, documentation, and binaries are available on GitHub:

https://github.com/canuconde/MemoriaForge

Conclusion

If you want to add local AI to a C++ application, you don't need to deal directly with the complexity of llama.cpp.

MemoriaForge provides a simple API, context persistence, and ready-to-use libraries so you can start developing in minutes.

Less time configuring models. More time building your product.

Support MemoriaForge

MemoriaForge is developed and maintained in my free time.

If this library saved you time, helped your project, or you simply want to support independent open-source software, consider making a small donation.

Every contribution helps improve the project and motivates future development.

Thank you for your support!

☕ Donate