Llama3 Assistant MVP

Early MVP of a local AI assistant powered by Meta Llama 3 (8B) with persistent conversation memory.

The project explores how to build a lightweight assistant that runs a local LLM, maintains conversation context, and organizes the system into modular components such as the LLM interface, memory management, and application loop.

This repository focuses on experimenting with local LLM interaction, simple memory persistence, and assistant architecture.

Features

• Local Meta Llama 3 8B model support
• Persistent memory storage
• Modular architecture
• JSON-based memory system

Project Structure

mvp/
│
├── main.py          # Entry point
├── llm.py           # LLM interaction layer
├── memory.py        # Memory management
├── memory.json      # Stored memory
├── schemas.py       # Data schemas
│
├── models/
│   └── meta-llama-3-8B/
│
├── requirements.txt
└── README.md

Installation

Clone the repository:

git clone https://github.com/yourname/llama3-assistant-mvp.git
cd llama3-assistant-mvp

Create virtual environment:

python -m venv .venv
source .venv/bin/activate

Install dependencies:

pip install -r requirements.txt

Run

python main.py

Notes

This repository represents an early MVP prototype and is still under development.

Future improvements may include:

• improved memory system
• conversation context management
• API interface
• better model integration

License

MIT

Contact

Gmail: maria.ilnitska11@gmail.com

LinkedIn: www.linkedin.com/in/maria-ilnitska

Telegram: @mariailnitska