llm-box
Run LLMs locally with a clean web UI
Overview
Automatic License Plate Recognition + gate control on Raspberry Pi — captures license plates via camera and triggers a GPIO relay to open a gate.
Details
This project demonstrates how to containerize and serve a Large Language Model (LLM) using FastAPI, Docker, and pre-trained transformer models, with a lightweight browser-based chat interface.
Built as part of the Introduction to Artificial Intelligence course at Southeast European University.
Serves a pre-trained transformer model (microsoft/phi-1_5) for text generation.
FastAPI-based backend with /chat endpoint.
Minimal browser UI for real-time chatting with the LLM.
Fully containerized with Docker (multi-stage build).
Benchmarking for latency, memory usage, and container performance.
Early CI/CD groundwork laid for automated builds.
