I built a local AI assistant fine-tuned on my own conversations
I wanted an AI assistant that is actually mine: it runs on my machine, knows how I talk, and can use my files and shell. So I built JERN, a small open-source assistant around a fine-tuned 1.5B model, and this is how it works and what broke. Why I built it Most assistants are someone else's product: they run on someone else's servers, know nothing about my files, and sound the same for everyone. I…
I constructed a personalized AI assistant called JERN, which runs on my own computer and is customized to understand how I communicate. This assistant, called JERN, uses a small open-source model that I fine-tuned with my past conversations, allowing it to answer in my voice and access my files and computer commands. JERN's construction involves a Chat model, a GPT-style transformer model I developed from scratch, and an app consisting of a FastAPI server, a simple web interface, and plug-in tools for file management, shell commands, web search, downloads, and system information.
To get JERN up and running, users execute the command python main.py, which launches the web interface at http://127.0.0.1:8000. The model's performance is limited by its size, sometimes misusing tools, duplicating information, or providing inaccurate details. However, JERN's tools are local, enabling access to actual files and commands on the user's machine, although this also poses potential risks.
For those with limited resources, JERN offers a Colab mode, running the model on Google Colab's GPU and linking it to the local web interface via an SSH tunnel. Despite its experimental nature, JERN has shown promise, and users are encouraged to try it by cloning the GitHub repository, setting up a virtual environment, installing necessary packages, and launching the assistant.
Written by urgent.news from Dev.to's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
