Over the past few weeks, I’ve been building LibreAI — a simple, privacy-focused AI chat app that runs entirely on your own infrastructure using open-source models like Mistral, LLaMA 3, and Phi via Ollama.

LibreAI is built with privacy and simplicity at its core. It streams responses in real-time, stores nothing, and has zero telemetry. No OpenAI, no account wall, no JavaScript bloat.

✅ Try it live

💬 Discuss it on Hacker News

🧠 Why I Built LibreAI

A lot of modern AI tools work well, but they come with trade-offs: vendor lock-in, data collection, or heavy dependencies.

I wanted an AI assistant I could trust and control, powered by models that run locally — without relying on big cloud providers or complex frontend stacks.

So I built LibreAI as a lightweight, clean alternative.

What I focused on:

A UI that loads fast, even on slow networks

Real-time streaming output

Support for multiple open models via Ollama

Full self-hostability — no cloud APIs required

No React, no tracking, no unnecessary complexity

⚙️ The Stack

LibreAI is built using:

Go Fiber – backend web framework

HTMX – frontend interactivity without JS frameworks

TailwindCSS – utility-first styling

Ollama – for serving local LLMs

Plausible – privacy-first analytics (no cookies or tracking)

🧪 What Works (and What’s Coming)

Right now, LibreAI supports:

Real-time streamed responses

Multiple open models via Ollama

Fast performance with a lightweight UI

What I’m improving:

Mobile layout and responsiveness

Model selection UX

Easier deployment options

LibreAI is intentionally minimal.

💬 I'd Love Your Feedback

If you're working on AI tools, local model deployments, or care about privacy-first UX, I’d love to hear your thoughts.

What models are you running locally?
What do you look for in a self-hosted AI tool?
Does lightweight and private beat "smart and centralized" for you?

👉 Try Libre

🗣 Join the Hacker News discussion

Thanks for reading 🙌