Ditch the Subscription, Self-Host Your AI Companion

Let's be real: the AI companion space is exploding, but so are the monthly subscription fees. What if you could have a capable, conversational AI sidekick without the recurring bill and with full control over your data? That's exactly the itch that Airi, a new open-source project, aims to scratch.

Instead of paying a platform, you run it yourself. It’s a straightforward proposition for developers who value privacy, customization, and just the simple satisfaction of self-hosting a cool piece of tech.

What It Does

Airi is a self-hosted, open-source web application that provides an interface for interacting with Large Language Models (LLMs) in a conversational, "companion"-style format. You connect it to your own AI backend (like an OpenAI-compatible API or a local model running via Ollama), and it gives you a clean, chat-focused UI to talk to it. It's essentially your own private portal to an AI, without the middleman.

Why It's Cool

The cool factor here isn't about a dizzying array of untested features—it's about intentionality and control.

• Privacy-First: All your conversations stay on your infrastructure. There's no external company logging your chats for "training purposes" or anything else.
• Model Agnostic: You're not locked into one AI. Point it at GPT-4 via the OpenAI API, use a local Llama or Mistral model through Ollama, or connect it to any other service that offers a compatible endpoint. Swap models as you please.
• The Self-Hosted Satisfaction: There's a unique developer joy in getting a service running on your own hardware or VPS. It's yours to modify, break, fix, and scale as you see fit.
• Cost Control: After the initial setup, your only cost is the compute for the model itself (which can be $0 if you're running small models locally on your own machine). No surprise monthly subscriptions.

How to Try It

Getting started with Airi is a classic Docker-based setup, which keeps things nicely contained.

1. Clone the repo:

   git clone https://github.com/moeru-ai/airi.git
   cd airi
   

2. Configure it: You'll need to set up your environment variables, primarily to point it to your LLM backend (like an OPENAI_API_BASE and OPENAI_API_KEY for compatible services).
3. Run it: Use Docker Compose to spin it up.

   docker-compose up -d
   

4. Open your browser: Navigate to http://localhost:3000 (or your server's IP) and start chatting.

For detailed configuration options, especially for different model backends, check the project's README.md on GitHub.

Final Thoughts

Airi won't have all the polished features of a massive commercial product, and that's not the point. It's a focused, developer-friendly tool that solves one problem well: giving you a private chat interface for the AI models you choose to use.

It's perfect for tinkerers, privacy-conscious users, or anyone who wants to decouple their AI interactions from a specific paid service. If you've already been running local models via Ollama or using OpenAI's API directly, Airi provides a much more pleasant way to interface with them daily. It’s a solid weekend project to set up and a satisfying way to take back a little control from the SaaS ecosystem.

Follow us for more cool projects: @githubprojects