We’ve all been there. You want to use an AI model, but you are tired of official interfaces that collect your telemetry and track your data. So, you look into open-source alternatives like Open WebUI. But guess what? You suddenly need Docker, a heavy Python/Node.js backend, hundreds of megabytes of dependencies, and a master's degree in DevOps just to chat with an API.
As a developer, this frustrated me. Why do we need heavy backends just to forward a JSON request to an OpenAI-compatible API?
So, I decided to rebel against standard web-app overengineering. I built Traliran AI Hub — a pure local-first playground that requires zero installation, zero backend, and zero telemetry.
- Live Demo: traliran.github.io/traliran-ai-hub
- Source Code: github.com/traliran/traliran-ai-hub
🛠️ The Tech Stack: Ultimate Minimalism
The core philosophy of Traliran AI Hub is absolute lightness and privacy.
- The Architecture: 100% Client-Side. No Node.js. No Docker. No servers.
- The Stack: Vanilla HTML, CSS, and pure JavaScript.
- Data Privacy: Everything is stored right in your browser via LocalStorage. Your API keys, history, and custom configurations never touch my servers because I don't even have a server.
- Communication: Direct HTTPS fetch requests using the standard OpenAI API schema.
🔥 3 Killer Features That Make It Different
Since the app is lightweight, it doesn't mean it lacks power. I built tools that focus on speed and pure utility.
1. Hot Switching (Models & Providers)
Imagine you are working in the built-in IDE or deeply discussing a complex architecture with a model via OpenRouter. Suddenly, you run out of credits. Normally, you'd stop, go to a billing page, top up your account, and lose your flow.
With Traliran AI Hub, you can do Hot Switching. In exactly two clicks, you can switch the entire provider (e.g., from OpenRouter to Google Gemini API) and keep the exact same conversation flowing without missing a beat.
2. Full-Context Prompt & Config Store
Most AI UIs allow you to save system prompts. That's fine, but incomplete. In Traliran AI Hub, you can export and import entire AI Configurations via JSON.
A configuration includes everything: the system prompt, max tokens, top_p, temperature, selected provider, endpoint, and exact model. You can download your perfect setup, share it with a friend, or import a pre-made tool from the built-in Config Store with a single click.
Here is exactly how a clean, self-contained preset looks under the hood:
{
"provider": "groq",
"apiKey": "[your_api_key_here]",
"endpoint": "https://api.groq.com/openai/v1",
"botName": "Translator",
"systemPrompt": "You are an advanced expert translator. Your sole purpose is to translate any text provided into high-quality, fluent English (or the language specified by user), keeping idioms and contextual tone perfectly intact. Avoid chatbot greetings.",
"selectedModel": "llama-3.3-70b-versatile",
"temperature": 0.9,
"topP": 0.95,
"maxTokens": 3840
}
3. AI-Powered Smart Notes
We all take messy notes. Your thoughts are all over the place, and it looks like chaos. I built an AI Notes feature into the hub to fix this.
You write down your raw thoughts, click one button, and the AI cleans it up. Even better, it supports instant inline prompting:
- Write
[note] Structure this-> AI returns a beautiful, clean markdown list. - Write
[note] Expand on this-> AI fills in the missing details automatically.
🗺️ The Roadmap? There is none.
I don’t believe in fake corporate roadmaps. I build features based on two metrics: if I need it, or if a user asks for it.
Right now, my main focus is expanding the built-in Prompt/Config Store. I want to add more ready-to-use configurations so you can import specialized AI assistants instantly and make your daily workflows even faster.
🚀 Give it a try!
If you want a blazing-fast, secure, and completely private way to interact with AI without the bloat of modern web frameworks, check out the project.
Launch it instantly on GitHub Pages: traliran.github.io/traliran-ai-hub
Feel free to break it, test it, and let me know your thoughts in the comments below or open an issue on the GitHub repository.
What features do you think a pure vanilla AI client should have next? Let's discuss!
Top comments (1)
Hey everyone! Keeping it short and sweet: the project is fully open-source under the GNU AGPLv3 license, built with pure HTML/CSS/JS without any heavy frameworks, and ready to use right now via GitHub Pages.If you find any bugs or want to suggest a feature, drop a comment below or open an Issue on GitHub. Let's test it out!