🦙 llama.ui - Minimal Interface for Local AI Companion ✨

Tired of complex AI setups? 😩 llama.ui is an open-source desktop application that provides a beautiful ✨, user-friendly interface for interacting with large language models (LLMs) powered by llama.cpp. Designed for simplicity and privacy 🔒, this project lets you chat with powerful quantized models on your local machine - no cloud required! 🚫☁️

⚡ TL;DR

This repository is a fork of llama.cpp WebUI with:

• Fresh new styles 🎨
• Extra functionality ⚙️
• Smoother experience ✨

🌟 Key Features

1. Multi-Provider Support: Works with llama.cpp, LM Studio, Ollama, vLLM, OpenAI,.. and many more!

2. Conversation Management:

   • IndexedDB storage for conversations
   • Branching conversation support (edit messages while preserving history)
   • Import/export functionality

3. Rich UI Components:

   • Markdown rendering with syntax highlighting
   • LaTeX math support
   • File attachments (text, images, PDFs)
   • Theme customization with DaisyUI themes
   • Responsive design for mobile and desktop

4. Advanced Features:

   • PWA support with offline capabilities
   • Streaming responses with Server-Sent Events
   • Customizable generation parameters
   • Performance metrics display

5. Privacy Focused: All data is stored locally in your browser - no cloud required!

6. Localized Interface: Most popular language packs are included in the app, and you can choose the language at any time.

🚀 Getting Started in 60 Seconds!

💻 Standalone Mode (Zero Installation)

1. ✨ Open our hosted UI instance
2. ⚙️ Click the gear icon → General settings
3. 🌐 Set "Base URL" to your local llama.cpp server (e.g. http://localhost:8080)
4. 🎉 Start chatting with your AI!

🔧 Need HTTPS magic for your local instance? Try this mitmproxy hack!

Uh-oh! Browsers block HTTP requests from HTTPS sites 😤. Since llama.cpp uses HTTP, we need a bridge 🌉. Enter mitmproxy - our traffic wizard! 🧙‍♂️

Local setup:

mitmdump -p 8443 --mode reverse:http://localhost:8080/

Docker quickstart:

docker run -it -p 8443:8443 mitmproxy/mitmproxy mitmdump -p 8443 --mode reverse:http://localhost:8080/

Pro-tip with Docker Compose:

services:
  mitmproxy:
    container_name: mitmproxy
    image: mitmproxy/mitmproxy:latest
    ports:
      - '8443:8443' # 🔁 Port magic happening here!
    command: mitmdump -p 8443 --mode reverse:http://localhost:8080/
    # ... (other config)

⚠️ Certificate Tango Time!

1. Visit http://localhost:8443
2. Click "Trust this certificate" 🤝
3. Restart 🦙 llama.ui page 🔄
4. Profit! 💸

Voilà! You've hacked the HTTPS barrier! 🎩✨

🖥️ Full Local Installation (Power User Edition)

1. 📦 Grab the latest release from our releases page
2. 🗜️ Unpack the archive (feel that excitement! 🤩)
3. ⚡ Fire up your llama.cpp server:

Linux/MacOS:

./server --host 0.0.0.0 \
         --port 8080 \
         --path "/path/to/llama.ui" \
         -m models/llama-2-7b.Q4_0.gguf \
         --ctx-size 4096

Windows:

llama-server ^
             --host 0.0.0.0 ^
             --port 8080 ^
             --path "C:\path\to\llama.ui" ^
             -m models\mistral-7b.Q4_K_M.gguf ^
             --ctx-size 4096

4. 🌐 Visit http://localhost:8080 and meet your new AI buddy! 🤖❤️

🌟 Join Our Awesome Community!

We're building something special together! 🚀

• 🎯 PRs are welcome! (Seriously, we high-five every contribution! ✋)
• 🐛 Bug squashing? Yes please! 🧯
• 📚 Documentation heroes needed! 🦸
• ✨ Make magic with your commits! (Follow Conventional Commits)

🛠️ Developer Wonderland

Prerequisites:

• 💻 macOS/Windows/Linux
• ⬢ Node.js >= 22
• 🦙 Local llama.cpp server humming along

Build the future:

npm ci       # 📦 Grab dependencies
npm run build  # 🔨 Craft the magic
npm start    # 🎬 Launch dev server (http://localhost:5173) for live-coding bliss! 🔥

🧰 Preconfiguring Defaults

Planning to redistribute the app with opinionated settings out of the box? Any JSON under src/config is baked into immutable defaults at build time (see src/config/index.ts).

If those baked defaults include a non-empty baseUrl, the inference server will auto-sync on first load so model metadata is fetched without requiring manual input.

🏗️ Architecture

Core Technologies

• Frontend: React with TypeScript
• Styling: Tailwind CSS + DaisyUI
• State Management: React Context API
• Routing: React Router
• Storage: IndexedDB via Dexie.js
• Build Tool: Vite

Key Components

1. App Context: Manages global configuration and settings
2. Inference Context: Handles API communication with inference providers
3. Message Context: Manages conversation state and message generation
4. Storage Utils: IndexedDB operations and localStorage management
5. Inference API: HTTP client for communicating with inference servers

📜 License - Freedom First!

llama.ui is proudly MIT licensed - go build amazing things! 🚀 See LICENSE for details.

---

Made with ❤️ and ☕ by humans who believe in private AI