Privacy Local LLM GUI

A cross-platform, privacy-focused desktop application that provides a graphical interface for interacting with local LLMs via Ollama. This tool automatically installs its dependencies, manages an Ollama server, pulls models on demand, and streams chat responses in real time.

---

---

Features

• 🚀 Zero-touch dependency management
  Automatically detects missing Python packages (ollama, langchain_ollama, tqdm, ttkthemes, etc.), installs them with progress feedback, and restarts the app.

• 🖥️ Built-in Ollama server management
  Checks if an Ollama server is running, starts it if needed, and gracefully shuts it down on exit.

• 📦 On-the-fly model pulling
  Downloads models (llama3.2:1b by default, plus a curated list) with a progress bar and allows canceling or switching mid-download.

• 💬 Interactive chat UI

  • System/human/assistant message history
  • “Customer Service”, “Translator”, “Chat AI” roles out of the box
  • Adjustable creativity (temperature) and reply length (max tokens) sliders
  • Syntax-colored chat bubbles for user vs. assistant
  • Light/Dark theme toggle

• 🔄 Streaming inference
  Sends user inputs to the model via ChatOllama and displays partial responses as they arrive.

---

Requirements

• Python 3.8 or newer
• Ollama CLI installed and on your PATH
• Internet connection (only to pull models the first time)

Installation & Quick Start

1. Clone this repository:

   git clone https://github.com/passimon/Privacy-Local-LLM-GUI.git
   cd Privacy-Local-LLM-GUI

2. Make sure python3 and ollama are on your PATH.
3. Run the GUI:

   ./gui.py

   On first launch, a small splash window will pop up, install the Python dependencies, and then restart the app automatically.

---

Usage

1. Model Selector
   Choose from a list of preconfigured LLMs. Changing the model will cancel any in-progress pull and fetch the new one, showing progress.
2. Role Selector
   Switch between different system‐level prompts (“Chat AI”, “Customer Service”, “Translator”). Changing role resets the conversation history.
3. Creativity & Reply Length
   Adjust temperature (0.0–1.0) and max tokens (128–512) with sliders.
4. Chat Area
   ­– Type your prompt in the bottom entry and press Enter or Send.
   ­– The assistant’s response streams in real time.
5. Theme Toggle
   Switch between dark (“equilux”) and light (“breeze”) themes.
6. Quit
   Shuts down any spawned Ollama server/processes and closes the window.

---

Code Structure

• Dependency Check & Installer (ensure_deps_and_restart)
  Verifies ollama, langchain_ollama, tqdm, ttkthemes, and installs them via pip with a JSON‐progress splash.
• Ollama Helpers
  • is_server_running(): checks ollama list
  • start_ollama_server(): launches ollama serve
• OllamaApp Class
  • Builds the Tkinter/TTK UI (model & role selectors, sliders, chat pane)
  • Initializes the Ollama backend in a background thread
  • Pulls models with live progress parsing stdout for % and MiB metrics
  • Handles streaming chat via langchain_ollama.ChatOllama
  • Manages conversation history, UI updates, and cleanup

---

Customization

• Add Models
  In the source, modify the self.models list.
• Define New Roles
  Extend the self.roles dict with a new key/value pair:

  self.roles["MyRole"] = "System prompt text for MyRole..."

• Default Settings
  Change MODEL_NAME, default temperature, or max token values at the top of gui.py.

---

Troubleshooting

• ollama not found
  Make sure you’ve installed Ollama and it’s on your PATH.
• Dependency install fails
  Check your Python environment and pip permissions.
• Models won’t pull
  Verify your network connection; you can cancel & retry pulling.

---

License

This project is released under the MIT License.