📱 PocketAI

A lightweight Python CLI that connects to a local Ollama server running on an Android phone (Termux) or a computer.

PocketAI lets you use a local Large Language Model (LLM) running on your own device without relying on cloud services. It communicates with Ollama through its REST API and provides a simple, clean command-line chat interface.

---

✨ Features

• 🤖 Local AI Chat: Converse with LLMs completely offline.
• 📱 Termux Ready: Works seamlessly with Ollama running on Android.
• 💻 Clean Interface: A minimal and easy-to-use command-line interface.
• 🔒 Fully Private: 100% local processing; your data never leaves your network.
• ⚡ API Powered: Communicates efficiently via the Ollama REST API.
• 🛠️ Built-in Commands: Use /help for guidance and /bye to exit securely.

---

📂 Project Structure

PocketAI/
│
├── pocketai/
│   ├── ai.py
│   └── main.py
│
├── README.md
├── LICENSE
├── requirements.txt
├── .gitignore
└── .gitattributes

⚙️ Requirements

💻 PC

• Python 3.13+
• Git
• Requests Library (pip install requests)

📱 Phone

• Android Device
• Termux
• Ollama for Android (via Termux)

🚀 Installing PocketAI (PC)

1. Clone the repository

git clone [https://github.com/kongogaming/PocketAI.git](https://github.com/kongogaming/PocketAI.git)
cd PocketAI

2. Create a virtual environment

python -m venv .venv

3. Activate the virtual environment

• Windows:

  .venv\Scripts\activate
  

• Linux / macOS:

  source .venv/bin/activate
  

4. Install dependencies

pip install -r requirements.txt

📱 Setting up Ollama on Android (Termux)

1. Install Termux

Download the latest version from the official GitHub release page or F-Droid.

2. Update packages

pkg update && pkg upgrade

3. Install Ollama

Follow the official Ollama Android/Termux installation instructions.

4. Download a model

For PocketAI v0.1.0, this project uses gemma3:270m by default:

ollama pull gemma3:270m

(You can replace this with any Ollama-compatible model later).

5. Allow network access

Start Ollama so other devices on your local Wi-Fi can connect to it.

export OLLAMA_HOST=0.0.0.0:11434
ollama serve

⚠️ Note: Leave this terminal running while you use PocketAI!

6. Verify the Connection

On the phone:

curl [http://127.0.0.1:11434](http://127.0.0.1:11434)

From your PC (Browser): Navigate to your phone's IP address on port 11434:

http://PHONE_IP:11434
# Example: [http://192.168.29.142:11434](http://192.168.29.142:11434)

Expected Output: Ollama is running

🛠️ Configure PocketAI

Open the pocketai/ai.py file in your code editor and set your Ollama server URL to match your phone's IP address:

# pocketai/ai.py
url = "http://YOUR_PHONE_IP:11434/api/generate"

# Example:
# url = "[http://192.168.29.142:11434/api/generate](http://192.168.29.142:11434/api/generate)"

🎮 Run PocketAI

Start the CLI application:

python pocketai/main.py

🖥️ Example Output

=============================================
               POCKET AI
=============================================

Type /help for commands.

You > Hello
AI  > Hello! How can I help you today?

You > /help

Commands:
/help - Show available commands
/bye  - Exit PocketAI

You > /bye
AI  > Goodbye!

⌨️ Commands

Command	Description
/help	Show available commands and help menu
/bye	Exit the PocketAI CLI

🧠 Models

PocketAI works with any Ollama model.

• Current Default: gemma3:270m
• Popular Alternatives:
  • qwen2.5:3b
  • llama3.2
  • phi4-mini Browse the Official Ollama Model Library for more options.

🔗 Ollama API

PocketAI communicates using the standard Ollama REST API.

• Example Endpoint: POST /api/generate
• Documentation: Ollama API Docs

🗺️ Roadmap

• Local AI chat
• Modular architecture
• GitHub release
• Configuration file support
• Streaming responses
• Chat history memory
• Model switching in CLI
• Voice support

📄 License

This project is licensed under the MIT License. See the LICENSE file for more details.

👨‍💻 Author

Somya Ranjan Pal (kongogaming)

• GitHub: @kongogaming