🧠 ClipLLM

A lightweight clipboard-to-Ollama bridge for macOS. Copy a prompt, get an AI response — right in your clipboard.

How It Works

1. Copy text starting with //ollama (e.g., //ollama explain quantum physics)
2. Wait a moment while ClipLLM sends your prompt to Ollama.
3. Paste — the AI response is now in your clipboard!

---

Choose Your Version

You can run ClipLLM using either Python or Java. Both do the same thing, but have different setups.

Option A: Python Version

Best if you prefer working with scripts and the requests library.

Setup:

# Create a virtual environment and install dependencies
python3 -m venv venv
source venv/bin/activate
pip install -r requirements.txt

Run:

source venv/bin/activate
python clipllm.py

---

☕ Option B: Java Version (Native & Fast)

Best for a "zero-dependency" experience. It uses pure Java 11+ features.

Setup: No installation needed! Just make sure you have a JDK installed.

Run:

# Compile and run in one step
java ClipLLM.java

---

📋 Prerequisites (For Both)

• Ollama installed and running — ollama.com
• Model pulled:

  ollama pull gemma4:e4b
  ollama pull nemotron-3-nano:4b

• Ollama serving: Make sure the Ollama app is open or run ollama serve.

---

⚙️ Configuration

Variable	Default	Description
PRIMARY_MODEL	gemma4:e4b	The main high-quality model
FALLBACK_MODEL	nemotron-3-nano:4b	Faster model used if the primary fails
TRIGGER	//ollama	Text prefix that triggers a prompt
POLL_INTERVAL	500ms	How often to check the clipboard

---

Technical Comparison

Feature	Python Implementation	Java Implementation
Clipboard	subprocess -> pbpaste/pbcopy	java.awt.Toolkit (Native API)
HTTP Client	requests (External Lib)	java.net.http.HttpClient (Native)
JSON Safety	Manual String replace	escapeJson() helper function
Concurrency	time.sleep()	Thread.sleep()
Dependencies	1 (requests)	0 (Zero!)

Why Java?

While the Python version is easy to write, the Java version is more powerful because it talks to the macOS system directly using the java.awt library, and it doesn't need any pip install to work.

What is the "Thread"?

Both versions use a Thread Sleep. This is crucial for your CPU's health. Instead of checking the clipboard millions of times per second (which would make your Mac hot), the program "naps" for 500ms between checks, keeping your CPU usage near 0%.

---

Troubleshooting

Problem	Solution
"Cannot connect to Ollama"	Run ollama serve or open the Ollama app.
"Model not found"	Run ollama pull gemma4:e4b.
"Slow responses"	The app will automatically try the Fallback model if it takes too long.
"Nothing happens"	Ensure your copied text starts exactly with //ollama (with the space).