AI Features & Privacy
Privacy-first AI that runs locally on your device. No data collection, no tracking, optional cloud features with your own API keys.
How It Works
Our AI runs entirely on your computer using Ollama. Nothing is sent to external servers unless you explicitly enable cloud features.
Local First
AI runs on your computer—no internet required
No Data Collection
We never see or store your data
Optional Cloud
Use your own API keys if desired
Graceful Fallback
Features work even without AI
Local AI with Ollama
We use Ollama, a free tool that runs AI models directly on your computer.
Recommended Models by VRAM
Select your GPU memory to see the best models for your system. All use Q4_K_M quantization—the optimal balance of quality and efficiency.
Integrated GPUs, entry-level cards
llama3.2:3b 2GB Fast Quick responses, basic tasks phi3:mini 2GB Fast Compact and efficient deepseek-r1:8b 5GB Reasoning Step-by-step problem solving qwen2.5-coder:7b 4.5GB Coding Code specialist moondream 1.7GB Vision Image understanding RTX 3060 12GB, RTX 4070
gemma2:9b 5GB Fast Google's efficient model deepseek-r1:14b 9GB Reasoning Strong reasoning ability qwen2.5-coder:14b 9GB Coding Professional grade llava:13b 8GB Vision Full vision capabilities RTX 4080, Mac 16GB+
phi4:14b 8GB Fast Microsoft's latest, excellent qwq:32b 18GB Reasoning Outstanding chain-of-thought qwen2.5-coder:14b 9GB Coding Room for longer context mistral:7b 4GB Creative Creative writing RTX 3090/4090, Mac Studio
deepseek-r1:32b 20GB Reasoning Near cloud quality qwen2.5-coder:32b 20GB Coding Expert-level coding llava:34b 20GB Vision Advanced visual analysis llama3.1:70b Q3 General 70B with quantization Multi-GPU, Mac Studio 64GB+
llama3.3:70b 40GB General Flagship open model qwen2.5-coder:32b +32K ctx Coding Full context for codebases qwen2-vl:72b Large Vision Best-in-class vision Quick Tips
Use fast models for simple tasks, larger for complex reasoning
More VRAM = longer context = fewer hallucinations
Coder models beat general models at coding tasks
Cloud Providers (Optional)
Use your own API keys for more powerful capabilities. We never see your keys—they're stored locally and sent directly to providers.
Haiku, GPT-4o-mini, Flash cost less per token
Opus, GPT-4 for complex, accuracy-critical tasks
Privacy Guarantees
❌ What We Never Do
- Collect or store your data
- Send data to our servers
- Track your AI usage
- Share with third parties
✅ What Stays Local
- All text you process
- Documents you summarize
- Images you analyze
- Conversation history
Quick Setup
Pull a Model
Run ollama pull llama3.2:3b in your terminal to download a model.
Open Our App
The app automatically detects Ollama at localhost:11434.
Recommended Models by VRAM
# 8GB VRAM - Entry level ollama pull llama3.2:3b ollama pull moondream # 12-16GB VRAM - Best experience ollama pull phi4:14b ollama pull deepseek-r1:14b # 24GB+ VRAM - Maximum quality ollama pull deepseek-r1:32b ollama pull qwen2.5-coder:32b Cloud API Setup (Optional)
- Get an API key from: Anthropic, OpenAI, Google, or Perplexity
- Open app settings → AI Settings → Cloud Providers
- Select provider, paste key, choose model
Troubleshooting
Ensure Ollama is running. Check localhost:11434 in browser. Restart Ollama if needed.
Try a smaller model. Close other apps. Check available memory.
Check internet connection. Ensure enough disk space. Try smaller model first.
Verify API key. Check account credits. Try a different model.