Prudii Mail - Private Email Client for Windows. Email that's █

Ollama is a free, open-source tool that lets you run large language models locally on your computer. Prudii uses it as the AI backend — instead of sending your emails to OpenAI or Google, everything stays on your machine. You install Ollama once, download a model, and Prudii talks to it on localhost. No API keys, no subscriptions, no cloud.

No. Ollama runs entirely on your machine. When Prudii asks for a summary or reply suggestion, it sends the email text to localhost:11434 — that's your own computer. Nothing leaves your network.

For email summaries and reply suggestions, 7B–8B parameter models hit the sweet spot between speed and quality:

• llama3.2 — Fast, solid all-rounder. Good starting point.
• mistral — Great at following instructions, good for replies.
• gemma2 — Compact and fast, decent summaries.

Install one with ollama pull llama3.2 and you're good to go. Bigger models (13B+) give better results but need more RAM and a decent GPU.

Not strictly, but it helps a lot. Ollama can run on CPU only, but responses will be slow (30–60 seconds for a summary). With a GPU that has 6+ GB VRAM (like an RTX 3060 or better), you'll get responses in 2–5 seconds. 16 GB RAM minimum is recommended either way.

1. Download and install Ollama from ollama.com
2. Open a terminal and run: ollama pull llama3.2
3. Make sure Ollama is running (it starts automatically on install)
4. In Prudii, go to Settings → AI and enable it
5. Click "Test Connection" — your models will show up
6. Select a model and you're done

Right now, two things:

• Summarize emails — Get a quick summary of long emails or entire threads. Click the sparkle icon on any email.
• Reply suggestions — The AI suggests three replies in different tones: professional, friendly, and concise. Pick one, edit it, send it.

Both features are available as a Premium feature. The AI is entirely optional — Prudii works perfectly fine without it.