I just got this working. The workflow is a little wonky to me, but better than I've seen from other AI-in-the-terminal solutions:<p>Setup:<p>1. Go to Settings (⌘+,)<p>2. General > AI. Put in your OpenAI key. I went with GPT-4o. Make sure you set the token limit to 4,096 or it wont go.<p>Using:<p>1. At the terminal window, hit Edit > Engage Artificial Intelligence, or just hit ⌘+Y.<p>2. Type in the command you want, then hit OK or Shift+Enter.<p>3. On the result window you can hit Shift+Enter again to paste the result into your terminal.<p>The whole Shift+Enter workflow is odd to me. I'd prefer just Enter. However, I like that I know exactly what's being sent to OpenAI. The integration doesn't try too hard to send everything over.<p>It's good to see iTerm get some updates too!
More discussion is over here: <a href="https://news.ycombinator.com/item?id=40425428">https://news.ycombinator.com/item?id=40425428</a>
I'd love to see the ability to hook into my self-hosted ollama (llama3) server running in my homelab. I'm not thrilled about giving OpenAI all my terminal history.