+ Common Issues — Cai Docs

Common Issues

Global hotkey doesn’t work

  • Open System Settings → Privacy & Security → Accessibility and make sure Cai is listed and enabled
  • If it’s listed but still not working, remove Cai from the list, then re-add it
  • Check that no other app is using ⌥C (or your custom hotkey) — e.g., Raycast, Alfred, BetterTouchTool

LLM not connecting

  • Verify your server is running: 
    curl http://127.0.0.1:1234/v1/models
  • Check that the URL in Preferences matches your server’s address and port
  • Ollama uses port 11434, LM Studio uses 1234 — make sure you selected the right provider
  • If using a Custom URL, make sure it includes the full path (e.g., http://127.0.0.1:1234/v1)

Date/meeting not detected

  • Detection works best with English dates (“Tuesday at 3pm”, “lunch tomorrow at noon”)
  • Try rephrasing: “3pm tomorrow” instead of “tomorrow 15h”
  • Some informal formats may not be recognized — explicit dates work most reliably

Calendar event has wrong details

Cai generates ICS files from detected text. For best results, include a clear date/time and optionally a location in the copied text. The ICS file opens in your default calendar app, where you can review and edit before saving.

Built-in model can’t start

Cai’s built-in LLM server uses ports 8690–8699, picking the first available one. If a previous server didn’t shut down cleanly, Cai automatically cleans up orphan processes on next launch.

If you still see issues, open Terminal:

# Find any leftover llama-server processes
lsof -i :8690-8699 | grep LISTEN

# Kill the process (replace <PID> with the actual process ID)
kill -9 <PID>

Apple Intelligence not working

Apple Intelligence requires:

  • Apple Silicon (M1 chip or later)
  • macOS 26 or later
  • Apple Intelligence enabled in System Settings > Apple Intelligence & Siri

If you meet the requirements but Apple Intelligence isn’t showing as a provider option, make sure you’ve enabled it in System Settings first, then restart Cai.

Result text looks odd

Cai renders results as Markdown. If your LLM returns unexpected formatting, try a different model or adjust your custom prompt to request plain text output.

Built-in model download fails

The built-in model (~2.15 GB) downloads on first launch. If the download fails:

  • Check your internet connection
  • Make sure you have enough disk space
  • Try quitting and relaunching Cai — the download resumes where it left off
  • Check the ~/Library/Application Support/Cai/models/ directory for partial downloads