Troubleshooting

Common issues when installing and running Didon on macOS.

Didon Opens But Does Not Analyze

Check that Ollama is running:

curl http://localhost:11434/api/tags

If it is not running, start Ollama.

Model Not Found

Pull the model you chose during installation. For example:

ollama pull qwen3-vl:2b-instruct

If you chose the optional 4B model (16 GB+ RAM Macs):

ollama pull qwen3-vl:4b-instruct

Screen Recording Error

Make sure Didon is enabled here:

System Settings > Privacy & Security > Screen & System Audio Recording

After changing this permission, quit and reopen Didon.

macOS Blocks The App

Didon is not yet notarized during the MVP stage, so macOS Gatekeeper may block it on first launch.

Try right-clicking the app and choosing Open first. Only if you still see a warning, remove the quarantine flag:

xattr -dr com.apple.quarantine /Applications/Didon.app

We plan to notarize the app with Apple in a future release so this step will no longer be needed.

High Ollama / llama-server Memory Usage

Didon analyzes a screenshot about every 30 seconds while monitoring. Each request sends one image to your local Ollama model.

Starting with v2.3, Didon passes keep_alive: 0 on every Ollama request so the model unloads after each analysis instead of staying in memory for several minutes. This keeps llama-server RAM much more stable during long sessions.

If memory still climbs over many hours:

Confirm you are on Didon v2.3 or newer.
Use the recommended Qwen 2B model (qwen3-vl:2b-instruct) unless you have 16 GB+ RAM and need the 4B model.
Restart Ollama: quit the Ollama app and open it again, or run ollama stop in Terminal.
Pause Didon when you are not tracking — no new analyses will run.

Check what Ollama has loaded:

ollama ps

Only the model you selected in Didon settings should appear while a screenshot is being analyzed.