Draggy
Living with it
Troubleshooting
The problems that actually come up, and what fixes them.
It will not get past the splash screen
Checking local AI service means Ollama is not running. Start it and relaunch. Downloading the model is genuinely several gigabytes.
Replies are extremely slow
Turn on Show speed and memory in
Settings → General and read the GPU figure. Anything under about
60% means the model is spilling onto the processor; step down a size in
Settings → Models.
Every reply says "model load"
Something else is evicting the model, usually the talk model or another application using the same card. Pick a smaller talk model, or set a fixed Context window so the window stops changing size.
The model describes a tool call instead of making one
It is too small. Below about 4B parameters, tool calling is unreliable and gets worse the more tools are switched on. Turn some off or move up a size.
An extension will not start
| Message | Means |
|---|---|
| Not configured yet | A required field is empty. Open the setup fields. |
| Did not answer in time | npm could not fetch it, or it needs a runtime you lack. |
| npm could not be found | Install Node.js and restart Draggy. |
Voice mode will not start
Permission denied is the microphone. No speech voices installed means your system has none; add one in its settings, or use the Natural voice instead.
It refused a file path or a local address
That is the guard working. The model may only reach the project folder, and may
only fetch public http and https addresses. Attach the
file yourself, or open the page in the browser.
Updates never appear
Updates only run in an installed build; from source the panel says so. macOS builds are unsigned and cannot update themselves.
Still stuck
Settings → Data → Open log folder has the log, which is where the
actual reason usually is. Then
open an issue with it.