Monitoring &health.

Mission control for every background service Typilot depends on. Checks run every few seconds. You can restart anything from here.

Ollama service

124 ms · VRAM 3.2 GB / 12 GB

Input monitor

X11 · /dev/input/event3

Local database

4.2 MB · SQLite

The three tiles.

Ollama service. The local inference engine. Tile shows reachability, response time, and GPU/VRAM usage from the backend.
Input monitor. The background keyboard listener. If it goes offline, you usually need to grant permissions or re-detect the device.
Local database. The settings store. Read-only tile - confirms health and shows the current file size.

Status badges.

Online. Service responded within the expected window.
Offline. Typilot cannot reach the service. Start manually or check Troubleshooting.
Starting. A start or restart command is in flight. Tile sits in loading state.
Attention. Slow response or recent warning - review the log stream.

Alerts drawer.

The right-hand panel surfaces the last five warnings or errors. Each alert names a source, a timestamp, and a one-line cause. Three recurring sources to watch:

Permissions. The keyboard listener wants accessibility access on macOS or uinput on Linux.
Model downloads. Ollama alerts during long downloads usually mean storage pressure or a dropped network.
Network checks. Proxy/firewall blocks show up with HTTP codes - update the allow list.

Quick fixes.

Ollama offline.

Click Start on the Ollama tile - Typilot launches with your default model.
If it bounces back to offline, open Monitoring → View logs for the detailed error.
Running an external Ollama instance? Ensure it is running and that Settings points at the right URL.

Keyboard listener offline.

Press Restart on the Input Monitor tile.
Just changed shortcuts? Settings → Shortcuts → Save reloads the listener.
Linux/Wayland: run keyboard detection to refresh the device path.

High response time.

Hover the Ollama tile for the latest latency + VRAM reading.
Close other GPU-heavy apps or switch to a smaller model when VRAM is near capacity.
Restart Ollama to clear cached workloads.

NEXT → 12 HISTORY & LOGS

Loading chapter

11 - HEALTH

Monitoring &health.

Mission control for every background service Typilot depends on. Checks run every few seconds. You can restart anything from here.

Ollama service

124 ms · VRAM 3.2 GB / 12 GB

Input monitor

X11 · /dev/input/event3

Local database

4.2 MB · SQLite

The three tiles.

Ollama service. The local inference engine. Tile shows reachability, response time, and GPU/VRAM usage from the backend.
Input monitor. The background keyboard listener. If it goes offline, you usually need to grant permissions or re-detect the device.
Local database. The settings store. Read-only tile - confirms health and shows the current file size.

Status badges.

Online. Service responded within the expected window.
Offline. Typilot cannot reach the service. Start manually or check Troubleshooting.
Starting. A start or restart command is in flight. Tile sits in loading state.
Attention. Slow response or recent warning - review the log stream.

Alerts drawer.

The right-hand panel surfaces the last five warnings or errors. Each alert names a source, a timestamp, and a one-line cause. Three recurring sources to watch:

Permissions. The keyboard listener wants accessibility access on macOS or uinput on Linux.
Model downloads. Ollama alerts during long downloads usually mean storage pressure or a dropped network.
Network checks. Proxy/firewall blocks show up with HTTP codes - update the allow list.

Quick fixes.

Ollama offline.

Click Start on the Ollama tile - Typilot launches with your default model.
If it bounces back to offline, open Monitoring → View logs for the detailed error.
Running an external Ollama instance? Ensure it is running and that Settings points at the right URL.

Keyboard listener offline.

Press Restart on the Input Monitor tile.
Just changed shortcuts? Settings → Shortcuts → Save reloads the listener.
Linux/Wayland: run keyboard detection to refresh the device path.

High response time.

Hover the Ollama tile for the latest latency + VRAM reading.
Close other GPU-heavy apps or switch to a smaller model when VRAM is near capacity.
Restart Ollama to clear cached workloads.

NEXT → 12 HISTORY & LOGS