Documentation

Monitoring & Health

The Monitoring page keeps an eye on every background service Typilot depends on. Use it as your mission control when something feels off.

🩺 Checks run in the background every few seconds • You can restart services safely from here • Alerts stay visible until the issue is resolved

Service tiles at a glance

Ollama service

Shows whether the Ollama daemon is reachable, how long it took to respond, and the GPU/VRAM usage reported by the backend.

Start
Stop
Restart

Input monitor

Tracks Typilot’s background keyboard listener. If it drops offline, you may need to grant permissions or re-detect your keyboard device.

Start
Stop
Restart

Local database

Confirms the SQLite settings store is healthy and displays the current file size. This tile is informational only and cannot be restarted.

Read-only

Status badges explained

Online

Service responded within the expected timeframe and is ready.

Offline

Typilot cannot reach the service. Start it manually or open the troubleshooting card for next steps.

Starting…

A start/restart command is in progress. Typilot keeps the tile in a loading state until it hears back from the service.

Attention

Something needs review (for example, Ollama responded slowly or the last check returned a warning).

Alerts drawer

The right-hand panel lists the last five warnings or errors pulled from the log stream. Each alert shows the source, time, and details.

Permission issues

Look for alerts sourced from the keyboard listener. Most mention missing accessibility access on macOS or uinput groups on Linux.

Model downloads

Ollama alerts can appear during large downloads if storage is full or your network connection drops mid-transfer.

Network checks

If a proxy or firewall blocks requests, monitoring records the HTTP error code so you can update your allow list.

Quick fixes

Ollama offline

  1. Click “Start” on the Ollama tile. Typilot launches it with your default model.
  2. If it immediately returns offline, open Monitoring → View logs to inspect the detailed error message.
  3. On Windows or macOS, ensure the Ollama desktop app is installed and logged in at least once.

Keyboard listener offline

  1. Press “Restart” on the Input Monitor tile.
  2. If you just changed shortcuts, open Settings → Shortcuts and click Save to reload the listener.
  3. On Linux/Wayland, run the keyboard detection to refresh the device path after plugging in new hardware.

High response time

  1. Hover over the Ollama tile to see the latest response latency and VRAM usage.
  2. Close unused applications or switch to a smaller model if VRAM is near capacity.
  3. Restart Ollama to clear cached workloads.

Best practices

  • Leave Monitoring open while you run automated tasks—alerts show up in real time.
  • After installing new models, restart Ollama from here to free unused GPU memory.
  • If a tile flickers between online and offline, check your firewall rules or VPN.