Loading chapter

14 - RECOVERY

Troubleshooting.

Five categories of failure, ten common questions, each with the exact fix. If your problem doesn't fit, contact support.

01 Installation
02 Connection
03 Performance
04 Shortcuts
05 Image analysis

Try these four first.

Restart everything. Typilot first - it restarts the bundled engine too - then the machine if that doesn't help.
Confirm the engine is running. Monitoring shows the Ollama tile - click Start if it's offline. http://localhost:11434 should respond in your browser.
Review settings. Default model selected, shortcut not conflicting, Ollama URL correct if you use an external instance.
Read the logs. Sidebar → Logs surfaces the actual error.

Installation

When the app won't launch or won't install.

Typilot won't start after installation.

Check minimum requirements: 4 GB RAM, 2 GB disk.
Run as administrator (Windows) or with sudo (Linux).
Inspect system logs for the launch error.
Reinstall if nothing changes.

The AI engine won't start.

The Ollama engine ships inside Typilot - no separate install is needed.
Open Monitoring and click Start on the Ollama tile.
Restart Typilot, then the machine if that doesn't help.
Running your own external Ollama? Confirm it installed correctly (direct download at ollama.com, admin/root privileges).

Connection

When Typilot can't reach Ollama or models won't show.

Typilot can't connect to Ollama.

Open Monitoring - the Ollama tile should show Online. Click Start if not; Typilot manages the bundled engine.
Open http://localhost:11434 in your browser - should respond.
Restart Typilot.
On an external Ollama instance: confirm the URL in Settings, check Defender Firewall (Windows) or sudo systemctl status ollama (Linux).

Models not showing.

Confirm Ollama connected.
Refresh the Models page.
Restart Typilot.
Run ollama list in terminal - confirm models are actually pulled.

Performance

When responses lag or the app freezes.

Responses are very slow.

Switch to a smaller model (llama3.2:1b or 3b).
Close other heavy apps to free RAM.
8 GB+ RAM recommended for 3b models.
Settings → AI behavior → lower typing speed for faster output.
Match the model to your hardware tier.

Application freezes during AI use.

Smaller model.
Close other apps to free RAM.
Check system resource use (Task Manager / Activity Monitor).
Restart Typilot.

Shortcuts

When the activation combo or send key won't fire.

Shortcut doesn't work.

Check for OS conflicts (Spotlight, IME switch).
Change the shortcut in Settings.
macOS: confirm Accessibility permission.
Linux: confirm input + uinput groups.
Restart Typilot after changes.

Shortcut works but the command doesn't run.

Confirm Ollama is connected.
A model is selected in Settings.
Default send key is . (period) - check yours.
Logs page surfaces the underlying error.

Image analysis

When img: returns nothing or garbage.

img: command doesn't work.

Vision-capable model downloaded (llama3.2-vision:11b).
OCR is built into Typilot - no separate install is needed.
Capture a clear region with readable text.
Restart Typilot if capture stops responding.

Text extraction fails or is inaccurate.

Larger region or full screen.
Higher-contrast text.
Some fonts and stylizations confuse OCR - try a clearer source.

STILL STUCK? Email contact@typilot.com · See also 01 GETTING STARTED

Loading chapter

14 - RECOVERY

Troubleshooting.

Five categories of failure, ten common questions, each with the exact fix. If your problem doesn't fit, contact support.

01 Installation
02 Connection
03 Performance
04 Shortcuts
05 Image analysis

Try these four first.

Restart everything. Typilot first - it restarts the bundled engine too - then the machine if that doesn't help.
Confirm the engine is running. Monitoring shows the Ollama tile - click Start if it's offline. http://localhost:11434 should respond in your browser.
Review settings. Default model selected, shortcut not conflicting, Ollama URL correct if you use an external instance.
Read the logs. Sidebar → Logs surfaces the actual error.

Installation

When the app won't launch or won't install.

Typilot won't start after installation.

Check minimum requirements: 4 GB RAM, 2 GB disk.
Run as administrator (Windows) or with sudo (Linux).
Inspect system logs for the launch error.
Reinstall if nothing changes.

The AI engine won't start.

The Ollama engine ships inside Typilot - no separate install is needed.
Open Monitoring and click Start on the Ollama tile.
Restart Typilot, then the machine if that doesn't help.
Running your own external Ollama? Confirm it installed correctly (direct download at ollama.com, admin/root privileges).

Connection

When Typilot can't reach Ollama or models won't show.

Typilot can't connect to Ollama.

Open Monitoring - the Ollama tile should show Online. Click Start if not; Typilot manages the bundled engine.
Open http://localhost:11434 in your browser - should respond.
Restart Typilot.
On an external Ollama instance: confirm the URL in Settings, check Defender Firewall (Windows) or sudo systemctl status ollama (Linux).

Models not showing.

Confirm Ollama connected.
Refresh the Models page.
Restart Typilot.
Run ollama list in terminal - confirm models are actually pulled.

Performance

When responses lag or the app freezes.

Responses are very slow.

Switch to a smaller model (llama3.2:1b or 3b).
Close other heavy apps to free RAM.
8 GB+ RAM recommended for 3b models.
Settings → AI behavior → lower typing speed for faster output.
Match the model to your hardware tier.

Application freezes during AI use.

Smaller model.
Close other apps to free RAM.
Check system resource use (Task Manager / Activity Monitor).
Restart Typilot.

Shortcuts

When the activation combo or send key won't fire.

Shortcut doesn't work.

Check for OS conflicts (Spotlight, IME switch).
Change the shortcut in Settings.
macOS: confirm Accessibility permission.
Linux: confirm input + uinput groups.
Restart Typilot after changes.

Shortcut works but the command doesn't run.

Confirm Ollama is connected.
A model is selected in Settings.
Default send key is . (period) - check yours.
Logs page surfaces the underlying error.

Image analysis

When img: returns nothing or garbage.

img: command doesn't work.

Vision-capable model downloaded (llama3.2-vision:11b).
OCR is built into Typilot - no separate install is needed.
Capture a clear region with readable text.
Restart Typilot if capture stops responding.

Text extraction fails or is inaccurate.

Larger region or full screen.
Higher-contrast text.
Some fonts and stylizations confuse OCR - try a clearer source.

STILL STUCK? Email contact@typilot.com · See also 01 GETTING STARTED