Navigation: Z-Cron Tools >

OLLAMA-START

Ollama can run AI models locally on your PC. The selected model provides the AI responses, while Ollama manages loading the model and providing access to it. With this Z-Cron command, you can configure the automatic start of Ollama as a scheduled job. This is useful if Ollama should be started automatically before other tasks or made available after the computer has started.

Point 1, Parameter Window

The button next to the command opens the Ollama Setup window. In this window, the start parameters for Ollama are configured, for example:

• Path to ollama.exe

• OLLAMA_HOST

• OLLAMA_ORIGINS

• Model folder

• Context length

• KV cache type

• Keep alive

• Queue

• Parallel requests

• GPU options

• Debug output

These settings define how Ollama will later be started by this Z-Cron job. Important: If the Ollama parameters are changed, Ollama must be restarted with the new settings.

Point 2, Start Ollama as a Windows System Service

If Ollama should run as a Windows system service, independent of a user logon, the corresponding service option in the job must be enabled. For this to work, Z-Cron itself must be installed and configured as a system service. This is already the case on many systems.

In this case, the job is not started by the logged-on user. Instead, the Z-Cron service runs the job in the background. This is useful if Ollama should already be available after Windows has started, for example:

• For a server

• For a computer in the local network

• For a workstation where Ollama should run without a user logon

Important: If Z-Cron is not installed as a system service, Ollama cannot be started as a system service this way. In this case, use the Autostart option instead. Then Ollama starts when the user logs on to Windows and Z-Cron is started.

Start Parameters, Parameter Window

In this window, you define how Ollama is started by the program in server mode. Ollama is the local AI runtime environment. To allow other parts of the program to access a local model, Ollama must run as a server. For this purpose, the program starts ollama.exe with the settings entered here. The settings are passed to Ollama when it is started. If Ollama is already running, you must first stop Ollama and then restart it with the new settings.

Link docs.ollama.com

Typical Causes of Problems

• The path to ollama.exe is incorrect.

• Ollama is already running with different settings.

• The specified model folder is not accessible.

• Port 11434 is blocked by a firewall.

• OLLAMA_HOST is set to localhost only, although access from the network is required.

• The computer does not have enough RAM or GPU memory for the selected model and context length.

• An experimental GPU option is enabled, although the hardware or driver does not support it properly.

• Several large models are loaded at the same time.

Note

These settings only affect the start of the Ollama server by Z-Cron. If Ollama is started in another way, for example through the normal Ollama Windows application or a manual command prompt, different settings may apply.

A separate stop command is not required for restarting Ollama. Before starting Ollama, OLLAMA-START terminates any running Ollama instances and then starts Ollama again with the configured settings.