• Microsoft unveiled "Microsoft Foundry Local," allowing users to execute LLMs directly on their PCs.
• The tool automates setup and optimizes PC performance for AI tasks.
• Users can install and run models like "Phi-3.5-mini" using simple command-line inputs.
• Local AI integrates seamlessly with Windows systems, reducing dependency on cloud-based AI services.
• The system supports various hardware configurations, including GPUs for enhanced performance.

https://github.com/LeapLabTHU/Absolute-Zero-Reasoner

https://www.arxiv.org/abs/2505.03335

IP - bound to 0.0.0.0 so that it serves external requests.

Windows - 

Set up an environment variable, OLLAMA_HOST, value 0.0.0.0 or 0.0.0.0:(port) - if you'd like to use different port number.  Then restart ollama

Linux -

sudo vi /etc/systemd/system/ollama.service

In [Service] section, add: 

Environment="OLLAMA_HOST=0.0.0.0"

$ sudo systemctl daemon-reload

$ sudo systemctl restart ollama

Keep model in memory longer

Environment variable, OLLAMA_KEEP_ALIVE to 10 (10 minutes, default is 5 minutes)

Model file location

Windows - %HOMEPATH%\.ollama\models

Linux - /usr/share/ollama/.ollama/models

Logs

Windows - %LOCALAPPDATA%\Ollama

Linux - 

sudo journalctl -u ollama.service > ollama_logs.txt

journalctl -e -u ollama

CMD

BASH