Feature Request: Enable System Time Synchronization for Models

[williamdefly]

Hello,

I am using Jan as a local assistant, and I would like to suggest a new feature: the ability to synchronize the system time and date with the AI model.

Currently, the model often lacks awareness of the current date and time, which limits its ability to handle time-sensitive tasks or provide accurate context. Having a feature that allows the model to retrieve the PC's system time would greatly improve its reliability as a personal assistant.

Could this be implemented as a configuration option or an automated synchronization feature?

Thank you for your hard work on Jan!

[qnixsynapse]

I think it is possible with the current version. If you go to assistant settings of your favorite assistant you can add a {{current_date}}

Although this is more of a model problem. Model like GPT-OSS already has this in its chat template.

[williamdefly]

Thank you very much for your reply and for your work on Jan. It really helps me understand how the current system works.

To take it a step further, would it be possible to consider adding a {{current_time}} variable, similar to the existing {{current_date}}? This would allow users to retrieve the system time in real-time within their system instructions, preventing the models from hallucinating incorrect times or having to rely on more complex tool calls.

Thanks again for your help and for developing this tool!

[qnixsynapse]

Adding {{current_time}} will cause cache invalidation. llama.cpp will prefill from the beginning since time changes faster than dates.