Only build on recently modified

[jwhco]

I'm having trouble finding in the code where a filter could be done that applies AI to only recently modified files. There is a .smart-connects/multi/ folder that updates every file -- rather than those recently modified notes.

Would embedding for files that haven't changed not need to be rerun? Could Smart Connects keep track of the last runtime, then refresh when triggered or on some period I could set in the configuration?

My main objective would be to conserve tokens and not duplicate work. There aren't many differences between runs. Yet all files get modified. The count of embeds is similar each time.

While running, I can work at half capacity with 27k notes. I'd be okay with nightly batch runs of more intense GPU-based builds than regular refreshes while in work mode.

This is why I run a different configuration on my writing machines without this plug-in. There could be a caching scheme that could focus workload, which would allow processing to go faster.

P. s. I'm working on recording use cases for writers and content marketers using Obsidian to write copy. So far, I'm not clear what running the Smart Connect App is doing other than making my computer run tokens faster.

[jwhco]

Smart Connections 2.1.94 is very effective updating .smart-connects/multi

It is only running on new notes. Very few files change on disk between embeds.