Could be a cool application to develop a QGIS process to do this, and then connect it in FarmOS 
Does that make sense?
1 Like
Maybe… I’d love it if QGIS were the right platform for this analysis work, but I’m guessing it’s less of an interactive problem and more of a backend process once the hands-on data science work happens - which probably happens in a notebook of some kind.
I guess it gets into questions of where that backend process is going to be run and on what frequency. Does the “orchestrator” (for lack of a better word) need access to metadata about the land assets (current crop type and geometry) or just the “AI parts”. Regarding frequency, if this is a tool used once every few years, it might be more practical to just dump the geometries and crop types to some sort of standardized file.
Wouldn’t a more conventional strategy be to implement an MCP for farmOS to expose the current state to the agents?