meal-scheduler VS FactGraph

I use Minizinc in a personal toy project (https://gitlab.com/dustin-space/meal-scheduler), and GECODE or Google's ortools solver at the backend. It's used for meal planning. Unfortunately it's way way slower than I'd hope. I suspect I just have the domain not modeled efficiently. Maybe if I had a few days to put into it, and learn how to properly debug the CSP solver step by step, it might help...

That sounds like a fun application, both the usage and the implementation.

I wonder if you have any interesting example data-files that could be used with the model, preferable both something small and something larger? Would be fun to test the model locally to see how it behaves.

Notes: I'm assuming here that https://gitlab.com/dustin-space/meal-scheduler/-/blob/master... is the model used.

I used to have a project like this. I was going to call it FactGraph: https://github.com/FactGraph/FactGraph/wiki

My idea was to build up a big community-maintained database containing facts and evidence, where everything is linked into a huge network. Everything would have a weight (sometimes automatically calculated from parent nodes), and the software would calculate probabilities for some big questions. Every user could also build their own personalized graph to explore their own worldview, and maybe even uncover some cognitive dissonance that they weren't aware of. Or you could use it to compare and contrast different philosophies, religions. Could even calculate a "coherence score" for each religion and denomination after crunching all of the available evidence.

Then I discovered RootClaim: https://www.rootclaim.com

They're doing something very similar, with a more targeted approach where they focus on some specific questions. e.g. COVID-19: https://www.rootclaim.com/analysis/what-is-the-source-of-cov...

RootClaim really seems to be nailing it so far, and hopefully they can continue to grow and become something like the project I was imagining.