Pivoting on cookingspace.com
Some of my readers may have experimented with my cookingspace.com web app. One big problem, according to user feedback, was the hassle of entering the food you have "on hand" in your kitchen - this information is used to suggest alternative ingredients if you are missing a few food items called for in a recipe. The problem was that it takes 5 or 10 minutes a week of time for a user to keep their "on hand" ingredient data up to date. There are 61122 rows in the user table for the web app, but of those people very few people entered ingredients.
So, this functionality should go. The problem is that I spent a lot of effort creating a model from 100K public domain recipes for suggesting alternative ingredients to any recipes - hard to let that work go. Actually, I am going to keep using this model but use it in a different way. I going to toss the Postgres database for each user's ingredients, and instead, let people search for recipes without pre-screening them for being good matches for what food the user has on hand. Instead, when a user is looking at any recipe, they will be able be able to select ingredients they don't have and the recipe will be re-displayed with alternative ingredients (which they may not have either). This new approach preserves some of the artificial intelligence functionality for users while hopefully removing the hassles of using the existing system.
I hope to make these changes in the next week or two.