I recently had an idea for an app that I would like to start developing for personal use and develop

You could use an OLS regression for this, or you could just use a machine learning algorithm. A decision tree will probably work just as well as any regression model (and possibly better).
If you definitely want to use a parametric regression model, there are three pretty standard models with probability as an outcome.
What you're describing is a linear probability model. Let Y be a binary dependent variable and X the vector of covariates. Then, the model has the form
$P[Y=1|X=x]=x^T\mathrm{\beta <!--\; \beta \; -->}.$
One nice thing about this type of model is that the coefficients are very easy to understand. So, for example, if x1 represents calories, then β^1 is the predicted change in Y associated with an increase of 1 calorie. It's also very easy to compute the predicted values, since it's just a product of two vectors.
The major drawback is that there's no limit on the predicted outcome. Imagine plugging in a meal with an absurd number of calories. Then, depending on whether β^1 is positive or negative, we could end up with P[Y=1|X=x]>1 or P[Y=1|X=x]<0, which shouldn't be possible for a probability.
The way to combat this issue is to use a probit or logit model. A probit model has the following form:
$P[Y=1|X=x]=\mathrm{\Phi <!--\; \Phi \; -->}(x^T\mathrm{\beta <!--\; \beta \; -->}),$
where $\mathrm{\Phi <!--\; \Phi \; -->}(\cdot <!--\; \cdot \; -->)$ is the standard normal cumulative distribution function.
A logit model has the form
$P[Y=1|X=x]=\frac{1}{1+{\exp }^{-<!--\; -\; -->x^T\mathrm{\beta <!--\; \beta \; -->}}}.$
Both the probit and logit models restrict the predicted probability to the interval [0,1], but on the flip side, the coefficients are more difficult to interpret.
The difference between probit and logit is in the assumption you're making about the distribution of the residuals - with probit, you're assuming that the residuals are normally distributed, and with logit, you're assuming they have a logistic distribution. In practice, they're usually pretty similar, and you'll likely get very similar outcomes with the two.