Add food handler in Bot

[bobheadxi]

Add food handler in Bot

[bwdmonkey]

@bobheadxi Are we actually gonna use some sort of ml with training data or just base it off one serving size?

Note: we still haven't decided on a dataset for food data.

[bobheadxi]

I was thinking about trying to ML keyword recognition by doing things like:

Bob: Hi I ate a cookie today
Calories: You ate a 'cookie' today? (Yes) (No)
Bob: Yes
Calories: Does 100 calories sound right? (Less) (Yes) (More)

Handy mechanisms like Quick Reply thanks to @ayaMakh's PR the other day ( #30 😉 ) will let us ask for feedback effectively like this. Obviously still have a lot to look into because none of us know anything, but an exciting opportunity.

As for food/exercise/data in general... we need to discuss that at some point. 🤔 I'm still in favour of using some online APIs, but if we can crudely machine learn everything that would be sick.

This will definitely be after we get some quality QA going though ( #31 which Aya is working on starting and we'll all help when she gets the ball rolling and #32, which I'll get going ), since we don't want to have to worry about random bugs when we begin hashing out our features. 🙅‍♂️

After that I will look into setting up some kind of POS (parts-of-speech) recognition ( #24 ) that will help us with all this - been using nltk for Python which is clean, hopefully an API for it or something else exists for Go. This should make it easier for us to work on natural language messages.

❗️ QA first, however ❗️Soz guys but once we reach a certain point we're just never going to get QA going, and it is essential that we do if we want to see this through to eventual Facebook approval + release for everyone to use 🎁

💬 Paging @ayaMakh @zhulia960821 because this is important discussion 💬

[bwdmonkey]

For Food Nutrient API: https://ndb.nal.usda.gov/ndb/doc/apilist/API-FOOD-REPORT.md

For exercise calories API: https://wger.de/en/software/api

I think these should be good enough for the API portion.

[bobheadxi]

I think we should use APIs to get initial "suggestions", but maintain our own database of likely values based on user input.

For example, the food handler could check our database, and if nothing is there, request for data from the APIs. Both the APIs you linked look great, just gotta figure out how best to use them now. I will update the issue descriptions with them.