Since the data building tasks run asynchronously, we cannot simply forward the client's auth0 token when calling CE server. Instead we'll need to authenticate the flask API either as a separate client in auth0 (there's a client-type for server-side apps), or authenticate using non-auth0 methods (e.g. signature in the request header similar to how AWS authenticates requests).

• Validate auth0 token: if valid proceed w/ request, else fail the request with 403 or 401 HTTP status code

Flask API should validate the auth0 tokens provided to it from the client.

• Add refresh to layouts api

Layout refresh

Currently when an existing layout is fetched, we simply return it (line 25 in layouts_api.py).
Instead, we need to check if a query param refresh is set to 1 or true and re-run the layout-creation logic (including adding widgets to the widget store and enqueueing tasks to generate the data) with one critical difference: update the layout in the store vs. add it.
The way update works is that it takes the primary key (pk) as the first argument and the change dictionary as the second: in this case the only field we're changing is widgets so the call would look something like l_store.update(pk, {'widgets': w_pks}). update.

Shared code

Both of the above API's will need to use _get_query_or_default from base_api.py to get the refresh query parameter.

Gotcha's

The task queue remembers job requests for an hour, so that it can de-dupe any simultaneous requests. This means that you can only refresh the data once per hour. We can always add a hard_refresh in the future that will actually delete/overwrite the status of the existing task in the queue, but for now it's probably better to throttle widget data creation to once per hour anyway (at least in production; for development purposes this would be tedious, but you can always just delete the entire task queue collection in mongo to reset it).

• 1. Add 'date_of_publication' field on the Article namedtuple and get the data from https://www.nlm.nih.gov/bsd/mms/medlineelements.html#dp for it. Refactor its use in author_info to hydrate the new field.
• 2. Add a function to ncbi_client.py that returns a KeywordInfo namedtuple (similar to AuthorInfo) with 'pmid_to_keywords', 'keyword_to_pmids', 'pmid_to_articles'.
• 3. Implement an initial strategy for assigning a year and a count to each word. For the year assignment we'll use the earliest year it was referenced, and use the number of associated PMID's for the count. Make the strategy abstract so that we can replace it easily going forward without reorganizing the rest of the code
• 4. Transform the data from (3) into the same format as mock_data.py's TIME_SERIES_DATA

from data viz_MeSHlvl1.xlsx