epidemics / covid

epidemicforcasting.org visualization repository
http://epidemicforecasting.org
GNU Affero General Public License v3.0
20 stars 12 forks source link

Custom client site #389

Open lagerros opened 4 years ago

lagerros commented 4 years ago

We want to deliver a "custom site" to clients who our pro bono team are doing consulting for, where they can explore the outputs of our modelling work in a similar interface as currently available on the site.

Basically this amounts to the client getting a specific URL, similar to http://epidemicforecasting.org/?channel=david-j&selection=BR that is used to visualised a file of traces uploaded to GCS.

However, on the site we also want to display:

  1. Active infection estimates for each region in a table. (These are the numbers usually displayed in bold red at the top of the site, right above "confirmed" cases)
  2. Active infection estimates on top of page (the ones we had before, that are currently gone)
  3. Cumulative + peak estimates of number of infections (the ones we had before, that are currently gone)
  4. The options on the left side (that say "0%", "20%", ...) should be editable to reflect the actual options available for the country. This can be hardcoded for now. But we need a way of hardcoding it that doesn't interfere with the main frontpage. (The easiest way would be to just build a hidden copy of the homepage for each new client)
  5. Similarly, the text in "How the model works" should be updatable and unique for each client. This should be doable without overriding the main page.

(I think 2. and 3. are bottlenecked by resolving current issues with data format and getting a batch of new simulation runs.)

For 4. and 5., it is fine to assume for the MVP that someone from the modelling team will be able to commit text updates directly to the relevant files (so need to build custom functionality for that).

Finally, assuming that we have the prerequisite input data files using v4 format, there should be an easy way to generate new custom sites as specified above. The workflow for the whole team should look like:

  1. Request comes in from country X
  2. Pro bono modelling team analyses country X and produces an output file in v4 format of their results
  3. Pro bono modelling team creates custom site for client
hnykda commented 4 years ago

2. and 3.

I created https://github.com/epidemics/covid/issues/395, but yes, we should be fine data-wise as soon as we have a new run (the functionality is there https://github.com/epidemics/epimodel/pull/34 but I need Tomas' approval it's correct).

1.

For 1. - how should it look like? Like an actual table like:

RegionID Active
US           43213
UK           53211
...

rendered on the model page?

4.

Mathijs has some proposal on #391 already. Will that always be 4 options? Could there be more or less? Bigger flexibility is of course making it harder.

5.

And that's going to be static per client and same for all his models or different per each model?

hnykda commented 4 years ago

Also, a side note, you Jacob mentioned that the complexity is low, that's IMO a bit underestimated.

lagerros commented 4 years ago

Fair, though I meant doesn't require any larger architectural decisions etc(?)

1,2,3 sound good. 4 yes could be more or less. 5 static per client (but different between different clients).

jk3141 commented 4 years ago

In my view we do not want this and work on this should be stopped