The Agora

Fine idea.
Given it can get extracted by script, we could simply instruct each vote should contain a list of rankings in a specific format, like ... brainstorming:

• Let's say, there are 5 options/candidates, a vote could look like
  4,5,2,1,3
  where candidate 4 would get 5 points, candidate 5 gets 4 points, and so on.
• Separators could be made flexible, "4, 5, 2, 1,3"... to make it a bit syntax-tolerant.
• The script/bot should be able to return two lists, each with username / home instance / points per candidate / original posted string; one list with positive recognition (and totals) and one for the erroneous. Could be put in spreadsheets or posted as tables.
• It would work with only two options as well.

Edit ...
The more i think about it, the more i come to the pont that this method would be elegantly simplistic. [shoulderpat] ;-)

• It's highly accessible, no fancy stuff or external tools required. All can be done with what is available in Lemmy anyway, and it's client-agnostic.
• A bot could post like three .csv lists; the two detailed ones i suggested above, and a "tally" one, that could already contain some fancy statistics, like votes per instance, or double usernames.
• All is openly accessible.

Cons:

• Perhaps not as error-resistant as a custom interface would be ... but errors can happen anyhow.
• Concealed voting is not possible.

Alerts:

• If there is no way to rank options equally, then in an equal-preference situation, the option which is listed first on the ballot will likely have an advantage. People are more likely to enter numbers in order than to swap them.
  Fix: should have a way to assign equal rank to options, like
  (4, 5), 2
  In this example, 4 and 5 would get (5+4)/2=4.5 points each (not 5 because that would give slightly more total weight to the vote), 3 points for option 2, and (2+1)/2=1.5 points to each of the remaining options 1 and 3. (And no such things are not too complex for my mind, as someone elsewhere wanted to suggest to me.)
• My suggestion here is counting all rankings, not having a winner determined by the sums of first choices, which would introduce some complexity and possible elimination of ballots: https://ballotpedia.org/Ranked-choice_voting_(RCV)
  So rather than giving the first choice an excessive gravity and counting second choices only if that didn't meet the threshold (thereby determining a majority from the rankings at the cost of ballots that didn't present a full ranking), this allows for determining even or "unconclusive" preferences. I also had the possibility of multiple selection in mind. Keep it simple.