Guess in this case it may be better to just write my own server, monitor a buffer every 100ms and place updates in a Postgres table + broadcast them.
On every buffer clear, the server would query the last average + its timestamp and compute aggregations (open, close, high, low, averages) if the present timestamp compared to the last timestamp has passed the interval for a certain time window.
Forgot my math and realized that there are formulas which are not approximations for the average hahah.
new_average = average * (n - length(items)) / n + Enum.sum(items) / n
