Are you able to find a way to reliably split the separate documents by newline in the string blob? Replacing "}{" with "}\n{" might work depending on the input data (but is sadly likely very fragile solution).
If you manage to achieve separation then just parsing the JSON documents line by line should be extremely easy:
string_blob
|> magic_way_to_split_the_blob()
|> StringIO.open() # emulate a device that reads from a string
|> elem(1) # get the device from the {:ok, device} tuple
|> IO.stream(:line) # wrap in a Stream-friendly object
|> Stream.map(&Jason.decode/1)
#|> ...do any other filtering or transformation by using Stream functions here...
|> Enum.to_list
Which would give you a bunch of {:ok, document} or {:error, reason} tuples in a list.
