the 'double' attachments showing up in Gmail seems to be a bug in Gmail ... the message source appears to be correct :)

The problem is that the code makes a wrong assumption in line 291 of envelope.py (https://github.com/tomekwojcik/envelopes/blob/master/envelopes/envelope.py#L291).

    for part in self._parts:
        type_maj, type_min = part[0].split('/')
        # Small correction. See thakkarparth007's comment on:
        # https://github.com/tomekwojcik/envelopes/issues/18
        if type_maj == 'text' and type_min in ('html', 'plain'):
            msg.attach(MIMEText(part[1], type_min, self._charset))
        else:
            msg.attach(part[1])

self._parts contains the (mimetype, part, charset) tuples for the "main"
content - the text/html part and the text/plain part. Here, part is a
(unicode?) string. But, it also contains the attachments - the (mimetype, part)
tuples which are added in the add_attachment() function. Here, the part
is a MIMEBase object. Now, consider what happens when you attach a text/html
or text/plain file - you'll be appending a ('text/html',MIMEBase) tuple
to the self._parts list. Now, in the above loop, the condition is satisfied
by this text-ish file. So, you try to create a MIMEText object with the first
argument being a MIMEBase instead of a (unicode?) string. This causes all the
trouble.

A simple (but maybe hackish) solution is to add another condition and make that
if statement look something like this:

       if type_maj == 'text' and type_min in ('html', 'plain') and\
            (isinstance(part[1], str) or isinstance(part[1], unicode)):
            msg.attach(MIMEText(part[1], type_min, self._charset))
        else:
            msg.attach(part[1])

This will work (at least for now).