Today, Last Tweets run into another issue. It's about the encoding

d = {'msg': u'Is still rather 17\xb0 in Auckland.Brr'}
print d['msg']
# Is still rather 17° in Auckland.Brr
print pickle.dumps(d, 0)
# "(dp0\nS'msg'\np1\nVIs still rather 17\xb0 in Auckland.Brr\np2\ns."

As you can see \xb0 is not in ASCII. If you assign the pickled result to a db.TextProperty[link], you will see an track back like

pickle.dumps(d, 0).decode('ascii')
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
UnicodeDecodeError: 'ascii' codec can't decode byte 0xb0 in position 34: ordinal not in range(128)
TextProperty will try to decode with ASCII decoder if you assign a str.

A bigger character set can resolve this issue:

print pickle.dumps(d, 0).decode('latin-1')
# u"(dp0\nS'msg'\np1\nVIs still rather 17\xb0 in Auckland.Brr\np2\ns."
to_db = pickle.dumps(d, 0).decode('latin-1')
print pickle.loads(to_db.encode('latin-1'))
# {'msg': u'Is still rather 17\xb0 in Auckland.Brr'}
print pickle.loads(to_db.encode('latin-1'))['msg']
# Is still rather 17° in Auckland.Brr

An working code should look like:

model.my_text = db.Text(pickle.dumps(my_dict), encoding='latin-1')
When this model try to set my_text, it sees a type db.Text object. It won't try to decode. I think you can also give an type unicode object directly (not tested):
model.my_text = pickle.dumps(my_dict).decode('latin-1')