Consider using classes instead of anonymous dictionaries

There are quite a few places in the code which deal with both mbox and source entries.
These have different structures, but in both cases the Python code uses a simple dictionary.
This can make it hard to maintain, and prevents any form of syntax checking.

It might make sense to create classes for the different database entries.
These can carry additional information with them, such as the _id

Start by locating the Python code that handles mbox and source entries, then compare the dictionary structures used for each. Determine the class boundaries and the additional information, such as _id, that each entry should carry. Done means the relevant anonymous dictionaries have been replaced consistently without changing their intended behavior.

Written by the indexing model from the issue text.