Custom Knowledge Bases

[Gau_Veldt]

According to the documentation:
...
It is also possible to create other kinds of knowledge bases that
determine the truth of statements in other ways. Pyke provides two of
these:
The question base which just poses the statement to an end
user as a question.
The special knowledge base which has several special one-off
knowledge entities to do different things like run a command on the
underlying system and examine its output and/or exit status.
There is only has one instance of this knowledge base --
called special.

So is it possible to make some sort of derived classes to make custom
knowledge bases (say something that imported some other module/class/
etc and was able to answer questions about its state) and
appropriately integrate with other fact, rule, question, etc, bases?

On that note it seems support for incrementally update existing
knowledge without resetting the entire works? This question seems to
be asked a lot but the answer seems that it may an intractable
feature to implement. Is this true? I can think of many real-time
scenarios where a change of plan needs to be evaluated as fast as
possible (incrementally) in reaction to unexpected new knowledge.
Unexpected meaning it can't be easily foreseen and pickled for later
since it would be akin to a chess playing machine able to enumerate
the facts of the entire game's possible moves (an intractable
complexity). We may not know to expect a change of state in a real
world parameter so it becomes troublesome even trying to factor such
state out using the universal_fact mechanism since there is no way to
know such a change is possible until it is observed in practice and it
may be incorrectly indicated as an immutable. We lost 10 years on
advance notice of ozone depletion because just such a system (a NASA/
NOAA atmosphere monitoring satellite) was unable to indicate declining
ozone level as anything other than a system fault (state was expected
to be immutable).