Really may be a separate project.
Allow blockd3 to be used in IPython. Should require changes to IPython's
notebook.js and cell.js and probably textcell.js
nice front-end stuff, sometimes just for documentation, sometimes for frontend/backend modification of the thing itself. maps. a target for d3 stuff. whatever.
initially, generating python code would be money... i think it would be pretty easy, and you can inject the other variables and procedures in the current kernel scope into the blockly document. adding a nice model of requests.py would make a lot of the data acquisition problems i have been experiencing trivial.
on the js side, the blockd3 stuff would be good here as well: SvgCells would be injected directly into the space, to give you a handy place from which to start.
Already wanted to do this anyway.
shameless self-promotion aside, reactive, model-based documentation would be great to add high-level insight to a notebook. real constraints that can be evaluated server-side would be mind blowing, alleviating my initial frustration with a pithy representation of bidrectional systems.
sliders and knobs, bound to variables? yes please