I've been having a heck of a time parameterizing a potential kit here.
It starts with the problem of a kerosene lantern. This is a prop that shows up on stage in various productions. Since we don't of course want to actually set fire to lamp oil, the usual trick is flashlight bulbs and batteries. For a brighter "flame," a 12v halogen (automotive use) and a battery pack of high enough voltage to run it (such as 8xAA batteries).
The more robust solution is LEDs. At the simplest, you could, indeed, use one of the automotive-use amber LED arrays and hook that up to your 8-pack of batteries. It would last longer with a higher average output and more consistent color temperature.
Or you get a little fancier. Use a 3W RGB LED, like the Cree I've been having fun with of late. With PWM control, you now have a portable light that you can set to a selected color and intensity. And you can even flicker it.
Now, sure, you could just hard-wire a Cree, plus PWM if wanted, onto a piece of perf. I have one around I was building for an effect that got scratched. But it is a neater circuit if you have the board printed.
And even neater than that if you have a reflow oven sitting around.
Doing it this way makes for a more compact and more reliable circuit. But the downside is that you aren't soldering something to fit just a lantern. Economies of scale become economical when there is, yes, scale. The development time of a circuit board pays off higher if the same board can be used for other things other than lanterns.
And this is the first problem I'm having. What are these "other things?"
The light-up coat I made for The Wiz is very much a unique application. I've done tens of shows with a lantern in them, but only two with light-up costumes. Really, I can't think of any other common theatrical situation that I would be reaching for a plug-and-play portable light source on.
Perhaps a flexible point source for general lighting; the kind of situations where you have a doorway or other inconvenient shadow and you just want a little face light. I'm willing to believe that a little firelight in such things as campfires and stoves would also be calling for a small portable RBG source if such were available. And I can't help thinking that there must be magic wands and crystal balls that could use a light.
Because there are two other givens with the PWM circuitry that gives us RGB control and potential flicker. The first being program-ability. The second being control-ability.
It goes without saying that the portable RGB source can be easily switched on and off. But you could also dim up and down, or change color on command.
And if you an empty socket, or the right kind of header, than it also becomes remote controllable. And you are no longer dependent on an actor getting over to the prop to turn it on and off.
Here's where creeping featuritis really comes into play, though.
Assume the "board" is an ATtiny-based PWM/program generator with a couple of controller inputs (perhaps capacitance sensing to save on external hardware). Assume it switches an arbitrary load through a trio of Power Darlingtons (or similar) and solder tabs or screw terminals. This detaches the LED/load itself so the circuit can be hidden in the base of a lamp or whatever. Constant-current drivers would be better for LEDs but would have to be matched; this allows us to re-purpose for relays or other tasks.
The board can easily power-regulate from a 3v to 12v source, so a 3-pack or 4-pack of rechargeable batteries is good enough. But a lipo is sexier; high density rechargeable battery built in, with charging circuit and charge indicators, so all you have to do is plug it into a USB charger (or similar) between shows.
(The main downside to the lipo is if you have back-to-back shows with heavy use of the circuit. Using swappable batteries means you can put the device back in service without having to wait on a charge.)
The bigger problem is program-ability. For me, I'm fine writing new code as needed and feeding it through an ISP port. But it might be easier for the general user -- heck, it would be easier for me too -- if you could adjust the behavior on the fly via nothing more than a USB cable. Better yet, through a USB cable -- or remotely through a radio link -- via a GUI that dealt with most of the details of selecting colors and setting up switches and so forth at a higher level of abstraction than typing fresh code.
I think at some point, you have to accept that one "board" shouldn't try to be too generalist. Perhaps it makes most sense to design it as if it will always be PWM'ing three channels of LED, with a preset of several hard-coded behaviors selected by resistor ladder and/or transmitted commands. And re-purpose that hardware with scratch-written code as unique applications arise.
And to ignore such fun ideas as lipo charging circuits, boost converters, constant-current drivers, and so forth. And restrict the immediate flexibility to setting jumpers on the PCB, and the load and source that get attached to the screw terminals.
Maybe that is a build-able circuit. Maybe that's enough to boot up Eagle and see what it might look like...