The first Forth I found that runs on the ESP8266 is PunyForth. I am still exploring it, but it seems to have lots of tools to work with ESP8266 capabilities.
I was having trouble figuring out how to determine how big the final code is as far as the 1K challenge is concerned, since a Forth environment is so extensible that the line between what is the system and what is user code is very indistinct. When this is completed, all the code will be included in the uploaded binary stored in with the base code uploaded with the esptool. Now that the rules are published, then the whole Forth interpreter is counted as far as memory space is concerned. So that excludes this project from a legitimate entrant to the 1K challenge, but I'll press ahead with it anyway because it may be useful to others anyway.
I suspect the source code will be under 1K, let alone the final code. And the final code includes the headers, which are only needed to build more code. Besides, PunyForth does not support headerless code (at least not yet!).
I have a couple of ESP-12E dev boards and several bare ESP-12E boards, so I intend to develop code to put onto a ESP-12E driving either a ULN2003 board or an A8266 driver. Maybe both, just because I can. If you have a preference, let me know and I'll do that first. The stepper can be whatever you want to drive using either board, with relevant power supply.