The fact that the process, especially for a small organization using variable sources, is why I suggested a fast responding system that can use an external setpoint input. And if the setpoint is controlled by a computer then that same system can have pressure and humidity sensor inputs that the control program can use to determine the required temperature at each instant. And from a safety point of view, solenoid valves for controlling natural gas or propane safely are available already certified. So all of the parts of the system would already be suitable for that sort of application. So the software would be the only part needing to be created for the application.The temperature profile you want changes dramatically as a function of batch size, bean origin, freshness, moisture content, etc. As I said before (in the linked post,) I freely admit that useful, effective coffee roasting automation should be *theoretically* possible, however the number of variables (many of them hard to quantify in meaningful units, much less measure without very expensive tools) is pretty high.
An especially fun relationship is centered around the moment when vapor pressure in the beans finally overcomes the strength of the cell walls (there's a fun variable to quantify) which allows the trapped moisture to release in sudden burst, with evaporative cooling suddenly dropping your air temperature readings. Of course, getting that moisture out of the way also opens the door to pyrolysis, so now you're coffee roasting shifts from being an endothermic system to an exothermic system. The result of this is that, right at the moment when your PID controller sees a sudden drop in temperature and wants to apply more heat, it should actually be applying LESS heat.
If you have the resources of Folgers, Starbucks, etc. it might be worth digging into, but for anyone looking at DIY work for a small-to-mid size roasting operation, roast profiling is pure folly.
I've worked with various roast profiling systems, and none of what's publicly available even factors in the moisture content of the beans, which should be a pretty basic one. I'm sure the mega-corps of the coffee world all have their own proprietary systems, but nothing you can buy comes close to doing what you'd really need. You can save and repeat temperature profiles to your heart's content, but you'll be lucky if those settings last you more than a week before you have to create a new profile (manually.) So you might as well roast manually.
This means that the users will need to understand just what needs to be happening during the process.
And the best part is that the whole system will not be so terribly expensive, the temperature controller will be the highest cost item, still below $1000 from Automation Direct.