About Fine-tuning ChatGPT format

ChatGPT
,
1

When I have multiple systems with similar features, how should I write my JSONL? Should I add a special tag object in the JSON to help the AI tell them apart?

For example, in the two JSONs below, I added product_name and section objects so that when the user asks the same question, the AI can use these categories to give the right response.

Is this approach right, or does fine-tuning already have a set format to handle this kind of issue?

{
 "input": {
   "messages": [
     {
       "role": "user",
       "content": "What kind of file does the export button create?"
     }
   ],
   "context": {
     "product_name": "Accounting System",
     "section": "Income and Expenditure Report"
   },
   "tools": [],
   "parallel_tool_calls": true
 },
 "preferred_output": [
   {
     "role": "assistant",
     "content": "The export format is EXCEL, and separate sheets will be created for each company \n (if you are exporting data for multiple companies at the same time)."   
   }
 ],
 "non_preferred_output": [
   {
     "role": "assistant",
     "content": "The export format is EXCEL"
   }
 ]
}

{
 "input": {
   "messages": [
     {
       "role": "user",
       "content": "What kind of file does the export button create?"
     }
   ],
   "context": {
     "product_name": "Operation System",
     "section": "Counter Payment Daily Report"
   },
   "tools": [],
   "parallel_tool_calls": true
 },
 "preferred_output": [
   {
     "role": "assistant",
     "content": "The exported file is in TXT format, and only the logged-in user's own account data can be queried and exported."   
   }
 ],
 "non_preferred_output": [
   {
     "role": "assistant",
     "content": "The export format is TXT"
   }
 ]
}