Undesired Playground new operation - clearing the completion when "stop' is pressed

API Bugs
1

For the OpenAI Staff lurkers:

The Chat Completions playground now replaces all the output just received with only an “Aborted” message if the cancel button is pressed.

  • Clearing response is unwanted

One needs to inspect what went wrong. “Cancel” is for "I don’t want to pay for another 2000 tokens of that, not “obliterate the response”.

Playground should be a tool for presenting and diagnosing what is delivered over the wire, as part of implementing what was discovered to be a feasible use-case.

Even showing alternate presentation of true metadata fields of API response items or giving a non-stream option and the API product delivered by each method.

Not 
Aborted
Yes 
data:
 object
  id: str
  choices: array
    object
      delta: object
        content: str
        function_call: NoneType
        refusal: NoneType
        role: str
        tool_calls: NoneType
      finish_reason: NoneType
      index: int
      logprobs: object
        content: array
          empty
        refusal: NoneType
  created: int
  model: str
  object: str
  service_tier: NoneType
  system_fingerprint: str
  usage: NoneType