Support for the promptFeedback.blockReason in the Gemini streaming output

[xukecheng]

Description

During the use of Gemini API, it is possible to encounter a return blockReason.

{
  "promptFeedback": {
    "blockReason": "PROHIBITED_CONTENT"
  },
  "usageMetadata": {
    "promptTokenCount": 12559,
    "totalTokenCount": 12559
  },
  "modelVersion": "gemini-1.5-flash-002"
}

However, it currently seems that there is no way to obtain the blockReason in the streaming output.

Code example

No response

AI provider

https://esm.sh/[email protected]

Additional context

No response

[lgrammel]

Duplicates #4236