Validating batch response with Google Search grounding against `GenerateContentResponse` causes validation error

[hakontonne]

It seems like the GenerateContentResponse objects and its nested objects are not designed properly for the any reponses that uses the built-in Google Search tool.

It returns:

response.candidates.0.groundingMetadata.groundingChunks.0.chunkMetadata
  Extra inputs are not permitted [type=extra_forbidden, input_value={'query': 'query'}, input_type=dict]
    For further information visit https://errors.pydantic.dev/2.12/v/extra_forbidden

This is caused by the GroundingChunk object, which does not actually match the groundingChunk returned by the Vertex AI API:

class GroundingMetadata(_common.BaseModel):
  """Information for various kinds of grounding."""

  grounding_chunks: Optional[list[GroundingChunk]] = Field(
      default=None,
      description="""A list of supporting references retrieved from the grounding
      source. This field is populated when the grounding source is Google
      Search, Vertex AI Search, or Google Maps.
      """,
  )
# Truncated for brevity

class GroundingChunk(_common.BaseModel):
  """A piece of evidence that supports a claim made by the model.

  This is used to show a citation for a claim made by the model. When grounding
  is enabled, the model returns a `GroundingChunk` that contains a reference to
  the source of the information.
  """

  image: Optional[GroundingChunkImage] = Field(
      default=None,
      description="""A grounding chunk from an image search result. See the `Image` message for details.""",
  )
  maps: Optional[GroundingChunkMaps] = Field(
      default=None,
      description="""A `Maps` chunk is a piece of evidence that comes from Google Maps.

      It contains information about a place, such as its name, address, and
      reviews. This is used to provide the user with rich, location-based
      information.""",
  )
  retrieved_context: Optional[GroundingChunkRetrievedContext] = Field(
      default=None,
      description="""A grounding chunk from a data source retrieved by a retrieval tool, such as Vertex AI Search. See the `RetrievedContext` message for details""",
  )
  web: Optional[GroundingChunkWeb] = Field(
      default=None,
      description="""A grounding chunk from a web page, typically from Google Search. See the `Web` message for details.""",
  )

The actual JSON groundingChunk JSON:

"groundingMetadata" : {
        "groundingChunks" : [ {
          "chunkMetadata" : {
            "query" : "dummy_query"
          },
          "web" : {
            "title" : "youtube.com",
            "uri" : "https://vertexaisearch.cloud.google.com/grounding-api-redirect/id"
          }
        } ],

Environment details

• Programming language: Python
• OS: MacOS
• Language runtime version: 3.12
• Package version: 2.6.0

Steps to reproduce

1. Create batch request or requests with google search enabled and ensure the model actually performs a Google search , dummy example:

   {
       "contents" : [ {
         "parts" : [ {
           "text" : "Dummy request"
         } ],
         "role" : "user"
       } ],
       "generationConfig" : {
         "responseMimeType" : "application/json",
         "responseSchema" : {
         },
         "temperature" : 0.2
       },
       "systemInstruction" : {
         "parts" : [ {
           "text" : "Dummy system instructions "
         } ]
       },
       "tools" : [ {
         "googleSearch" : {
           "excludeDomains" : [ "example.com" ] // google search on batch inference. does not work unless you actually specify an exclusion domain... please fix as well
         }
       } ]
     }

2. After results are ready, read the results and validate the GenerationResponse object againts the "response" part of the JSON response:

 from google.genai import types
 result = types.GenerateContentResponse.model_validate_json(parsed_json_dict["response"]) # Raises validation error