Spring-ai feat: Add responseMimeType option in vertexAiGeminiChatOptions

The Gemini model provides the responseMimeType parameter, as documented in the Gemini Model Reference.

However, when I attempting to call the Gemini model using Spring AI, there is no direct option to set this parameter. To work around this, I used reflection in my project:

        val chatModel = VertexAiGeminiChatModel(
            VertexAI(System.getenv("GEMINI_PROJECT_ID"), System.getenv("GEMINI_LOCATION"))
        )

        val generationConfig = GenerationConfig.newBuilder()
            .setResponseMimeType("application/json") // responseMimeType 을 설정할 수 없어서. 직접 접근하여 설정하였다.
            .setTemperature(0.8f)
            .build()

        val generationConfigField = VertexAiGeminiChatModel::class.java.getDeclaredField("generationConfig")
        generationConfigField.isAccessible = true
        generationConfigField.set(chatModel, generationConfig)

Therefore, it would be beneficial if the responseMimeType parameter was officially supported. The model generated by default in VertexAiGeminiChatModel is GEMINI_1_5_PRO, so this option is available from this version onwards, so it should be added as needed.

Comment From: markpollack

Thanks! Important feature indeed. I updated the docs. Merged in bc55bc7196f9d02115c32fa1d831b85d7c35e86f