OpenRouter provides a unified API which should be compatible with OpenAI. I tried using tools and MCP integrations with the spring-ai-openai-spring-boot-starter to get it to work.

I am testing with the Filesystem MCP Server. The integration works great when using the openai/gpt-4o-mini model, but I am getting a warning when using google/gemini-2.5-pro-exp-03-25:free model which should also be compatible with the tools (if OpenRouter is using a unified API and the model supports it):

2025-03-26T22:04:13.586+01:00  WARN 39241 --- [ai-demo] [           main] o.s.ai.openai.OpenAiChatModel            : No choices returned for prompt: Prompt{messages=[UserMessage{content='What folders does the current active user have on the desktop?
Your response should be in JSON format.
Do not include any explanations, only provide a RFC8259 compliant JSON response following this format without deviation.
Do not include markdown code blocks in your response.
Remove the ```json markdown from the output.
Here is the JSON Schema instance your output must adhere to:
```{
  "$schema" : ";,
  "type" : "object",
  "properties" : {
    "desktopLocation" : {
      "type" : "string"
    },
    "files" : {
      "type" : "array",
      "items" : {
        "type" : "string"
      }
    }
  },
  "additionalProperties" : false
}```
', properties={messageType=USER}, messageType=USER}], modelOptions=OpenAiChatOptions: {"streamUsage":false,"model":"google/gemini-2.5-pro-exp-03-25:free","temperature":0.7}}

And the result is null. Here is how I am calling the API:

@Configuration
class Testing {
    @Bean
    fun demo(chat: ChatClient.Builder, provider: ToolCallbackProvider): ApplicationRunner {
        return ApplicationRunner {
            println(
                chat
                    .build()
                    .prompt("What folders does the current active user have on the desktop?")
                    .tools(provider)
                    .call()
                    .entity(Files::class.java)
            )
        }
    }

    data class Files(
        val desktopLocation: String,
        val files: List<String>
    )
}

And my application.properties:

spring.ai.openai.chat.options.model=google/gemini-2.5-pro-exp-03-25:free
spring.ai.openai.base-url=
spring.ai.openai.api-key=MY_TOKEN
spring.ai.mcp.client.stdio.servers-configuration=classpath:fs-config.json

If I comment out the last line in the properties file, everything works as expected, except the results are hallucinated. I thought this would work the same way as it does with the openai/gpt-4o-mini model since they should both support tools.

Is there a way to configure spring-ai to use the tools with another model other than openai/gpt-4o-mini?

-- EDIT --

After some more digging, I was able to dig this error out:

raw -> {
  "error": {
    "code": 400,
    "message": "* GenerateContentRequest.tools[0].function_declarations[10].parameters.properties: should be non-empty for OBJECT type\n",
    "status": "INVALID_ARGUMENT"
  }
}

Is this a spring-ai issue or an issue on the OpenRouter side?