No serializer found for class software.amazon.awssdk.services.bedrockruntime.model.TokenUsage

[robert-niestroj]

Bug description
After upgrade from Spring AI 1.1.2 to 1.1.3 following Exception is thrown:

com.fasterxml.jackson.databind.exc.InvalidDefinitionException: No serializer found for class software.amazon.awssdk.services.bedrockruntime.model.TokenUsage and no properties discovered to create BeanSerializer (to avoid exception, disable SerializationFeature.FAIL_ON_EMPTY_BEANS) (through reference chain: org.springframework.ai.chat.model.ChatResponse["metadata"]->org.springframework.ai.chat.metadata.ChatResponseMetadata["usage"]->org.springframework.ai.chat.metadata.DefaultUsage["nativeUsage"])
	at com.fasterxml.jackson.databind.exc.InvalidDefinitionException.from(InvalidDefinitionException.java:77) ``` 


in this code:

```java
  @GetMapping("/hello-response")
  public ChatResponse helloResponse(@RequestParam String prompt) {
    String promptFromDb = "Plan me a {days}-day trip to {city}";
    return chatClient
        .prompt()
        .user(u -> u.text(promptFromDb).param("days", 3).param("city", prompt))
        .call()
        .chatResponse();
  }

Environment
Java 25
Spring Boot 3.5
Spring AI 1.1.3
Amazon Bedrock Converse

Steps to reproduce
Execute above code.

Expected behavior
No exception is thrown like in Spring AI 1.1.2

[ZhangDT-sky]

Hi, I was able to reproduce this locally.

It looks like the problem is that Bedrock Converse puts the AWS SDK TokenUsage object into DefaultUsage.nativeUsage, and Jackson then fails to serialize it when ChatResponse is returned directly from a controller.

I opened a PR with a possible fix and a test:
PR: #5690