[Bug]:Anthropic SDK throws AnthropicInvalidDataException when using proxy services that strip the `id` field

[hubigsea]

Describe the bug

When using AnthropicChatModel with a proxy service (anthropic proxy api) that forwards Anthropic API responses, the SDK throws AnthropicInvalidDataException: id is not set.

This issue occurs because some proxy services strip the id field from the response body for privacy reasons.

To Reproduce

Steps

1. Configure AnthropicChatModel with a proxy URL:

AnthropicChatModel model = AnthropicChatModel.builder()
    .apiKey("sk-ant-...")  // Your API key
    .modelName("claude-sonnet-4-5-20250929")
    .baseUrl("Anthropic Proxy URL")  // Proxy URL
    .stream(false)
    .build();

ReActAgent agent = ReActAgent.builder()
    .name("TestAgent")
    .sysPrompt("You are a math assistant.")
    .model(model)
    .build();

Msg response = agent.call(Msg.builder()
    .textContent("2+3等于多少？")
    .build()).block();

2. Execute the code

3. See error: AnthropicInvalidDataException: id is not set

Curl Command (Alternative Reproduction)

curl -X POST "https://anthropicProxyApi/v1/messages" \
  -H "Content-Type: application/json" \
  -H "x-api-key: sk-ant-YOUR_API_KEY" \
  -H "anthropic-version: 2023-06-01" \
  -H "anthropic-dangerous-direct-browser-access: true" \
  -d '{
    "model": "claude-sonnet-4-5-20250929",
    "max_tokens": 1024,
    "messages": [
      {
        "role": "user",
        "content": "2+3等于多少？"
      }
    ]
  }'

Expected behavior

The SDK should handle responses that are missing the id field gracefully. The id field is important for tracking but not critical for response parsing.

Error messages

com.anthropic.errors.AnthropicInvalidDataException: `id` is not set
    at com.anthropic.core.JsonField.getRequired$anthropic_java_core(Values.kt:174)
    at com.anthropic.models.messages.Message.id(Message.kt:68)
    at io.agentscope.core.formatter.anthropic.AnthropicResponseParser.parseMessage(AnthropicResponseParser.java:97)
    at io.agentscope.core.formatter.anthropic.AnthropicChatFormatter.parseResponse(AnthropicChatFormatter.java:47)
    at io.agentscope.core.model.AnthropicChatModel.lambda$doStream$1(AnthropicChatModel.java:194)

Response from Proxy (missing id field)

{
  "type": "message",
  "role": "assistant",
  "content": [
    {
      "type": "text",
      "text": "2 + 3 = 5"
    }
  ],
  "model": "claude-sonnet-4-5-20250929",
  "stop_reason": "end_turn",
  "stop_sequence": null,
  "usage": {
    "input_tokens": 57,
    "output_tokens": 8
  }
}

Environment

• AgentScope-Java Version: 1.0.7
• Java Version: 17+
• OS: macOS

Additional context

Root Cause

The Anthropic SDK 2.11.0 requires the id field in the message response and validates it with @Required annotation during deserialization.

Affected Components

• AnthropicChatModel (io.agentscope.core.model.AnthropicChatModel)
• AnthropicResponseParser (io.agentscope.core.formatter.anthropic.AnthropicResponseParser)

Behavior Comparison

Scenario	Current Behavior	Expected Behavior
Proxy without id field	❌ Throws exception	✅ Should handle gracefully
Official Anthropic API	❓ Not tested yet	✅ Should work

Suggested Fixes

1. Fix in Anthropic SDK (Recommended): Make the id field optional in the SDK's Message class instead of required.

2. Fix in AgentScope Response Parser: Modify AnthropicResponseParser.parseMessage() to use a custom ObjectMapper that can deserialize responses without the id field.

3. Proxy-Friendly Configuration: Add a configuration option to AnthropicChatModel.Builder to generate fallback IDs for responses missing the id field.

Note

This issue is specific to proxy services that modify Anthropic API responses. The official Anthropic API always includes the id field.

[Alexxigang]

I traced this to AnthropicResponseParser: both the non-streaming path and the message-start streaming path call message.id() directly, so proxy responses that omit id fail before we can build a ChatResponse. I'm working on a fix that reads the SDK's raw _id() field safely, falls back when it is missing, and adds regression tests for both parse paths.

[Alexxigang]

Opened a fix here: #1156

It makes the Anthropic parser tolerate proxy responses that omit the top-level id field, reuses the SDK raw _id() value when present, and adds parser/formatter regression coverage for non-streaming and message-start streaming paths.