[BUG] AI Proxy plugin returns Spring AI ChatResponse instead of OpenAI Chat Completions format

[eye-gu]

Is there an existing issue for this?

• I have searched the existing issues

Current Behavior

The ai-proxy plugin currently uses Spring AI's ChatClient API to call upstream LLM providers. This causes the gateway to return Spring AI's internal ChatResponse object structure instead of the standard OpenAI Chat Completions API format.

Clients expecting an OpenAI-compatible API (e.g., ChatCompletion / ChatCompletionChunk) receive a non-standard response and cannot parse it correctly.

Additionally, Spring AI's createRequest() reconstructs ChatCompletionMessage from AssistantMessage, which loses fields such as reasoning_content, refusal, and annotations from the original request.

Expected Behavior

No response

Steps To Reproduce

No response

Environment

ShenYu version(s):

Debug logs

No response

Anything else?

by the way, Do shenyu need to support other ai protocols, such as Anthropic? Do shenyu need to convert different protocols, for example, if the client is the OpenAI Chat Completions protocol and the provider is Anthropic?

[github-actions]

👋 Hello @yuluo-yx

A new issue has been opened by @eye-gu:

• Title: [BUG] AI Proxy plugin returns Spring AI ChatResponse instead of OpenAI Chat Completions format
• Link: [BUG] AI Proxy plugin returns Spring AI ChatResponse instead of OpenAI Chat Completions format #6340

Please review when you have time. Thank you! 🙏