Skip to content

在Ubuntu 22.04.4 LTS使用Spring AI调用GUPtack中qwen3-embedding-8b模型报HTTP 400 - Invalid HTTP request received.异常 #3821

Open
@shenxq214

Description

@shenxq214

各位大佬好:
我在Mac机器上使用IDEA运行Spring AI 1.0.0 release版本的项目调用GUPtack中qwen3-embedding-8b模型能正常向量化,
但打包镜像部署到Ubuntu 22.04.4 LTS服务器上,调用向量化,则报HTTP 400 - Invalid HTTP request received.异常。

在部署镜像服务的Linux服务器上直接使用curl访问qwen3-embedding-8b能正常访问,可以排除网络问题。

测试方法如下:
@slf4j
@RestController
@RequestMapping("/milvusRag")
public class MilvusEmbeddingController {

@Resource
private EmbeddingModel embeddingModel;

@GetMapping("/embedding")
public Map embed(@RequestParam(value = "message", defaultValue = "Tell me a joke") String message) {
    EmbeddingResponse embeddingResponse = this.embeddingModel.embedForResponse(List.of(message));
    log.info("Embedding response embedding size: {}", embeddingResponse.getResults().get(0).getOutput().length);
    log.info("Embedding response: {}", JSON.toJSONString(embeddingResponse));
    return Map.of("embedding", embeddingResponse);
}

}

nacos配置如下:
spring:
ai:
openai:
api-key: gpustack_xxx
base-url: http://127.0.0.x/
embedding:
options:
model: qwen3-embedding-8b
chat:
options:
model: deepseek-r1-distill-qwen-32b

请教各位大佬,这种情况该如何排查和解决?

Metadata

Metadata

Assignees

No one assigned

    Type

    No type

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions