環境：

vllm 0.8.5

java 17

Qwen3-32B-FP8

問題描述：

JAVA請求vllm的api服務報錯Unsupported upgrade request、 Invalid HTTP request received.

WARNING:  Unsupported upgrade request.
INFO:      - "POST /v1/chat/completions HTTP/1.1" 400 Bad Request
WARNING:  Invalid HTTP request received.

服務端:

解決方案：

1.原來請求腳本

import okhttp3.*;public class VllmApiDemo (public static void main(String[] args) throws Exception {OkHttpClient client = new OkHttpClient();String json = "{"+ "\"model\": \"Qwen3-32B-FP8\","+ "\"messages\": [{\"role\": \"user\", \"content\": \"你好\")]"+ "}";RequestBody body = RequestBody.create(json, MediaType.parse("application/json"));Request request = new Request.Builder().url("http://192.168.16.2:8700/v1/chat/completions").post(body).build();Response response = client.newCall(request).execute();System.out.println(response.body().string());}
}

錯誤日志分析：

WARNING:  Unsupported upgrade request.
INFO:      - "POST /v1/chat/completions HTTP/1.1" 400 Bad Request
WARNING:  Invalid HTTP request received.

服務端提示請求格式有誤，導致 400 錯誤。 Java 代碼，有幾處明顯的語法和格式問題，可能導致請求體JSON格式不正確，服務端無法解析：

---

1.代碼語法錯誤

• 第一行 mport okhttp3.*; 應為 import okhttp3.*;
• 類定義括號用錯了，應該是 { 而不是 (
• JSON 字符串中，messages 數組結尾用了 )]，應為 }]
• JSON 字符串最后的引號關閉處格式有問題

---

2. 正確的 JSON 請求體應為

{"model": "Qwen3-32B-FP8","messages": [{"role": "user","content": "你好"}]
}

---

3. 修正后的 Java 示例代碼

import okhttp3.*;public class VllmApiDemo {public static void main(String[] args) throws Exception {OkHttpClient client = new OkHttpClient();String json = "{"+ "\"model\": \"Qwen3-32B-FP8\","+ "\"messages\": [{\"role\": \"user\", \"content\": \"你好\"}]"+ "}";RequestBody body = RequestBody.create(json, MediaType.parse("application/json"));Request request = new Request.Builder().url("http://192.168.16.2:8700/v1/chat/completions").post(body).build();Response response = client.newCall(request).execute();System.out.println(response.body().string());}
}

---

4. 重點說明

• JSON格式正確，數組和對象括號匹配
• 類定義和import語句正確
• 請求地址和模型名稱確認無誤
• 請求體內容用 MediaType.parse("application/json") 指定媒體類型

---

4. 建議調試步驟

• 你可以先用 curl 命令測試接口，確認服務端正常響應，例如：

curl -X POST http://192.168.16.2:8700/v1/chat/completions \
-H "Content-Type: application/json" \
-d '{"model":"Qwen3-32B-FP8","messages":[{"role":"user","content":"你好"}]}'

• 確認 curl 成功返回后，再運行 Java 程序。

---