LocalFoodAI_FerRo988/test_qwen_perf.py

import httpx
import time
import json

def test_performance():
    url = "http://localhost:11434/api/chat"
    payload = {
        "model": "qwen3.5:9b",
        "messages": [{"role": "user", "content": "Hi."}],
        "stream": True
    }
    
    print(f"Starting streaming test for qwen3.5:9b...")
    start_time = time.time()
    
    try:
        with httpx.Client(timeout=300.0) as client:
            with client.stream("POST", url, json=payload) as response:
                response.raise_for_status()
                print("\n--- Response ---")
                for line in response.iter_lines():
                    if line:
                        chunk = json.loads(line)
                        content = chunk.get("message", {}).get("content", "")
                        print(content, end="", flush=True)
                        if chunk.get("done"):
                            break
                print("\n----------------")
            
            end_time = time.time()
            duration = end_time - start_time
            print(f"Duration: {duration:.2f} seconds")

            
    except Exception as e:
        print(f"Error during test: {e}")

if __name__ == "__main__":
    test_performance()