무공해 상담봇

import requests
import time
import numpy as np

# Ollama API 설정
OLLAMA_URL = "http://localhost:11434/api/chat"

# 테스트할 모델 목록
MODELS = [
    "phi3:mini",
    "llama-3.2-Korean",
    "gemma3:4b"
]

# 시스템 프롬프트
SYSTEM_PROMPT = """당신은 상대방 말에 무조건 공감해주는 공감봇입니다.
상대방의 상황과 감정을 판단해 함께 기뻐하고, 슬퍼하고, 화내며 공감합니다.
친한 친구처럼 자연스럽고 올바른 한국어 문장으로 말하세요.
문장에는 이모지를 사용하지 않습니다.

중요 규칙:
- 반드시 JSON 형식으로만 응답해야 합니다.

출력 스키마:
{
  "emotion": "happy" | "sad" | "angry" | "soso",
  "reply": "사용자에게 보여줄 한글 한 문장"
}
"""

# 테스트 입력 문장들
TEST_INPUTS = [
    "오늘 시험이 너무 어려웠어요",
    "친구가 생일 선물을 줬어요",
    "회사에서 상사가 너무 화를 내요",
    "주말에 영화를 봤는데 재미없었어요",
    "새로운 취미를 시작했어요"
]

def test_model(model: str, test_input: str, num_runs: int = 3) -> float:
    """Measure model response time"""
    response_times = []
    
    for _ in range(num_runs):
        try:
            # Request start time
            start_time = time.time()
            
            # Call Ollama API
            payload = {
                "model": model,
                "messages": [
                    {"role": "system", "content": SYSTEM_PROMPT},
                    {"role": "user", "content": test_input}
                ],
                "stream": False,
                "format": "json",
                "options": {
                    "temperature": 0.0
                }
            }
            
            response = requests.post(OLLAMA_URL, json=payload, timeout=60)
            response.raise_for_status()
            
            # Response received time
            end_time = time.time()
            response_time = end_time - start_time
            response_times.append(response_time)
            
        except Exception as e:
            print(f"Error testing model {model}: {e}")
            response_times.append(float('inf'))
    
    # Calculate average (excluding inf)
    valid_times = [t for t in response_times if t != float('inf')]
    avg_response_time = np.mean(valid_times) if valid_times else float('inf')
    
    return avg_response_time

# Test all models
results = {}

print("Starting model tests...")
for model in MODELS:
    print(f"\nTesting {model}...")
    model_response_times = []
    
    for test_input in TEST_INPUTS:
        print(f"  Input: {test_input[:30]}...")
        response_time = test_model(model, test_input)
        model_response_times.append(response_time)
        print(f"    Response time: {response_time:.2f}s")
    
    results[model] = {
        "response_times": model_response_times,
        "avg_response_time": np.mean(model_response_times)
    }

print("\nAll tests completed!")

import matplotlib.pyplot as plt

# Visualize results as graph
fig, ax = plt.subplots(1, 1, figsize=(10, 6))

# Format model names for display
model_labels = [m.replace(":", "\n") for m in MODELS]

# Response time graph
response_times_avg = [results[m]["avg_response_time"] for m in MODELS]
colors = ['#FF6B6B', '#4ECDC4', '#45B7D1']

# Calculate max value excluding inf
valid_times = [t for t in response_times_avg if t != float('inf')]
max_valid_time = max(valid_times) if valid_times else 10

# Prepare graph data (inf displayed as 0 but handled separately)
graph_times = [t if t != float('inf') else 0 for t in response_times_avg]

bars = ax.bar(model_labels, graph_times, color=colors, alpha=0.8, edgecolor='black', linewidth=1.2)
ax.set_ylabel('Average Response Time (seconds)', fontsize=12, fontweight='bold')
ax.set_xlabel('Model', fontsize=12, fontweight='bold')
ax.set_title('Average Response Time by Model', fontsize=14, fontweight='bold', pad=20)
ax.grid(axis='y', alpha=0.3, linestyle='--')

# Display values
for i, (bar, val) in enumerate(zip(bars, response_times_avg)):
    if val != float('inf'):
        ax.text(bar.get_x() + bar.get_width()/2, bar.get_height() + max_valid_time * 0.02,
                f'{val:.2f}s', ha='center', va='bottom', fontsize=10, fontweight='bold')
    else:
        ax.text(bar.get_x() + bar.get_width()/2, max_valid_time * 0.05,
                'Failed', ha='center', va='bottom', fontsize=9, fontweight='bold', color='red')

plt.tight_layout()
plt.show()

# Print summary results
print("\n=== Measurement Results Summary ===")
print(f"{'Model':<25} {'Avg Response Time':<20}")
print("-" * 45)
for model in MODELS:
    rt = results[model]["avg_response_time"]
    rt_str = f"{rt:.2f}s" if rt != float('inf') else "Failed"
    print(f"{model:<25} {rt_str:<20}")