# mcp_server.py
from fastmcp import FastMCP
import numpy as np
import pandas as pd
import matplotlib.pyplot as plt
# 1. MCP Server 설정 (HTTP 모드)
mcp = FastMCP(
name="WRG",
instructions="You are an AI assistant who utilizes tools that can help with game development",
)
# 2. Tool Definitions
## 3.1 Machinegun Recoil
@mcp.tool()
def machinegun_recoil_points(shots):
"""
machinegun_recoil_points(shots: int):
기관총 반동 궤적을 생성합니다.
반동이 강하고 연속적인 사격 패턴을 모사합니다.
"""
try:
shots = int(shots)
if shots < 1 or shots > 100:
raise ValueError
except:
raise ValueError("Shots must be an integer between 1 and 100")
dx_list, dy_list = [], []
for i in range(1, shots + 1):
# 초탄
if i == 1:
dx = np.random.uniform(0.0, 0.0)
dy = np.random.uniform(0.0, 0.0)
elif i <= int(shots / 3):
dx = np.random.uniform(0.1, 0.1)
dy = np.random.uniform(0.3, 0.3)
# 중탄
elif i <= int((shots / 3) * 2):
dx = np.random.normal(0.1, 0.3)
dy = np.random.normal(0.1, 0.3)
# 후탄
else:
dx = np.random.uniform(1, 0.5)
dy = np.random.normal(0.0, 0.2)
dx_list.append(dx)
dy_list.append(dy)
# Offset 계산용
offset_yaw = dx_list
offset_pitch = dy_list
# csvResult 포맷 문자열 생성
csv_result = ",".join(
[f"(OffsetPitch={round(pitch, 2)},OffsetYaw={round(yaw, 2)})"
for pitch, yaw in zip(offset_pitch, offset_yaw)]
)
return f"({csv_result})"
## 3.2 Pistol Recoil
@mcp.tool()
def pistol_recoil_points(shots):
"""
pistol_recoil_points(shots: int):
권총 반동 궤적을 생성합니다.
상대적으로 짧고 일관된 반동 패턴을 가집니다.
"""
try:
shots = int(shots)
if shots < 1 or shots > 15:
raise ValueError
except:
raise ValueError("shots must be an integer between 1 and 15")
dx_list, dy_list = [], []
for i in range(1, shots + 1):
# 초탄
if i <= int(shots/3):
dx = np.random.uniform(0.0, 0.2) # 약간의 X축 흔들림 부여
dy = np.random.uniform(0.0, 0.5) # 세로 반동 (조금 줄임)
# 중탄
elif i <= int((shots/3)*2):
dx = np.random.uniform(0.3, 0.3) # X축 미세 흔들림 증가
dy = np.random.uniform(0.3, 0.3)
# 후탄
else:
dx = np.random.uniform(-0.1, 0.3) # X축 흔들림 강함
dy = np.random.uniform(-0.1, 0.2) # Y축 거의 없음
dx_list.append(dx)
dy_list.append(dy)
# Offset 계산용
offset_yaw = dx_list
offset_pitch = dy_list
# csvResult 포맷 문자열 생성
csv_result = ",".join(
[f"(OffsetPitch={round(pitch, 2)},OffsetYaw={round(yaw, 2)})"
for pitch, yaw in zip(offset_pitch, offset_yaw)]
)
return f"({csv_result})"
## 3.3 Shotgun Recoil
@mcp.tool()
def shotgun_recoil_points(shots, pelletsPerShot=16):
"""
shotgun_recoil_points(shots: int, pellets_per_shots: int):
산탄총 반동 궤적을 생성합니다.
퍼짐이 강한 산포형 탄착군을 모사합니다.
"""
try:
shots = int(shots)
pelletsPerShot = int(pelletsPerShot)
if shots < 1 or shots > 8:
raise ValueError("shots must be an integer between 1 and 8")
if pelletsPerShot < 4 or pelletsPerShot > 50:
raise ValueError("pelletsPerShot must be an integer between 1 and 50")
except:
raise ValueError("shots and pelletsPerShot must be integers")
x = []
y = []
for _ in range(shots):
for _ in range(pelletsPerShot):
dx = np.random.normal(0.0, 2.5)
dy = np.random.normal(0.0, 2.5)
x.append(dx)
y.append(dy)
return x, y
## 3.4 Plot Recoil Pattern
@mcp.tool()
def plot_recoil_pattern(data):
"""생성된 좌표값을 시각화하는 메서드"""
import ast
# 문자열로 전달된 데이터를 파싱
if isinstance(data, str):
try:
data = ast.literal_eval(data)
except:
return "Error: Invalid data format"
# 데이터 검증
if not isinstance(data, list) or len(data) != 2:
return "Error: Data must be [x_list, y_list] format"
x_coords, y_coords = data[0], data[1]
plt.figure(figsize=(8, 8))
plt.scatter(x_coords, y_coords, c='red', s=30, marker='s', alpha=0.7)
# 궤적 연결선 추가
plt.plot(x_coords, y_coords, 'b-', alpha=0.3, linewidth=1)
# 시작점과 끝점 강조
if len(x_coords) > 0:
plt.scatter(x_coords[0], y_coords[0], c='green', s=100, marker='o', label='Start')
plt.scatter(x_coords[-1], y_coords[-1], c='black', s=100, marker='x', label='End')
plt.axhline(0, color='black', linestyle='--', linewidth=1)
plt.axvline(0, color='black', linestyle='--', linewidth=1)
plt.grid(True, linestyle=':', color='grey', alpha=0.3)
plt.gca().invert_yaxis() # y축 반전
plt.title(f"Recoil Pattern ({len(x_coords)} shots)")
plt.xlabel("Horizontal Deviation")
plt.ylabel("Vertical Recoil")
plt.legend()
# 이미지를 Base64로 인코딩하여 반환
import io
import base64
buffer = io.BytesIO()
plt.savefig(buffer, format='png', dpi=150, bbox_inches='tight')
buffer.seek(0)
image_base64 = base64.b64encode(buffer.getvalue()).decode()
plt.close() # 메모리 정리
return f"data:image/png;base64,{image_base64}"
## 3.5 Dataset Conversion
@mcp.tool()
def dataset(data):
"""생성된 좌표값을 데이터프레임화하는 메서드"""
import ast
# 문자열로 전달된 데이터를 파싱
if isinstance(data, str):
try:
data = ast.literal_eval(data)
except:
return [{"error": "Invalid data format"}]
# 데이터 검증
if not isinstance(data, list) or len(data) != 2:
return [{"error": "Data must be [x_list, y_list] format"}]
x_coords, y_coords = data[0], data[1]
# 길이 검증
if len(x_coords) != len(y_coords):
return [{"error": "X and Y coordinates length mismatch"}]
# 데이터프레임 생성 및 추가 정보 포함
records = []
for i, (x, y) in enumerate(zip(x_coords, y_coords)):
records.append({
"shot_number": i + 1,
"x_coordinate": round(x, 4),
"y_coordinate": round(y, 4),
"distance_from_center": round((x**2 + y**2)**0.5, 4),
"cumulative_recoil": round(sum(y_coords[:i+1]), 4)
})
return records
# 4. Server Start
if __name__ == "__main__":
mcp.run(transport='stdio')
