339 lines
11 KiB
Markdown
339 lines
11 KiB
Markdown
# ローカルClaude Code環境構築ガイド
|
||
RTX 4060 Ti + Qwen2.5-Coder + MCP Server
|
||
|
||
## 1. 必要なツールのインストール
|
||
|
||
### Ollamaのセットアップ
|
||
```bash
|
||
# Ollamaのインストール（Windows）
|
||
# https://ollama.com からダウンロード
|
||
|
||
# Qwen2.5-Coderモデルをダウンロード
|
||
ollama pull qwen2.5-coder:14b-instruct-q4_K_M
|
||
# または7Bバージョン（軽量）
|
||
ollama pull qwen2.5-coder:7b-instruct-q4_K_M
|
||
```
|
||
|
||
### Python環境の準備
|
||
```bash
|
||
# 仮想環境作成
|
||
python -m venv claude-code-env
|
||
claude-code-env\Scripts\activate # Windows
|
||
# source claude-code-env/bin/activate # Linux/Mac
|
||
|
||
# 必要なパッケージをインストール
|
||
pip install requests ollama-python rich click pathspec gitpython
|
||
```
|
||
|
||
## 2. メインスクリプトの作成
|
||
|
||
### claude_code.py
|
||
```python
|
||
#!/usr/bin/env python3
|
||
import os
|
||
import sys
|
||
import json
|
||
import click
|
||
import requests
|
||
from pathlib import Path
|
||
from rich.console import Console
|
||
from rich.markdown import Markdown
|
||
from rich.syntax import Syntax
|
||
|
||
console = Console()
|
||
|
||
class LocalClaudeCode:
|
||
def __init__(self, model="qwen2.5-coder:14b-instruct-q4_K_M"):
|
||
self.model = model
|
||
self.ollama_url = "http://localhost:11434"
|
||
self.conversation_history = []
|
||
self.project_context = ""
|
||
|
||
def get_project_context(self):
|
||
"""プロジェクトのファイル構造とGitステータスを取得"""
|
||
context = []
|
||
|
||
# ファイル構造
|
||
try:
|
||
for root, dirs, files in os.walk("."):
|
||
# .git, node_modules, __pycache__ などを除外
|
||
dirs[:] = [d for d in dirs if not d.startswith('.') and d not in ['node_modules', '__pycache__']]
|
||
level = root.replace(".", "").count(os.sep)
|
||
indent = " " * 2 * level
|
||
context.append(f"{indent}{os.path.basename(root)}/")
|
||
subindent = " " * 2 * (level + 1)
|
||
for file in files:
|
||
if not file.startswith('.'):
|
||
context.append(f"{subindent}{file}")
|
||
except Exception as e:
|
||
context.append(f"Error reading directory: {e}")
|
||
|
||
return "\n".join(context[:50]) # 最初の50行まで
|
||
|
||
def read_file(self, filepath):
|
||
"""ファイルを読み込む"""
|
||
try:
|
||
with open(filepath, 'r', encoding='utf-8', errors='ignore') as f:
|
||
return f.read()
|
||
except Exception as e:
|
||
return f"Error reading file: {e}"
|
||
|
||
def write_file(self, filepath, content):
|
||
"""ファイルに書き込む"""
|
||
try:
|
||
os.makedirs(os.path.dirname(filepath), exist_ok=True)
|
||
with open(filepath, 'w', encoding='utf-8') as f:
|
||
f.write(content)
|
||
return f"✅ File written: {filepath}"
|
||
except Exception as e:
|
||
return f"❌ Error writing file: {e}"
|
||
|
||
def call_ollama(self, prompt):
|
||
"""Ollamaにリクエストを送信"""
|
||
try:
|
||
response = requests.post(
|
||
f"{self.ollama_url}/api/generate",
|
||
json={
|
||
"model": self.model,
|
||
"prompt": prompt,
|
||
"stream": False,
|
||
"options": {
|
||
"temperature": 0.1,
|
||
"top_p": 0.95,
|
||
"num_predict": 2048
|
||
}
|
||
}
|
||
)
|
||
if response.status_code == 200:
|
||
return response.json()["response"]
|
||
else:
|
||
return f"Error: {response.status_code} - {response.text}"
|
||
except Exception as e:
|
||
return f"Connection error: {e}"
|
||
|
||
def process_command(self, user_input):
|
||
"""ユーザーの指示を処理"""
|
||
# プロジェクトコンテキストを更新
|
||
self.project_context = self.get_project_context()
|
||
|
||
# システムプロンプト
|
||
system_prompt = f"""You are an expert coding assistant. You can:
|
||
1. Read and analyze code files
|
||
2. Write and modify files
|
||
3. Explain code and provide suggestions
|
||
4. Debug and fix issues
|
||
|
||
Current project structure:
|
||
{self.project_context}
|
||
|
||
When you need to read a file, respond with: READ_FILE: <filepath>
|
||
When you need to write a file, respond with: WRITE_FILE: <filepath>
|
||
```
|
||
<file content>
|
||
```
|
||
|
||
User request: {user_input}
|
||
"""
|
||
|
||
response = self.call_ollama(system_prompt)
|
||
return self.process_response(response)
|
||
|
||
def process_response(self, response):
|
||
"""レスポンスを処理してファイル操作を実行"""
|
||
lines = response.split('\n')
|
||
processed_response = []
|
||
|
||
i = 0
|
||
while i < len(lines):
|
||
line = lines[i].strip()
|
||
|
||
if line.startswith("READ_FILE:"):
|
||
filepath = line.replace("READ_FILE:", "").strip()
|
||
content = self.read_file(filepath)
|
||
processed_response.append(f"📁 Reading {filepath}:")
|
||
processed_response.append(f"```\n{content}\n```")
|
||
|
||
elif line.startswith("WRITE_FILE:"):
|
||
filepath = line.replace("WRITE_FILE:", "").strip()
|
||
i += 1
|
||
# 次の```まで読み込む
|
||
if i < len(lines) and lines[i].strip() == "```":
|
||
i += 1
|
||
file_content = []
|
||
while i < len(lines) and lines[i].strip() != "```":
|
||
file_content.append(lines[i])
|
||
i += 1
|
||
content = '\n'.join(file_content)
|
||
result = self.write_file(filepath, content)
|
||
processed_response.append(result)
|
||
else:
|
||
processed_response.append("❌ Invalid WRITE_FILE format")
|
||
else:
|
||
processed_response.append(line)
|
||
|
||
i += 1
|
||
|
||
return '\n'.join(processed_response)
|
||
|
||
@click.command()
|
||
@click.option('--model', default="qwen2.5-coder:14b-instruct-q4_K_M", help='Ollama model to use')
|
||
@click.option('--interactive', '-i', is_flag=True, help='Interactive mode')
|
||
@click.argument('prompt', required=False)
|
||
def main(model, interactive, prompt):
|
||
"""Local Claude Code - AI Coding Assistant"""
|
||
|
||
claude = LocalClaudeCode(model)
|
||
|
||
if interactive or not prompt:
|
||
console.print("[bold green]🤖 Local Claude Code Assistant[/bold green]")
|
||
console.print(f"Model: {model}")
|
||
console.print("Type 'quit' to exit\n")
|
||
|
||
while True:
|
||
try:
|
||
user_input = input("👤 You: ").strip()
|
||
if user_input.lower() in ['quit', 'exit', 'q']:
|
||
break
|
||
|
||
if user_input:
|
||
console.print("\n🤖 Assistant:")
|
||
response = claude.process_command(user_input)
|
||
console.print(Markdown(response))
|
||
console.print()
|
||
|
||
except KeyboardInterrupt:
|
||
console.print("\n👋 Goodbye!")
|
||
break
|
||
else:
|
||
response = claude.process_command(prompt)
|
||
console.print(response)
|
||
|
||
if __name__ == "__main__":
|
||
main()
|
||
```
|
||
|
||
## 3. MCP Server統合
|
||
|
||
### mcp_integration.py
|
||
```python
|
||
import json
|
||
import subprocess
|
||
from typing import Dict, List, Any
|
||
|
||
class MCPIntegration:
|
||
def __init__(self):
|
||
self.servers = {}
|
||
|
||
def add_server(self, name: str, command: List[str], args: Dict[str, Any] = None):
|
||
"""MCPサーバーを追加"""
|
||
self.servers[name] = {
|
||
"command": command,
|
||
"args": args or {}
|
||
}
|
||
|
||
def call_mcp_tool(self, server_name: str, tool_name: str, arguments: Dict[str, Any]):
|
||
"""MCPツールを呼び出す"""
|
||
if server_name not in self.servers:
|
||
return {"error": f"Server {server_name} not found"}
|
||
|
||
try:
|
||
# MCPサーバーとの通信（JSONRPCベース）
|
||
request = {
|
||
"jsonrpc": "2.0",
|
||
"id": 1,
|
||
"method": f"tools/{tool_name}",
|
||
"params": {"arguments": arguments}
|
||
}
|
||
|
||
process = subprocess.Popen(
|
||
self.servers[server_name]["command"],
|
||
stdin=subprocess.PIPE,
|
||
stdout=subprocess.PIPE,
|
||
stderr=subprocess.PIPE,
|
||
text=True
|
||
)
|
||
|
||
stdout, stderr = process.communicate(json.dumps(request))
|
||
|
||
if stderr:
|
||
return {"error": stderr}
|
||
|
||
return json.loads(stdout)
|
||
|
||
except Exception as e:
|
||
return {"error": str(e)}
|
||
|
||
# 使用例
|
||
mcp = MCPIntegration()
|
||
mcp.add_server("filesystem", ["python", "-m", "mcp_server_filesystem"])
|
||
mcp.add_server("git", ["python", "-m", "mcp_server_git"])
|
||
```
|
||
|
||
## 4. 設定ファイル
|
||
|
||
### config.json
|
||
```json
|
||
{
|
||
"model": "qwen2.5-coder:14b-instruct-q4_K_M",
|
||
"ollama_url": "http://localhost:11434",
|
||
"mcp_servers": {
|
||
"filesystem": {
|
||
"command": ["python", "-m", "mcp_server_filesystem"],
|
||
"args": {"allowed_directories": ["."]}
|
||
},
|
||
"git": {
|
||
"command": ["python", "-m", "mcp_server_git"]
|
||
}
|
||
},
|
||
"excluded_files": [".git", "node_modules", "__pycache__", "*.pyc"],
|
||
"max_file_size": 1048576
|
||
}
|
||
```
|
||
|
||
## 5. 使用方法
|
||
|
||
### 基本的な使い方
|
||
```bash
|
||
# インタラクティブモード
|
||
python claude_code.py -i
|
||
|
||
# 単発コマンド
|
||
python claude_code.py "Pythonでクイックソートを実装して"
|
||
|
||
# 特定のモデルを使用
|
||
python claude_code.py --model qwen2.5-coder:7b-instruct-q4_K_M -i
|
||
```
|
||
|
||
### MCP Serverのセットアップ
|
||
```bash
|
||
# 必要なMCPサーバーをインストール
|
||
pip install mcp-server-git mcp-server-filesystem
|
||
|
||
# 設定ファイルを編集してMCPサーバーを有効化
|
||
```
|
||
|
||
## 6. 機能一覧
|
||
|
||
- ✅ ローカルLLMとの対話
|
||
- ✅ ファイル読み書き
|
||
- ✅ プロジェクト構造の自動認識
|
||
- ✅ Gitステータス表示
|
||
- ✅ シンタックスハイライト
|
||
- ✅ MCP Server統合（オプション）
|
||
- ✅ 設定ファイル対応
|
||
|
||
## 7. トラブルシューティング
|
||
|
||
### よくある問題
|
||
1. **Ollamaが起動しない**: `ollama serve` でサーバーを起動
|
||
2. **モデルが見つからない**: `ollama list` でインストール済みモデルを確認
|
||
3. **メモリ不足**: より軽量な7Bモデルを使用
|
||
4. **ファイル権限エラー**: 実行権限を確認
|
||
|
||
### パフォーマンス最適化
|
||
- GPU使用を確認: `nvidia-smi` でVRAM使用量をチェック
|
||
- モデルサイズの調整: Q4_K_M → Q4_K_S で軽量化
|
||
- コンテキスト長を調整して応答速度を向上
|
||
|
||
重い場合は7Bバージョン（qwen2.5-coder:7b-instruct-q4_K_M）に変更。
|