- Repository URL consistency: Updated all references to BeehiveInnovations/zen-mcp-server format - Documentation clarity: Fixed misleading table headers and improved Docker configuration examples - File conventions: Added missing final newlines to all files - Configuration consistency: Clarified API key placeholder format in documentation Addresses all points raised in PR #17 review by Gemini Code Assist. 🤖 Generated with Claude Code Co-Authored-By: Claude <noreply@anthropic.com>
15 KiB
Code Style Guide
Overview
This document establishes coding standards and style guidelines for the Gemini MCP Server project. Consistent code style improves readability, maintainability, and collaboration efficiency.
Python Style Guidelines
PEP 8 Compliance
Base Standard: Follow PEP 8 as the foundation for all Python code.
Automated Formatting: Use Black formatter with default settings:
black tools/ utils/ tests/ --line-length 88
Line Length: 88 characters (Black default)
# Good
result = some_function_with_long_name(
parameter_one, parameter_two, parameter_three
)
# Avoid
result = some_function_with_long_name(parameter_one, parameter_two, parameter_three)
Import Organization
Import Order (enforced by isort):
# 1. Standard library imports
import asyncio
import json
import os
from pathlib import Path
from typing import Dict, List, Optional
# 2. Third-party imports
import redis
from pydantic import BaseModel
# 3. Local application imports
from tools.base import BaseTool
from utils.file_utils import validate_file_path
Import Formatting:
# Good - Explicit imports
from typing import Dict, List, Optional
from utils.conversation_memory import ThreadContext, ConversationMemory
# Avoid - Wildcard imports
from utils.conversation_memory import *
Naming Conventions
Functions and Variables: snake_case
def process_file_content(file_path: str) -> str:
context_tokens = calculate_token_count(content)
return formatted_content
Classes: PascalCase
class GeminiClient:
pass
class ThreadContext:
pass
Constants: UPPER_SNAKE_CASE
MAX_CONTEXT_TOKENS = 1000000
THINKING_MODE_TOKENS = {
'minimal': 128,
'low': 2048,
'medium': 8192
}
Private Methods: Leading underscore
class ToolBase:
def execute(self):
return self._process_internal_logic()
def _process_internal_logic(self):
# Private implementation
pass
Type Hints
Mandatory Type Hints
Function Signatures: Always include type hints
# Good
def validate_file_path(file_path: str) -> bool:
return os.path.exists(file_path)
async def process_request(request: dict) -> ToolOutput:
# Implementation
pass
# Avoid
def validate_file_path(file_path):
return os.path.exists(file_path)
Complex Types: Use typing module
from typing import Dict, List, Optional, Union, Any
def process_files(files: List[str]) -> Dict[str, Any]:
return {"processed": files}
def get_config(key: str) -> Optional[str]:
return os.getenv(key)
Generic Types: Use TypeVar for reusable generics
from typing import TypeVar, Generic
T = TypeVar('T')
class Repository(Generic[T]):
def get(self, id: str) -> Optional[T]:
# Implementation
pass
Documentation Standards
Docstring Format
Use Google Style docstrings:
def execute_tool(name: str, arguments: dict, context: Optional[str] = None) -> ToolOutput:
"""Execute a tool with given arguments and context.
Args:
name: The name of the tool to execute
arguments: Tool-specific parameters and configuration
context: Optional conversation context for threading
Returns:
ToolOutput containing the execution result and metadata
Raises:
ToolNotFoundError: If the specified tool doesn't exist
ValidationError: If arguments don't match tool schema
Example:
>>> output = execute_tool("chat", {"prompt": "Hello"})
>>> print(output.content)
"Hello! How can I help you today?"
"""
# Implementation
pass
Class Documentation:
class ConversationMemory:
"""Manages conversation threading and context persistence.
This class handles storing and retrieving conversation contexts
using Redis as the backend storage. It supports thread-based
organization and automatic cleanup of expired conversations.
Attributes:
redis_client: Redis connection for storage operations
default_ttl: Default time-to-live for conversation threads
Example:
>>> memory = ConversationMemory("redis://localhost:6379")
>>> context = ThreadContext("thread-123")
>>> await memory.store_thread(context)
"""
def __init__(self, redis_url: str, default_ttl: int = 86400):
"""Initialize conversation memory with Redis connection.
Args:
redis_url: Redis connection string
default_ttl: Default TTL in seconds (default: 24 hours)
"""
pass
Inline Comments
When to Comment:
# Good - Explain complex business logic
def calculate_token_budget(files: List[str], total_budget: int) -> Dict[str, int]:
# Priority 1 files (source code) get 60% of budget
priority_1_budget = int(total_budget * 0.6)
# Group files by priority based on extension
priority_groups = defaultdict(list)
for file in files:
ext = Path(file).suffix.lower()
priority = FILE_PRIORITIES.get(ext, 4)
priority_groups[priority].append(file)
return allocate_budget_by_priority(priority_groups, total_budget)
# Avoid - Stating the obvious
def get_file_size(file_path: str) -> int:
# Get the size of the file
return os.path.getsize(file_path)
Security and Performance Notes:
def validate_file_path(file_path: str) -> bool:
# Security: Prevent directory traversal attacks
if '..' in file_path or file_path.startswith('/etc/'):
return False
# Performance: Early return for non-existent files
if not os.path.exists(file_path):
return False
return True
Error Handling
Exception Handling Patterns
Specific Exceptions:
# Good - Specific exception handling
try:
with open(file_path, 'r') as f:
content = f.read()
except FileNotFoundError:
logger.warning(f"File not found: {file_path}")
return None
except PermissionError:
logger.error(f"Permission denied: {file_path}")
raise SecurityError(f"Access denied to {file_path}")
except UnicodeDecodeError:
logger.warning(f"Encoding error in {file_path}")
return f"Error: Cannot decode file {file_path}"
# Avoid - Bare except clauses
try:
content = f.read()
except:
return None
Custom Exceptions:
class GeminiMCPError(Exception):
"""Base exception for Gemini MCP Server errors."""
pass
class ToolNotFoundError(GeminiMCPError):
"""Raised when a requested tool is not found."""
pass
class ValidationError(GeminiMCPError):
"""Raised when input validation fails."""
pass
Logging Standards
Logging Levels:
import logging
logger = logging.getLogger(__name__)
# DEBUG: Detailed diagnostic information
logger.debug(f"Processing file: {file_path}, size: {file_size}")
# INFO: General operational information
logger.info(f"Tool '{tool_name}' executed successfully")
# WARNING: Something unexpected but recoverable
logger.warning(f"File {file_path} exceeds recommended size limit")
# ERROR: Error condition but application continues
logger.error(f"Failed to process file {file_path}: {str(e)}")
# CRITICAL: Serious error, application may not continue
logger.critical(f"Redis connection failed: {connection_error}")
Structured Logging:
# Good - Structured logging with context
logger.info(
"Tool execution completed",
extra={
"tool_name": tool_name,
"execution_time": execution_time,
"files_processed": len(files),
"thinking_mode": thinking_mode
}
)
# Avoid - Unstructured string formatting
logger.info(f"Tool {tool_name} took {execution_time}s to process {len(files)} files")
Async/Await Patterns
Async Function Design
Async When Needed:
# Good - I/O operations should be async
async def fetch_gemini_response(prompt: str) -> str:
async with aiohttp.ClientSession() as session:
async with session.post(GEMINI_API_URL, json=payload) as response:
return await response.text()
# Good - CPU-bound work remains sync
def parse_stack_trace(trace_text: str) -> List[StackFrame]:
# CPU-intensive parsing logic
return parsed_frames
Async Context Managers:
class AsyncRedisClient:
async def __aenter__(self):
self.connection = await redis.connect(self.url)
return self
async def __aexit__(self, exc_type, exc_val, exc_tb):
await self.connection.close()
# Usage
async with AsyncRedisClient(redis_url) as client:
await client.store_data(key, value)
Security Best Practices
Input Validation
Path Validation:
def validate_file_path(file_path: str) -> bool:
"""Validate file path for security and accessibility."""
# Convert to absolute path
abs_path = os.path.abspath(file_path)
# Check for directory traversal
if not abs_path.startswith(PROJECT_ROOT):
raise SecurityError(f"Path outside project root: {abs_path}")
# Check for dangerous patterns
dangerous_patterns = ['../', '~/', '/etc/', '/var/']
if any(pattern in file_path for pattern in dangerous_patterns):
raise SecurityError(f"Dangerous path pattern detected: {file_path}")
return True
Data Sanitization:
def sanitize_user_input(user_input: str) -> str:
"""Sanitize user input to prevent injection attacks."""
# Remove null bytes
sanitized = user_input.replace('\x00', '')
# Limit length
if len(sanitized) > MAX_INPUT_LENGTH:
sanitized = sanitized[:MAX_INPUT_LENGTH]
# Remove control characters
sanitized = ''.join(char for char in sanitized if ord(char) >= 32)
return sanitized
Secret Management
Environment Variables:
# Good - Environment variable with validation
GEMINI_API_KEY = os.getenv('GEMINI_API_KEY')
if not GEMINI_API_KEY:
raise ConfigurationError("GEMINI_API_KEY environment variable required")
# Avoid - Hardcoded secrets
API_KEY = "sk-1234567890abcdef" # Never do this
Secret Logging Prevention:
def log_request_safely(request_data: dict) -> None:
"""Log request data while excluding sensitive fields."""
safe_data = request_data.copy()
# Remove sensitive fields
sensitive_fields = ['api_key', 'token', 'password', 'secret']
for field in sensitive_fields:
if field in safe_data:
safe_data[field] = '[REDACTED]'
logger.info(f"Processing request: {safe_data}")
Performance Guidelines
Memory Management
Generator Usage:
# Good - Memory efficient for large datasets
def process_large_file(file_path: str) -> Generator[str, None, None]:
with open(file_path, 'r') as f:
for line in f:
yield process_line(line)
# Avoid - Loading entire file into memory
def process_large_file(file_path: str) -> List[str]:
with open(file_path, 'r') as f:
return [process_line(line) for line in f.readlines()]
Context Managers:
# Good - Automatic resource cleanup
class FileProcessor:
def __enter__(self):
self.temp_files = []
return self
def __exit__(self, exc_type, exc_val, exc_tb):
# Cleanup temporary files
for temp_file in self.temp_files:
os.unlink(temp_file)
Caching Patterns
LRU Cache for Expensive Operations:
from functools import lru_cache
@lru_cache(maxsize=128)
def parse_file_content(file_path: str, file_hash: str) -> str:
"""Parse file content with caching based on file hash."""
with open(file_path, 'r') as f:
return expensive_parsing_operation(f.read())
Testing Standards
Test File Organization
Test Structure:
# tests/test_tools.py
import pytest
from unittest.mock import Mock, patch
from tools.chat import ChatTool
from tools.models import ToolOutput
class TestChatTool:
"""Test suite for ChatTool functionality."""
def setup_method(self):
"""Set up test fixtures before each test method."""
self.chat_tool = ChatTool()
self.mock_gemini_client = Mock()
def test_basic_chat_execution(self):
"""Test basic chat tool execution with simple prompt."""
# Arrange
request = {"prompt": "Hello"}
# Act
result = self.chat_tool.execute(request)
# Assert
assert isinstance(result, ToolOutput)
assert result.status == "success"
@patch('tools.chat.GeminiClient')
def test_chat_with_mocked_api(self, mock_client):
"""Test chat tool with mocked Gemini API responses."""
# Test implementation
pass
Test Naming Conventions
Test Method Names:
def test_should_validate_file_path_when_path_is_safe():
"""Test that safe file paths are correctly validated."""
pass
def test_should_raise_security_error_when_path_contains_traversal():
"""Test that directory traversal attempts raise SecurityError."""
pass
def test_should_return_none_when_file_not_found():
"""Test that missing files return None gracefully."""
pass
Configuration Management
Environment-Based Configuration
Configuration Class:
class Config:
"""Application configuration with validation."""
def __init__(self):
self.gemini_api_key = self._require_env('GEMINI_API_KEY')
self.redis_url = os.getenv('REDIS_URL', 'redis://localhost:6379')
self.project_root = os.getenv('PROJECT_ROOT', '/workspace')
self.max_context_tokens = int(os.getenv('MAX_CONTEXT_TOKENS', '1000000'))
# Validate configuration
self._validate_configuration()
def _require_env(self, key: str) -> str:
"""Require environment variable or raise error."""
value = os.getenv(key)
if not value:
raise ConfigurationError(f"Required environment variable: {key}")
return value
def _validate_configuration(self) -> None:
"""Validate configuration values."""
if not os.path.exists(self.project_root):
raise ConfigurationError(f"PROJECT_ROOT not found: {self.project_root}")
Pre-commit Hooks
Automated Quality Checks
Required Tools:
# Install development dependencies
pip install black isort flake8 mypy pytest
# Format code
black tools/ utils/ tests/
isort tools/ utils/ tests/
# Check code quality
flake8 tools/ utils/ tests/
mypy tools/ utils/
# Run tests
pytest tests/ -v --cov=tools --cov=utils
Pre-commit Configuration (.pre-commit-config.yaml):
repos:
- repo: https://github.com/psf/black
rev: 23.3.0
hooks:
- id: black
language_version: python3.9
- repo: https://github.com/pycqa/isort
rev: 5.12.0
hooks:
- id: isort
- repo: https://github.com/pycqa/flake8
rev: 6.0.0
hooks:
- id: flake8
- repo: https://github.com/pre-commit/mirrors-mypy
rev: v1.3.0
hooks:
- id: mypy
Following these code style guidelines ensures consistent, maintainable, and secure code across the Gemini MCP Server project. All team members should adhere to these standards and use the automated tools to enforce compliance.