- Create examples directory with detailed tool usage documentation
- Add example request/response for each tool
- Update main README with links to examples
- Update .gitignore to include examples directory
- Fix tool response type compatibility with MCP SDK
- Add toolResponseToServerResult helper function
- Update server to properly convert tool responses
- Improve type safety in request handling
- Clean up error handling
- Add validation for all API configuration values
- Add bounds checking for numeric values
- Add model name validation
- Update example configuration with all available options
- Add helpful error messages for invalid configuration
- Make all Deepseek API settings configurable via MCP client config
- Add configuration validation in config.ts
- Update types to ensure type safety
- Add comprehensive example in mcp-client-config.example.json
Settings now include:
- API key and base URL
- Model selection (e.g., deepseek-reasoner)
- Max tokens (1-8192)
- Max retries
- Timeout
This change gives users more control over the server's behavior
while maintaining proper validation and defaults.
- Switch to deepseek-reasoner model for improved analysis
- Add Chain of Thought (CoT) reasoning capabilities
- Update types to support reasoning_content in responses
- Update second-opinion tool to showcase CoT reasoning
- Add max_tokens configuration for response length control
- Update documentation and examples
The deepseek-reasoner model provides step-by-step reasoning before
generating final responses, improving the quality and transparency
of the analysis. The reasoning is now exposed in tool responses
under the 'reasoning' field.
- Switch to OpenAI SDK for Deepseek API calls
- Add openai package dependency
- Update configuration for Deepseek API
- Update environment variable examples
- Improve error handling for API responses
This change improves reliability and maintainability by using the official SDK
with proper typing and error handling.
Implements the following tools:
- code-review: Code analysis with security and performance focus
- design-critique: Design feedback for UI/UX and architecture
- writing-feedback: Writing analysis with type-specific criteria
- brainstorm-enhancements: Creative improvement suggestions
Each tool includes:
- Type-safe implementation
- Comprehensive error handling
- Input validation and sanitization
- Rate limiting
- Structured prompts
- Clear documentation
Updates:
- Server implementation to support all tools
- Type definitions for all tool arguments
- Type guards for argument validation
