smallest.ai Customer Service Chatbot

A TypeScript implementation of a Model Context Protocol (MCP) for customer service chatbots using smallest.ai API.

Features

• TypeScript implementation with strong typing
• Built with Bun for fast execution
• Mock API for local development without needing API credentials
• Context management for maintaining conversation state
• Support for multiple languages and communication styles
• Error handling and graceful fallbacks

Getting Started

Prerequisites

• Bun installed on your system

Installation

1. Clone this repository or create a new project:

mkdir smallestai-mcp
cd smallestai-mcp

2. Initialize the project:

bun init -y

3. Create the project structure:

mkdir -p src/types src/config src/services src/utils

4. Copy the files from the provided implementation into the corresponding directories.

5. Install dependencies:

bun add -d typescript @types/node

Running the Project

To run the project in development mode with the mock API:

bun start

Project Structure

smallestai-mcp/
├── src/
│   ├── types/
│   │   └── index.ts        # Type definitions
│   ├── config/
│   │   └── api.ts          # API configuration
│   ├── services/
│   │   ├── smallest-ai.ts  # API service with mock implementation
│   │   ├── translation.ts  # Language adaptation service
│   │   └── storage.ts      # Session storage service
│   ├── utils/
│   │   ├── context-manager.ts   # Conversation context management
│   │   └── response-processor.ts # Response processing utilities
│   └── index.ts            # Main application entry point
├── .env                    # Environment variables (create this file)
├── package.json            # Project configuration
└── tsconfig.json           # TypeScript configuration

Configuration

Create a .env file in the root directory with your API key (not required for mock mode):

SMALLEST_AI_API_KEY=your_api_key_here

Using the Chatbot

The main entry point for the chatbot is the customerServiceChat function in src/index.ts:

const response = await customerServiceChat(
  "Hello, I'm having an issue with my recent order", 
  "user-session-123",
  "Example Corp"
);
console.log(response);

Switching to the Real API

To use the real smallest.ai API instead of the mock implementation:

1. Make sure you have a valid API key in your .env file
2. Edit src/services/smallest-ai.ts to use the real API implementation instead of the mock

Customization

Knowledge Base

Populate the knowledge base in src/services/storage.ts with your company's products, policies, and FAQs.

Communication Style

Modify the communication styles in src/services/translation.ts to match your brand voice.

Adding Features

• Add more follow-up action types in src/utils/response-processor.ts
• Implement persistent storage in src/services/storage.ts
• Add authentication and rate limiting

Production Deployment

Before deploying to production:

1. Replace the mock API with the real API implementation
2. Implement proper session storage (database)
3. Add comprehensive error handling and logging
4. Set up monitoring and alerts

License

MIT