MCP 服务器

Dialogflow CX MCP Server

Enables AI assistants to integrate with Google Dialogflow CX for intent detection, session management, and conversational capabilities.

README

🤖 Dialogflow CX MCP Server 🚀

A powerful Model Control Protocol (MCP) server implementation for Google Dialogflow CX, enabling seamless integration between AI assistants and Google's advanced conversational platform.

💡 Pro Tip: This server bridges the gap between AI assistants and Dialogflow CX, unlocking powerful conversational capabilities!

📋 Overview

This project provides a suite of tools that allow AI assistants to interact with Dialogflow CX agents through a standardized protocol. The server handles all the complexity of managing conversations, processing intent detection, and interfacing with Google's powerful NLU systems.

✨ Key Features

🔄 Bidirectional communication with Dialogflow CX
🎯 Intent detection and matching capabilities
🎤 Audio processing for speech recognition
🔌 Webhook request/response handling
📝 Session management for persistent conversations
🔒 Secure API authentication

🔧 Requirements

Requirement	Description	Version
🐍 Python	Programming language	3.12+
☁️ Google Cloud	Project with Dialogflow CX enabled	Latest
🤖 Dialogflow CX	Conversational agent	Latest
🔑 API Credentials	Authentication for Google services	-

🚀 Installation

🐳 Using Docker

# Clone the repository
git clone https://github.com/Yash-Kavaiya/mcp-server-conversation-agents.git
cd mcp-server-conversation-agents

# Build the Docker image
docker build -t dialogflow-cx-mcp .

# Run the container
docker run -it dialogflow-cx-mcp

💻 Manual Installation

# Clone the repository
git clone https://github.com/Yash-Kavaiya/mcp-server-conversation-agents.git
cd mcp-server-conversation-agents

# Create a virtual environment (optional but recommended)
python -m venv venv
source venv/bin/activate  # On Windows: venv\Scripts\activate

# Install the package
pip install -e .

⚙️ Configuration

You'll need to provide the following configuration parameters:

Parameter	Description	Example
`dialogflowApiKey`	Your Dialogflow API key	`"abc123def456"`
`projectId`	Google Cloud project ID	`"my-dialogflow-project"`
`location`	Location of the agent	`"us-central1"`
`agentId`	ID of your Dialogflow CX agent	`"12345-abcde-67890"`

These can be set as environment variables:

export DIALOGFLOW_API_KEY=your_api_key
export PROJECT_ID=your_project_id
export LOCATION=your_location
export AGENT_ID=your_agent_id

📊 Architecture

graph TD
    A[AI Assistant] <-->|MCP Protocol| B[MCP Server]
    B <-->|Google API| C[Dialogflow CX]
    C <-->|NLU Processing| D[Intent Detection]
    C <-->|Conversation Management| E[Session Management]
    B <-->|Webhooks| F[External Services]

🛠️ Usage

The MCP server exposes the following tools for AI assistants:

🔍 initialize_dialogflow

Initialize the Dialogflow CX client with your project details.

await initialize_dialogflow(
    project_id="your-project-id",
    location="us-central1",
    agent_id="your-agent-id",
    credentials_path="/path/to/credentials.json"  # Optional
)

💬 detect_intent

Detect intent from text input.

response = await detect_intent(
    text="Hello, how can you help me?",
    session_id="user123",  # Optional
    language_code="en-US"  # Optional
)

🎤 detect_intent_from_audio

Process audio files to detect intent.

response = await detect_intent_from_audio(
    audio_file_path="/path/to/audio.wav",
    session_id="user123",  # Optional
    sample_rate_hertz=16000,  # Optional
    audio_encoding="AUDIO_ENCODING_LINEAR_16",  # Optional
    language_code="en-US"  # Optional
)

🎯 match_intent

Match intent without affecting the conversation session.

response = await match_intent(
    text="What are your hours?",
    session_id="user123",  # Optional
    language_code="en-US"  # Optional
)

🔄 Webhook Handling

Parse webhook requests and create webhook responses:

# Parse a webhook request
parsed_request = await parse_webhook_request(request_json)

# Create a webhook response
response = await create_webhook_response({
    "messages": ["Hello! How can I help you today?"],
    "parameter_updates": {"user_name": "John"}
})

🔧 Response Format

Here's an example of the response format:

<details> <summary>📋 Click to expand</summary>

{
  "messages": [
    {
      "type": "text",
      "content": "Hello! How can I help you today?"
    }
  ],
  "intent": {
    "name": "greeting",
    "confidence": 0.95
  },
  "parameters": {
    "user_name": "John"
  },
  "current_page": "Welcome Page",
  "session_id": "user123",
  "end_interaction": false
}