Txtai Assistant MCP

🚀 TxtAI Assistant MCP Project

TxtAI Assistant MCP is a Memory Cognitive Processor (MCP) based on the txtai library, used for semantic search and memory management. This project aims to enhance the context awareness of dialogue systems by storing, retrieving, and managing information.

🚀 Quick Start

TxtAI Assistant MCP is a Memory Cognitive Processor (MCP) implemented based on the txtai library, which is used for semantic search and memory management. The project aims to enhance the context awareness of dialogue systems by storing, retrieving, and managing information.

✨ Features

Introduction to TxtAI

TxtAI is a powerful text processing library that supports the following core functions:

• Semantic Search: Text retrieval based on similarity.
• Neural Search: Text matching using transformer models.
• Summary Generation: Extracting key information from long texts.
• Question - Answering System: Answering context - related questions.

Project Features

• Semantic Search: Retrieve relevant memories through similarity calculation.
• Persistent Storage: Save memories and label indexes in the file system.
• Label Organization: Support efficient retrieval based on labels.
• Metadata Support: Allow attaching custom information.
• Multithreaded Processing: Optimize performance and resource utilization.
• Security Mechanism: Prevent directory traversal attacks and ensure input validation.

📦 Installation

1. Clone the project repository:

   git clone https://github.com/yourusername/txtai-assistant-mcp.git
   cd txtai-assistant-mcp
   

2. Run the startup script:

   ./scripts/start.sh
   

3. Create and activate a virtual environment (recommended):

   python -m venv venv
   source venv/bin/activate  # For Linux/Mac
   venv\Scripts\activate      # For Windows
   

📚 Documentation

Configuration File

The project's configuration information is stored in the .env.template file, with the following specific content:

# Log settings
LOG_LEVEL=INFO
LOG_FILE=./logs/server.log

# Storage path
DATA_PATH=./data

# Network settings
SERVER_HOST=127.0.0.1
SERVER_PORT=5000

Integration with Claude and Cline

Modify the configuration file

To enable the MCP function in Claude or Cline, you need to edit their configuration files:

In Claude:

// config.js
{
  "mcp": {
    "enabled": true,
    "host": "127.0.0.1",
    "port": 5000
  }
}

In Cline:

// settings.json
{
  "integrations": {
    "txtai_mcp": {
      "url": "http://localhost:5000"
    }
  }
}

MCP Toolset

Store Memories

Use the /store endpoint to create new memories:

{
  "content": "Information content to be stored",
  "metadata": {
    "source": "Example",
    "timestamp": "2023-01-01T00:00:00Z"
  },
  "tags": ["Example", "Memory"],
  "type": "General"
}

Retrieve Memories

Retrieve content through semantic queries:

{
  "query": "Search keywords",
  "result_count": 5,
  "similarity_threshold": 0.7
}

Tag - based Retrieval

Find memories based on tags:

{
  "tags": ["Example", "Memory"]
}

Delete Memories

Delete specific memories by content hash:

DELETE /memory/{content_hash}

API Endpoints

Store Memories

POST /store

Request Body:

{
  "content": "Information to be stored",
  "metadata": {
    "source": "Example",
    "timestamp": "2023-01-01T00:00:00Z"
  },
  "tags": ["Important", "Memory"],
  "type": "Default"
}

Retrieve Memories

POST /search

Request Body:

{
  "query": "Search content",
  "result_count": 5,
  "similarity_threshold": 0.6
}

Project Structure

txtai-assistant-mcp/
├── data/          # Folder for storing memories and label indexes
├── logs/         # Location for storing log files
├── scripts/      # Contains startup scripts and other tools
└── txtai_mcp.py  # Core processing logic

💻 Usage Examples

Basic Usage

1. Start the service:

   python txtai_mcp.py
   

2. Send a storage request:

   curl -X POST http://localhost:5000/store -H "Content-Type: application/json" -d '{"content":"This is the first memory","tags":["Test","Initial"]}'
   

3. Retrieve memories:

   curl -X POST http://localhost:5000/search -H "Content-Type: application/json" -d '{"query":"Test"}'
   

🔗 Related Resources

• TxtAI Official Documentation
• MCP Project Homepage

📄 License

This project follows the MIT license. Please refer to the LICENSE file for specific terms.