Semantic MCP Layer

A working implementation of a semantic layer that enhances the MCP memory server with real pattern matching and type inference capabilities.

Features

• SQLite-based pattern storage
• Real-time type inference with confidence scoring
• Pattern matching with attribute validation
• API endpoints for inference and pattern management

Installation

pip install -r requirements.txt

Usage

1. Start the server:

python src/api.py

2. Test the functionality:

python tests/test_semantic.py

API Endpoints

Infer Types

POST /infer

{
    "id": "doc1",
    "attributes": {
        "title": "Project Document",
        "format": "pdf"
    }
}

Get Patterns

GET /patterns

How It Works

1. Pattern Matching:

   • Validates required attributes
   • Checks string patterns and values
   • Applies keyword matching

2. Confidence Scoring:

   • Weights required vs optional attributes
   • Adjusts for pattern specificity
   • Considers attribute match quality

3. Storage:

   • Patterns stored in SQLite
   • Inference results cached
   • Pattern confidence tracked

Development

• Add new patterns in semantic_core.py
• Extend pattern matching in _match_value
• Add new API endpoints in api.py