mdsmart 0.1.0

Multi-Database Smart Query Generator - Convert natural language to SQL using AI

MDSmart - Multi-Database Smart Query Generator

MDSmart converts natural language questions into SQL using local LLMs via Ollama. It supports SQLite by default, with optional connectors for PostgreSQL and MySQL/MariaDB, plus a built-in FAISS vector store for training examples and a Knowledge Base for injecting business rules and documentation.

Features

• Local LLMs with Ollama (no external keys required)
• SQLite, PostgreSQL, and MySQL/MariaDB support with automatic schema extraction (install extras for non-SQLite)
• Retrieval-augmented prompts using FAISS (examples + knowledge)
• Retry flow to fix SQL when execution fails

Quick Start

from mdsmart import MDSmart

# Initialize (ensure `ollama serve` is running)
md = MDSmart(model="llama3", embedding_model="nomic-embed-text")

# Connect to database
md.connect_database(
    db_id="sales_db",
    db_type="sqlite",
    connection_params={"path": "sales.db"},
    description="Sales and orders data",
    keywords=["sales", "orders", "revenue"]
)

# Ask a question
result = md.ask("What are the top 10 products by revenue?")
print(result)

Requirements

• Python 3.8+
• Ollama installed and running
• FAISS for vector search (pip install faiss-cpu)

More Documentation

• Getting Started: docs/getting-started.md
• API Reference: docs/api.md