dsllm - Domain Specific Language from LLM

Generate Domain Specific Language (DSL) statements from Natural Language input using Large Language Models (LLMs).

Overview

dsllm is a Python library that converts natural language descriptions into structured DSL statements. Currently supports SQL generation with OpenAI models, designed to be extensible to other DSLs and LLM providers.

Features

• SQL Generation: Convert natural language to SQL queries
• OpenAI Integration: Built-in support for GPT-3.5/4 models
• Context Injection: Support for database schemas and constraints
• Validation: Pre-flight and post-flight validation checks
• Safety Features: Read-only mode, dangerous operation detection
• Retry Mechanism: Automatic retries with exponential backoff

Installation

pip install dsllm

Quick Start

import asyncio
import os
from dsllm import DSLLMGenerator, OpenAIProvider, SQLGenerator

async def main():
    # Initialize components
    provider = OpenAIProvider(api_key=os.getenv("OPENAI_API_KEY"))
    sql_generator = SQLGenerator()
    generator = DSLLMGenerator(provider=provider, dsl_generator=sql_generator)
    
    # Generate SQL from natural language
    result = await generator.generate(
        natural_language="Find all active users who registered this month",
        context={
            "schema": '''
                CREATE TABLE users (
                    id SERIAL PRIMARY KEY,
                    email VARCHAR(255),
                    is_active BOOLEAN DEFAULT true,
                    created_at TIMESTAMP
                );
            '''
        }
    )
    
    print("Generated SQL:")
    print(result.dsl_statement)

# Run the example
asyncio.run(main())

Basic Usage

Simple Query Generation

result = await generator.generate("Show me all users")
print(result.dsl_statement)  # SELECT * FROM users;

With Context and Constraints

result = await generator.generate(
    "Find recent orders with high values",
    context={"tables": ["orders", "customers"]},
    constraints={"max_rows": 50}
)

Read-Only Mode

# Enable read-only mode for safety
sql_generator = SQLGenerator(enforce_read_only=True)
generator = DSLLMGenerator(provider=provider, dsl_generator=sql_generator)

# Only SELECT statements will be generated
result = await generator.generate("Get user data")  # ✅ Works
result = await generator.generate("Delete old records")  # ❌ Fails validation

Configuration

Environment Variables

• OPENAI_API_KEY: Your OpenAI API key

Provider Options

provider = OpenAIProvider(
    model="gpt-4",           # Model to use
    temperature=0.1,         # Generation temperature
    max_tokens=500          # Maximum tokens
)

Roadmap

• Additional DSLs: GraphQL, MongoDB queries, Elasticsearch DSL
• More LLM Providers: Anthropic Claude, Google PaLM, Azure OpenAI
• Advanced Features: Agentic loops, automated testing, observability
• Enhanced Validation: Semantic validation, performance analysis

Contributing

Contributions are welcome! Please see our Contributing Guide for details.

License

This project is licensed under the MIT License - see the LICENSE file for details.

Author

Frenk Dragar (@frenkd)