QueryTune: A Cross-Platform Tool for Intelligently Optimizing SQL Queries Using Large Language Models

QueryTune is a cross-platform tool that uses large language models (LLMs) to intelligently optimize SQL queries. It natively supports macOS and extends to Windows and Linux. Through dual-mode AI processing—local (Ollama-integrated open-source models) and cloud (OpenAI-compatible services)—it provides structured optimization suggestions (SQL refactoring, index recommendations, technical explanations) and flexible chat interactions, addressing the pain point where developers/DBAs can optimize queries without relying on expensive tools or deep professional knowledge.

QueryTune was created by developer meob, with the core goal of solving the long-standing pain point in SQL optimization: quickly obtaining high-quality optimization suggestions without relying on expensive commercial tools or deep database expertise. As a native macOS application, it differs from traditional query analysis tools by fully leveraging LLMs' capabilities in code understanding and generation, supporting optimization from simple SELECT queries to complex multi-table JOIN queries.

QueryTune offers two complementary analysis modes:

1. Optimization Mode: Generates structured results, including refactored SQL (to improve execution efficiency), index recommendations (matching query characteristics), and technical explanations (illustrating optimization effectiveness and expected improvements);
2. Chat Mode: Conversational interaction with streaming responses, supporting open exploration of optimization ideas, suitable for scenarios where the optimization direction is unclear.

QueryTune supports a hybrid architecture of local and cloud AI models:

• Local Deployment: Integrates open-source models like Qwen2.5-Coder, DeepSeek-R1, and Llama3.1 via Ollama, ensuring sensitive queries stay within the local environment, suitable for privacy scenarios;
• Cloud Services: Compatible with OpenAI API and services like Groq and LM Studio, supporting models such as GPT-4o and GPT-4o-mini, ideal for scenarios pursuing extreme speed and optimization quality.

Context Awareness: Provides a collapsible context area, allowing addition of instructions, table sizes, and schema details (it is recommended to only provide DDL of tables involved in the query to avoid noise from irrelevant information), improving the accuracy of suggestions; Model Recommendations:

• Local default: Qwen2.5-Coder7B (balances logical understanding and speed), DeepSeek-R1 series is suitable for reasoning analysis;
• Cloud recommendations: GPT-4o, Claude3.5/3.7 Sonnet (excellent at complex refactoring); Temperature Parameter: Set to 0.0-0.2 for SQL optimization suggestions to ensure grammatical correctness; for reasoning models, set to 0.6 to stimulate performance.

QueryTune extends support to Windows and Linux platforms: macOS offers native menus, Windows uses a standard UI, and Linux provides AppImage and standalone binaries; Personalization options include SQL formatting (indentation, keyword case, etc.), themes, font settings, API key management, and configuration persistence to retain user preferences.

QueryTune is suitable for:

• Application developers: Daily query optimization assistant to identify performance bottlenecks;
• DBAs: Quickly evaluate and optimize queries, reducing manual analysis workload;
• SQL learners: Understand optimization principles through AI suggestions; Core value: Improve query performance, lower the technical threshold for optimization, and represent the development direction of AI-assisted database management.

QueryTune demonstrates the innovative value of combining LLMs with traditional database tools. As the capabilities of open-source models improve, AI-assisted optimization tools will further lower the technical threshold. For developers looking to improve SQL quality, QueryTune is a choice worth trying.