GenAI Database Explorer: An Intelligent Database Exploration Tool Based on Semantic Models

GenAI Database Explorer is an innovative tool that uses generative AI to explore relational databases. Its core lies in building and enriching a semantic model as an intermediate layer instead of querying the database directly, helping users solve the pain points of understanding complex schemas and stored procedures. This tool is applicable to scenarios such as new member onboarding and legacy system maintenance, with advantages including controllability, versioned management, and offline availability.

In enterprise application development, developers often face issues with historical databases: lack of standardized design, non-standard naming, missing documentation, and complex stored procedures. Traditional database documentation tools only mechanically list table structures and cannot help understand business relationships, which gave rise to the solution of GenAI Database Explorer.

GenAI Database Explorer adopts a "semantic model-first" strategy, with a workflow divided into three steps:

1. Extraction phase (extract raw structure from the schema);
2. Enrichment phase (add semantic explanations using generative AI);
3. Query phase (answer questions based on the semantic model). The advantages of this method include: controllability (ability to review and adjust the model), versioned management (stored in JSON and included in Git), and offline availability (no need to connect to the actual database).

The tool is provided as a console application, with core commands including:

• init-project: Initialize the project and create the settings.json directory structure;
• extract-model: Extract database schema information to generate the initial semantic model;
• data-dictionary: Supplement additional information through the data dictionary;
• enrich-model: Use GenAI services to generate descriptions, business purposes, etc., for tables, columns, and stored procedures;
• show-object: View detailed information of a specific object;
• query-model: Answer natural language questions based on the semantic model (e.g., explain business concepts, recommend SQL, analyze stored procedure logic).

GenAI Database Explorer is suitable for the following scenarios:

1. New team member onboarding: Quickly understand the system by asking natural language questions (e.g., "Which tables store order data?");
2. Legacy system maintenance: Locate related tables, understand stored procedures, and evaluate the impact of changes;
3. Data migration projects: Understand the source system model, identify core entities, and plan the target schema.

Architecture Decisions: The reasons for choosing a semantic model instead of direct query include: non-standard naming of organically grown databases, integrable value of data dictionaries, and support for injecting domain knowledge. Future Plans: Develop a web application interface to lower the entry barrier; enhance the natural language understanding ability of query-model to support more complex analysis.

Steps to Use: 1. Create a project (genai-database-explorer init-project); 2. Configure the connection (edit settings.json); 3. Extract the model (extract-model); 4. Enrich semantics (enrich-model);5. Start exploring (query-model). Open Source License: Adopts the MIT license, maintained by Daniel Scott-Raynsford, encourages community customization and extension, and is positioned as a starting point for custom projects rather than an off-the-shelf product.