Halo

Halo is an AI-powered database schema architect and query builder. It allows you to converse with an AI assistant to design, modify, and query relational databases visually.

Features

• AI-Powered Design: Converse naturally with the AI to generate complete database schemas.
• Visual Canvas: Instantly see your generated tables, columns, and relationships rendered on an interactive canvas using React Flow.
• Dynamic Modifications: Ask the AI to add, remove, or modify existing tables and relationships, and watch the canvas update in real time.
• Query Generation: Once your schema is built, ask the AI to write complex SQL queries based on the generated structure.

Tech Stack

• Framework: Next.js
• AI Integration: Vercel AI SDK
• LLM Provider: NVIDIA NIM (google/gemma-4-31b-it)
• State Management: Zustand
• Visualization: React Flow & Dagre (for auto-layout)
• Styling: Tailwind CSS & shadcn/ui

Getting Started

Prerequisites

Ensure you have Node.js and a package manager (npm, yarn, pnpm, or bun) installed.

Installation

1. Clone the repository and install dependencies:

   pnpm install
   

2. Set up your environment variables by creating a .env.local file in the root directory:

   NVIDIA_API_KEY=your_nvidia_api_key_here
   NVIDIA_BASE_URL=https://integrate.api.nvidia.com/v1
   NVIDIA_MODEL=google/gemma-4-31b-it
   

3. Start the development server:

   pnpm dev
   

4. Open http://localhost:3000 in your browser.

How to Use

1. Start a Conversation: Use the chat panel on the left to ask the AI to create a database schema (e.g., "Design a schema for an e-commerce platform").
2. Interact with the Canvas: The AI will generate the schema and the right-hand canvas will automatically render the tables and relationships.
3. Iterate: Ask the AI to make changes, like "Add a stripe_customer_id to the users table".
4. Query: Ask for SQL queries, such as "Write a query to get the top 5 spending users".