Renovation-LLM – AI-Powered Interior Design Assistant

An AI-powered tool that assists interior designers and architects by analyzing real-world renovation datasets and generating high-quality design visuals based on user prompts.

📌 Overview

Renovation-LLM is a workflow-driven AI assistant that helps interior designers and architects streamline their creative process. This project is a small-scale version of what Renovate.ai does, utilizing structured interior design data and AI-powered image generation to offer personalized design inspiration.

The tool analyzes a large dataset of interior design images, extracts structured insights, and generates SQL queries based on user prompts. It then retrieves relevant design data and feeds it into OpenAI’s image generation models to create stunning AI-powered interior design concepts.

🔥 Key Features

✅ Interior Design Data Analysis – Processes real-world renovation datasets from Kaggle.
✅ PostgreSQL Storage – Stores structured image metadata for efficient querying and retrieval.
✅ AI-Powered Workflow – Uses an agentic framework to process prompts and generate SQL queries dynamically.
✅ Keyword Extraction – Extracts key design elements from user inputs to retrieve the most relevant images.
✅ AI Image Generation – Uses OpenAI’s image generator to create custom design concepts based on structured data.

🏗️ Tech Stack

• Database: PostgreSQL (Stores interior design image metadata)
• Data Processing: Python (for dataset processing and SQL integration)
• AI Agent Framework: Custom workflow for function calling & query generation
• Image Generation: OpenAI API (Generates new interior design concepts)
• Dataset Source: Kaggle (Synthetic Home Interior Dataset)

📂 Dataset & Processing

The dataset was downloaded from Kaggle:

import kagglehub

# Download latest version
path = kagglehub.dataset_download("luznoc/synthetic-dataset-for-home-interior")

Each image was analyzed and stored in a PostgreSQL database with the following schema:

Column Name	Data Type	Description
id	INT (PK)	Unique identifier for each image
image_url	TEXT	Link to the original image
image_description	TEXT	AI-generated description of the image
room_type	TEXT	Type of room (e.g., kitchen, living room)
price	FLOAT	Estimated renovation cost
color	TEXT	Primary colors in the image
style	TEXT	Design style (e.g., modern, vintage)

🛠️ Installation & Setup

Clone the repository:

git clone https://github.com/MozartofCode/Renovation-LLM.git
cd Renovation-LLM

Set up the PostgreSQL database

1. Install PostgreSQL and create a database:

   CREATE DATABASE renovation_db;

2. Set up the table schema:

   CREATE TABLE renovation_photos (
       id SERIAL PRIMARY KEY,
       image_url TEXT,
       image_description TEXT,
       room_type TEXT,
       price INT,
       color TEXT,
       style TEXT
   );

3. Load the dataset into PostgreSQL using a Python script (provided in the repo).

🎯 How It Works

1️⃣ User enters a design prompt (e.g., "modern kitchen with white marble countertops").
2️⃣ The AI extracts key design elements (e.g., "modern", "kitchen", "marble").
3️⃣ It generates an SQL query to retrieve matching images from the database.
4️⃣ The retrieved image descriptions are fed into OpenAI’s image generator to create a high-quality AI-generated design.

🚧 Future Enhancements

🔹 Integration with 3D Rendering Software – Connect with Blender or Unreal Engine for realistic renderings.
🔹 User Preference Learning – Train AI models to recommend designs based on past preferences.
🔹 Multi-Modal AI – Combine text, image, and video for richer design recommendations.

📬 Contact

Author: Bertan Berker