https://github.com/pakagronglb/python-rag-agent-pydantic-ai

A powerful property information assistant built with Python, OpenAI, and LanceDB that uses Retrieval Augmented Generation (RAG) to answer questions about property listings.
https://github.com/pakagronglb/python-rag-agent-pydantic-ai

lancedb openai pydantic pydantic-ai python rag streamlit

Last synced: 4 months ago
JSON representation

A powerful property information assistant built with Python, OpenAI, and LanceDB that uses Retrieval Augmented Generation (RAG) to answer questions about property listings.

• Host: GitHub
• URL: https://github.com/pakagronglb/python-rag-agent-pydantic-ai
• Owner: pakagronglb
• Created: 2025-03-03T00:34:25.000Z (7 months ago)
• Default Branch: main
• Last Pushed: 2025-03-03T01:25:37.000Z (7 months ago)
• Last Synced: 2025-03-03T01:27:49.643Z (7 months ago)
• Topics: lancedb, openai, pydantic, pydantic-ai, python, rag, streamlit
• Language: Python
• Homepage:
• Size: 0 Bytes
• Stars: 0
• Watchers: 1
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

README

          
# 🏠 Property Information RAG Assistant

A powerful property information assistant built with Python, OpenAI, and LanceDB that uses Retrieval Augmented Generation (RAG) to answer questions about property listings.

[![Python](https://img.shields.io/badge/Python-3.13-blue.svg)](https://www.python.org/)

[![OpenAI](https://img.shields.io/badge/OpenAI-GPT--4o-green.svg)](https://openai.com/)

[![LanceDB](https://img.shields.io/badge/LanceDB-0.20.0-orange.svg)](https://lancedb.com/)

[![Streamlit](https://img.shields.io/badge/Streamlit-1.42.2-red.svg)](https://streamlit.io/)

[![Pydantic](https://img.shields.io/badge/Pydantic-2.10.6-purple.svg)](https://pydantic.dev/)

## 🚀 Features

- **Vector Search**: Fast and accurate semantic search using LanceDB and OpenAI embeddings

- **Conversational UI**: Clean user interface built with Streamlit

- **Rich Property Information**: Query details about properties, including:

  - Pet policies (allowed pets, fees, weight limits)

  - Location information

  - Amenities and features

  - Pricing details

- **Intelligent Query Understanding**: Converts natural language questions into optimized search queries

- **Well-formatted Responses**: Clear, structured answers with relevant property details

## 📋 Table of Contents

- [Installation](#installation)

- [Usage](#usage)

- [Project Structure](#project-structure)

- [How It Works](#how-it-works)

- [Credits](#credits)

- [License](#license)

## 🔧 Installation

1. Clone the repository:

   ```bash

   git clone http://www.github.com/pakagronglb/python-rag-agent-pydantic-ai

   cd python-rag-agent-pantic-ai

   ```

2. Create and activate a virtual environment:

   ```bash

   python -m venv venv

   source venv/bin/activate  # On Windows: venv\Scripts\activate

   ```

3. Install the required packages:

   ```bash

   pip install -r requirements.txt

   ```

4. Set up your OpenAI API key in a `.env` file:

   ```

   OPENAI_API_KEY=your_api_key_here

   ```

## 💻 Usage

1. Run the setup script to initialize the knowledge base:

   ```bash

   python lancedb_setup.py

   ```

2. Launch the Streamlit web interface:

   ```bash

   python -m streamlit run app.py

   ```

3. Open your browser and navigate to:

   ```

   http://localhost:8501

   ```

4. Enter your property-related questions in the input field and click "Search"

## 🏗️ Project Structure

- `app.py`: Streamlit web interface

- `agent_run.py`: Command-line version of the query agent

- `lancedb_setup.py`: Knowledge base setup and vector database operations

- `knowledge-file/`: Directory containing markdown property information files

## ⚙️ How It Works

1. **Data Ingestion**: Property information is stored as markdown files in the `knowledge-file` directory

2. **Vector Embedding**: The text is chunked and embedded using OpenAI's text-embedding-3-small model

3. **Query Processing**: User queries are converted to optimized search queries using an LLM

4. **Vector Search**: LanceDB retrieves the most semantically similar documents

5. **Response Generation**: An LLM generates a well-formatted response based on the retrieved context

## 🙏 Credits

This project was inspired by [Jie Jenn's YouTube tutorial](https://www.youtube.com/watch?v=P212vYt6Xo8) on building RAG applications with Python. His clear explanations and practical approach were instrumental in creating this application.

## 📄 License

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

🏠 Property Information Assistant

    [Header]

Ask a Question

    [Input Form with background]

    [Full-width Search button]

📝 Response

    [Response area with background]