LlamaCloud RAG Demo

This repository contains a Streamlit application that demonstrates Retrieval-Augmented Generation (RAG) using LlamaCloud for document retrieval and SQL for structured data queries, powered by Gemini 2.0.

Overview

The application showcases a hybrid approach to question answering by:

• Using a SQL database for structured city data (population statistics)
• Leveraging LlamaCloud for unstructured document retrieval
• Employing Gemini 2.0 for natural language understanding and response generation

This combination allows for precise answers to factual questions while maintaining the flexibility to handle nuanced, context-dependent queries.

Features

• Natural Language Queries: Ask questions about US cities in plain English
• Hybrid RAG Architecture:
  • SQL database for population statistics
  • LlamaCloud for pre-indexed city documents
• Smart Query Routing: Automatically directs queries to the appropriate backend
• Interactive Chat Interface: Built with Streamlit for a user-friendly experience

Demo Screenshot

Installation

Prerequisites

• Python 3.10 or higher
• Google API key (for Gemini access)
• LlamaCloud API key and organization ID

Setup

1. Clone this repository:

git clone https://github.com/pandyaved98/Assignment.git
cd Assignment/advanced_rag

2. Create a virtual environment:

python -m venv venv
source venv/bin/activate

3. Install the required packages:

pip install -r requirements.txt

Required API Keys

This application requires the following API keys:

1. Google API Key for Gemini models:

   • Get your API key from Google AI Studio
   • Required for both LLM inference and embeddings

2. LlamaCloud Credentials:

   • API Key
   • Organization ID
   • Project Name
   • Index Name

You can input these credentials directly in the Streamlit UI or set them as environment variables.

Usage

1. Start the Streamlit application:

streamlit run app.py

2. Open your browser and navigate to http://localhost:8501

3. Input your API keys in the sidebar

4. Start asking questions about US cities like:

   • "What is the population of New York City?"
   • "Which city has the highest population?"
   • "Tell me about the history of Chicago."
   • "Compare the populations of Los Angeles and Houston."

How It Works

The application follows this workflow:

1. User Query Input: The user enters a natural language question about cities
2. Query Analysis: The system determines if the query is about population/statistics or general information
3. Query Routing:
   • Population queries → SQL database (via direct SQL or NLSQLTableQueryEngine)
   • General information queries → LlamaCloud document retrieval
4. Response Generation: The appropriate data is retrieved and used to generate a natural language response
5. Result Display: The answer is shown to the user along with the source of information

Architecture

Requirements

The project requires the following Python packages:

streamlit
llama-index
llama-index-llms-gemini
llama-index-embeddings-gemini
llama-index-indices-managed-llama-cloud
sqlalchemy
pandas

Contributing

Contributions are welcome! Please feel free to submit a Pull Request.

License

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