Streamlit LlamaIndex RAG with Citations

This project is a web application built with Streamlit that implements a Retrieval-Augmented Generation (RAG) pipeline using LlamaIndex. It allows users to ask questions about a collection of documents and receive answers with citations, pointing to the specific sources of information within the documents.

The application uses ChromaDB as a vector store to efficiently retrieve relevant context from the documents in the data directory.

These are screenshots of an ongoing conversation, showing the response of the LLM with source citation:

Features

• Interactive chat interface powered by Streamlit.
• Query documents using natural language.
• Receive answers generated by a Large Language Model (LLM).
• Answers include citations to the source documents for verification.
• Uses LlamaIndex for the RAG pipeline.
• Uses ChromaDB for vector storage.

Tech Stack

• Streamlit
• OpenAI
• LlamaIndex
• NLTK
• ChromaDB

Setup and Installation

1. Clone the repository:

   git clone <repository-url>
   cd streamlit-llamaindex-rag-citation

2. Create and activate a Python virtual environment:

   python -m venv citationenv
   source citationenv/bin/activate
   # On Windows, use:
   # citationenv\Scripts\activate

3. Install the required dependencies:

   pip install -r requirements.txt

4. Set up your API keys: Create a file named secrets.toml inside a .streamlit directory.

   .streamlit/
   └── secrets.toml
   

5. Add your OpenAI API key to secrets.toml:

   OPENAI_API_KEY = "sk-..."

Usage

1. Add your data: Place the PDF documents you want to query in the data directory. The project comes with some example documents.

2. Run the Streamlit application:

   streamlit run citation_app.py

3. Open your browser: Navigate to the local URL provided by Streamlit (usually http://localhost:8501).

Project Structure

.
├── .gitignore
├── citation_app.py         # Main Streamlit application
├── readme.md               # This file
├── requirements.txt        # Python dependencies
├── .streamlit/
│   └── secrets.toml        # Secrets management for Streamlit
├── chroma_db/              # ChromaDB vector store
├── citation/               # LlamaIndex storage
├── citationenv/            # Python virtual environment
└── data/                   # Source documents (PDFs)