QA System with Google Gemini and LlamaIndex

This project is a Question Answering (QA) system that leverages Google Gemini LLM and document embeddings using llama-index. It enables semantic search and question answering over your own documents, with an interactive Jupyter notebook and a Streamlit app for experimentation.

---

Features

• Document Ingestion: Load and process documents from a directory.
• Embeddings: Use Gemini embeddings for semantic understanding.
• LLM Integration: Query Google Gemini LLM for answers.
• Indexing: Build a vector store index for efficient retrieval.
• Interactive Notebook: Step-by-step code in notebooks/expriments.ipynb.
• Streamlit App: User-friendly interface for QA (see notebooks/StreamlitApp.py).

---

Project Structure

qasystem/
│
├── Data/                      # Place your PDF or text documents here
├── QAWithPDF/                 # Core modules (if any)
├── notebooks/
│   ├── expriments.ipynb       # Main Jupyter notebook
│   └── StreamlitApp.py        # Streamlit app for QA
├── requirements.txt           # Python dependencies
└── README.md                  # Project documentation

---

Setup

1. Clone the repository:

   git clone https://github.com/yourusername/qasystem.git
   cd qasystem

2. Install dependencies:

   pip install -r requirements.txt

3. Set up your Google API key:

   • Create a .env file in the root directory.
   • Add your key:

     GOOGLE_API_KEY=your_google_api_key_here
     

---

Usage

Jupyter Notebook

1. Place your documents (PDFs, text files) in the Data/ directory.
2. Open notebooks/expriments.ipynb in Jupyter.
3. Run the cells to:
   • Load your API key and dependencies.
   • List available Gemini models.
   • Load and preview your documents.
   • Initialize Gemini LLM and embedding models.
   • Build the vector index.
   • Query the index with natural language questions.

Streamlit App

To run the Streamlit app for a web-based interface:

streamlit run notebooks/StreamlitApp.py

---

Example Workflow (Notebook)

• Load API Key:
  Uses dotenv to securely load your Google API key.
• List Gemini Models:
  See which Gemini models are available for use.
• Document Loading:
  Reads all files from the Data/ directory.
• Embedding & Indexing:
  Embeds documents and builds a vector index.
• Querying:
  Ask questions like "What is X?" and get answers from your documents.

---

Requirements

See requirements.txt for all dependencies, including:

• llama-index
• google-generativeai
• llama-index-llms-gemini
• pypdf
• python-dotenv
• IPython
• streamlit

---

Author: Aditya Sethi