Data-Aware RAG Bot

A PDF chatbot built with Streamlit. Users can upload a document, build an index, and ask questions about the uploaded content.

Live app: https://dmb8exalpgpnr57uxxsuyu.streamlit.app/

Source code: https://github.com/dp24723/Data-Aware-RAG-bot-

About

This project is a document question-answering app. It extracts text from uploaded PDFs or text files, creates a searchable index, and uses Gemini to answer questions based on the document content.

I built this project to make it easier to search, summarize, and understand long documents without manually reading page by page.

How to use

1. Open the live app.
2. Upload a PDF, TXT, or Markdown file from the sidebar.
3. Click Build index.
4. Wait for the indexing process to finish.
5. Type your question in the chat box.
6. The app will generate an answer based on the uploaded document.
7. Open the Sources section to check the text used for the answer.

Example questions:

• Give me a summary of this document.
• What is the email address in this resume?
• What projects are mentioned?
• What technical skills are listed?
• Summarize the experience section.

Features

• Upload PDF, TXT, and Markdown files
• Build a searchable document index
• Ask questions in a chat interface
• Generate answers using Gemini
• Show source snippets used for each answer
• Run locally or deploy on Streamlit Community Cloud

Tech stack

• Python
• Streamlit
• Gemini API
• PyPDF
• NumPy
• Scikit-learn
• Docker

Project structure

app.py is the main Streamlit application.

The core logic is inside the src/pdf_chatbot folder:

• document_loader.py extracts text from uploaded files
• store.py builds and searches the document index
• answer.py generates answers using retrieved context
• config.py handles app settings
• embeddings.py handles text representation

Run locally

Clone the repository:

git clone https://github.com/dp24723/Data-Aware-RAG-bot-.git

cd Data-Aware-RAG-bot-

Create a Python environment:

conda create -n raglive python=3.11 -y

conda activate raglive

Install dependencies:

python -m pip install --upgrade pip

python -m pip install -r requirements.txt

Create a local Streamlit secrets file:

mkdir -p .streamlit

touch .streamlit/secrets.toml

Add your Gemini key inside .streamlit/secrets.toml:

LLM_PROVIDER = "gemini" GEMINI_API_KEY = "YOUR_GEMINI_API_KEY" GEMINI_MODEL = "gemini-3.1-flash-lite" PDF_CHAT_STORAGE = "storage" TOP_K = "6"

Start the app:

python -m streamlit run app.py

Open:

http://localhost:8501

Deployment

The app is deployed on Streamlit Community Cloud.

For deployment, add the Gemini API key in Streamlit Cloud under:

App settings → Secrets

Use this format:

LLM_PROVIDER = "gemini" GEMINI_API_KEY = "YOUR_GEMINI_API_KEY" GEMINI_MODEL = "gemini-3.1-flash-lite" PDF_CHAT_STORAGE = "storage" TOP_K = "6"

The main file path is:

app.py

Security

API keys are not included in this repository.

Do not commit these files or folders:

• .env
• .streamlit/secrets.toml
• storage/
• pycache/
• .venv/

Future work

• Add DOCX support
• Improve retrieval ranking
• Add chat history export
• Support multiple document collections
• Add persistent storage for indexed documents