An interactive chatbot built with Streamlit, LangChain, FAISS, and Google Gemini API.
Upload any PDF and chat with it — the app extracts the text, creates vector embeddings, and answers your questions using Gemini + retrieval.
- Upload any PDF via sidebar
- PDF text is chunked & indexed in FAISS
- Ask natural questions about the content
- Answers generated with Google Gemini (via LangChain)
- Chat history stored during the session
- Clean and modern Streamlit UI
├── app.py # Main Streamlit app ├── config/ │ └── settings.py # Loads GOOGLE_API_KEY from .env ├── src/ │ ├── pipeline.py # Indexing, QA pipeline │ ├── pdf_loader.py # PDF text extraction │ ├── text_processing.py # Cleaning & chunking │ ├── vectorstore.py # FAISS vector DB utilities │ └── llm.py # LLM + embeddings providers ├── data/ # Uploaded PDFs will be stored here ├── faiss_index/ # Persisted FAISS vector store ├── requirements.txt └── README.md |___.env
- Clone the repo
git clone https://github.com/Shubham9403-git/Advanced-Chat-Bot.git cd pdf-chatbot - **Create a virtual environment
python -m venv venv source venv/bin/activate # macOS/Linux venv\Scripts\activate # Windows
- **Install dependencies
pip install -r requirements.txt
- **Set up environment variables Create a .env file in the project root: GOOGLE_API_KEY=your_google_api_key_here
📖 Usage
Upload a PDF from the sidebar Wait for processing (text extraction + FAISS indexing) Ask questions in the chat box Get answers grounded in your document
🛠️ Tech Stack
Streamlit – Web UI LangChain – Orchestration FAISS – Vector database Google Gemini – LLM & embeddings PyPDF – PDF text extraction
Requires a Google API key for Gemini models Currently supports one PDF at a time All uploaded PDFs are stored in the data/ folder
📌 Roadmap
Support multiple PDFs in one session Add persistent chat history Add support for non-PDF formats (DOCX, TXT) Deploy to cloud (Streamlit Community Cloud, Hugging Face Spaces, etc.)
📜 License
MIT License © 2025