🤖 Personal AI Assistant (RAG Pipeline)

A powerful retrieval-augmented generation (RAG) application that allows you to chat with your own documents (PDFs and FAQs) using Google Gemini and FAISS.

🚀 Features

• PDF Upload: Upload any PDF and ask questions about its content.
• FAQ Support: Pre-loaded knowledge base from structured JSON FAQs.
• Local Embeddings: Fast, local vectorization using sentence-transformers.
• Smart Retrieval: Uses FAISS for efficient similarity search.
• Strict Answering: The AI only answers based on the provided documents to prevent "hallucinations."
• Source Attribution: See exactly which page or FAQ the AI used to generate the answer.

---

🏗️ Project Architecture (A-Z)

This project follows a modular RAG architecture. Here is a breakdown of "what is where":

Core Entry Point

• app.py: The main Streamlit interface. It orchestrates the entire flow from document upload to chat interactions.

Source Code (src/)

• config.py: Handles environment variables and API key management using python-dotenv.
• loaders/:
  • pdf_loader.py: Uses pypdf to extract text from PDF files.
  • text_loader.py: Processes JSON files into a structured document format.
• utils/:
  • text_cleaning.py: Normalizes extracted text and fixes common PDF character-spacing issues.
  • chunking.py: Splits long documents into manageable chunks (700 chars with overlap) for better search accuracy.
• embeddings/:
  • embedder.py: Converts text into 384-dimensional vectors using the all-MiniLM-L6-v2 model.
• vectorstore/:
  • faiss_store.py: Manages the FAISS index for high-speed vector similarity searching.
• llm/:
  • gemini_client.py: Wrapper for the Google Gemini API (gemini-2.5-flash-lite) to generate responses.
• rag/:
  • pipeline.py: The logic that builds the "Context + Question" prompt for the LLM.

---

🛠️ Tech Stack

• Frontend: Streamlit
• LLM: Google Gemini API
• Vector DB: FAISS
• Embeddings: Sentence-Transformers (all-MiniLM-L6-v2)
• Lang: Python 3.10+

---

⚙️ Installation & Setup

1. Clone the repository:

   git clone <repo-url>
   cd final-project

2. Set up virtual environment:

   python -m venv .venv
   source .venv/bin/activate  # On Windows: .venv\Scripts\activate

3. Install dependencies:

   pip install -r requirements.txt

4. Configuration: Create a .env file in the root directory and add your Gemini API Key:

   GEMINI_API_KEY=your_api_key_here

5. Run the App:

   streamlit run app.py

---

📄 Usage

1. Open the URL provided by Streamlit (usually http://localhost:8501).
2. The app will automatically load the default FAQ and sample PDF.
3. Upload a new PDF using the sidebar/uploader to chat with specific documents.
4. Type your question in the chat input at the bottom.
5. Check the "Sources" dropdown below the AI's answer to verify its facts.