PDF RAG Assistant

A clean Python project for building a Retrieval-Augmented Generation assistant over local PDF files.

This project lets you:

• Load one or more PDF files from a local folder
• Split the extracted text into searchable chunks
• Create an OpenAI embedding index using FAISS
• Ask questions against the indexed PDF content
• Optionally rerank retrieved chunks with Cohere
• Run a simple evaluation against your own test questions

Project Structure

pdf-rag-assistant/
├── data/                         # Place your PDF files here
├── vector_store/                 # FAISS index is saved here
├── src/
│   └── rag_pdf_assistant/
│       ├── __init__.py
│       ├── chat.py
│       ├── config.py
│       ├── evaluation.py
│       ├── indexer.py
│       ├── pdf_loader.py
│       └── prompts.py
├── main.py
├── requirements.txt
├── .env.example
└── .gitignore

Setup

1. Create a virtual environment

python -m venv .venv

Windows:

.venv\Scripts\activate

Mac/Linux:

source .venv/bin/activate

2. Install dependencies

pip install -r requirements.txt

3. Add environment variables

Create a .env file in the project root:

OPENAI_API_KEY=your_openai_api_key_here
COHERE_API_KEY=optional_cohere_api_key_here

Cohere is optional. If you do not add a Cohere key, the project will use normal FAISS similarity retrieval.

4. Add PDF files

Place your PDFs inside the data/ folder.

5. Build the vector index

python main.py build

6. Ask a question

python main.py ask "What is this document about?"

7. Start interactive chat

python main.py chat

8. Run sample evaluation

python main.py evaluate

Notes for GitHub

Do not upload your .env file, API keys, or private PDFs. The .gitignore file already excludes common sensitive/local files.