VitaeRAG: Local Knowledge Base Q&A with RAG

This is a local knowledge base Q&A project based on Retrieval-Augmented Generation (RAG) technology.
You can upload your own PDF papers, and the project will process them into a vector database.
Then, through a command-line interface, you can query the contents of these papers using a Large Language Model (LLM).

📂 Project Structure

rag-llm-project/
├── data/
│   └── (Please put your PDF papers here)
├── vector_store/
│   └── (This folder will be automatically created by ingest.py to store vector indexes)
├── ingest.py          # 1. Script to process PDFs and create the vector database
├── app.py             # 2. Main application for Q&A with your papers
├── requirements.txt   # 3. Python dependencies required for the project
└── README.md          # 4. Project documentation

🚀 Usage Steps

1. Environment Setup

First, ensure you have Python 3.8 or higher installed.
Then, clone or download this project, open a terminal in the project root directory, and (optionally) create a virtual environment before installing dependencies:

pip install -r requirements.txt

2. Configure the LLM API

This project is set up to call a large language model (such as DeepSeek) via API.

First, create a file named .env in the project root directory.

Then, open the .env file and add your API key and API endpoint in the following format.This repository currently supports DeepSeek and ZhipuAI. For example, using DeepSeek:

LLM_PROVIDER="deepseek"
DEEPSEEK_API_KEY="your_deepseek_api_key_here"
DEEPSEEK_API_BASE="https://api.deepseek.com/v1"
DEEPSEEK_EMBEDDING_MODEL="text-embedding-v2"
DEEPSEEK_CHAT_MODEL="deepseek-chat"

Replace your_deepseek_api_key_here with your actual key.
The app.py script will automatically load this information from the file.

3. Add Your PDF Files

Place the PDF papers or documents you want to query inside the data folder in the project root.

4. Create the Vector Database

Run the ingest.py script in the terminal.
This script will read all the PDFs in the data folder, process and split them, generate embeddings, and finally create a FAISS index file in the vector_store folder.

python ingest.py

This step only needs to be done once unless your document library changes.

5. Start Asking Questions!

Once the vector database has been created, you can run the main application app.py to start asking questions.

python app.py

After the program starts, you can type questions in the terminal, and the model will respond based on the content of your papers. Type exit to quit the program.

✨ Enjoy your usage!

Some screenshots: