DeepSeek PDF Assistant

This project is a Streamlit-based interactive PDF question-answering assistant powered by LangChain, FAISS, and Ollama's DeepSeek LLM. The application allows users to upload PDF documents, process them into semantic chunks, and retrieve accurate, context-aware answers.

---

Requirements

• Python 3.8 or higher
• Streamlit
• LangChain
• Ollama CLI
• DeepSeek Model (DeepSeek-r1)

---

Installation

Step 1: Clone the Repository

git clone https://github.com/akshayks13/DeepSeek-PDF-Assistant.git
cd DeepSeek-PDF-Assistant

Step 2: Set Up Virtual Environment (Optional)

python3 -m venv venv
source venv/bin/activate  # For Mac/Linux
.\venv\Scripts\activate  # For Windows

Step 3: Install Dependencies

pip install -r requirements.txt

Step 4: Install Ollama

👉 Follow the Ollama Installation Guide

Alternatively, use the command below:

brew install ollama

Step 5: Download the DeepSeek Model

Ollama provides different variants of the DeepSeek model to suit various computational needs. Start by downloading the default 7B model or choose a variant that fits your requirements.

To Download the Default 7B Model:

ollama pull deepseek-r1

To Download Other Model Variants:

• 1.5B Model:

  ollama pull deepseek-r1:1.5b

  Suitable for lightweight applications where memory is constrained.

• 14B Model:

  ollama pull deepseek-r1:14b

  Balances reasoning power and resource usage.

• 70B Model:

  ollama pull deepseek-r1:70b

  Offers advanced reasoning but requires significant RAM.

• 671B Model:

  ollama pull deepseek-r1:671b

  Best for heavy computational tasks requiring superior reasoning.

Tip

Choose a model variant based on your system's RAM and performance requirements.

---

Running the Application

1. Start the Streamlit application.

   streamlit run app.py

2. Open your browser and navigate to the URL

   

---

Usage

1. Upload a PDF file through the UI.
2. Enter your question in the provided input box.
3. View the response based on the uploaded document.

---

Notes

• Ensure you have the DeepSeek model (deepseek-r1) downloaded before running the app.

---

Acknowledgements

Special thanks to LangChain, FAISS, and Ollama for their amazing tools and libraries.