In this tutorial, I’ll guide you through running DeepSeek-R1 locally, step by step, and setting it up with Ollama. We’ll also build a simple RAG application that runs on your laptop using the R1 model, LangChain, and Gradio.

If you're looking for an overview of the R1 model, check out this DeepSeek-R1 article. For fine-tuning instructions, refer to this tutorial on fine-tuning DeepSeek-R1.

Why Run DeepSeek-R1 Locally?

Running DeepSeek-R1 on your machine gives you full control over execution without relying on external servers. Key benefits include:

• Privacy & Security: Keeps all data on your system.

• Uninterrupted Access: Avoids rate limits, downtime, and service disruptions.

• Performance: Delivers faster responses by eliminating API latency.

• Customization: Allows parameter adjustments, prompt fine-tuning, and local application integration.

• Cost Efficiency: Removes API costs by running the model locally.

• Offline Availability: Enables use without an internet connection once downloaded.

Setting Up DeepSeek-R1 Locally with Ollama

Ollama simplifies local LLM execution by managing model downloads, quantization, and deployment.

Step 1: Install Ollama

Download and install Ollama from the official website.

Once the download is complete, install the Ollama application as you would any other software.

Step 2: Download and Run DeepSeek-R1

Now, let's test the setup and download the model. Open a terminal and run the following command:

Model Variants

Ollama provides multiple versions of DeepSeek-R1, ranging from 1.5B to 671B parameters. The 671B model is the original DeepSeek-R1, while the smaller models are distilled versions based on Qwen and Llama architectures.

If your hardware cannot support the 671B model, you can run a smaller version by replacing X in the command below with the desired parameter size (1.5b, 7b, 8b, 14b, 32b, 70b, or 671b):

This flexibility allows you to use DeepSeek-R1 even without high-end hardware.

Step 3: Running DeepSeek-R1 in the Background

To keep DeepSeek-R1 running continuously and make it available via an API, start the Ollama server:

This will enable integration with other applications.

Using DeepSeek-R1 Locally

Step 1: Running Inference via CLI

Once the model is downloaded, you can interact with DeepSeek-R1 directly from the terminal.

Step 2: Accessing DeepSeek-R1 via API

To integrate DeepSeek-R1 into applications, use the Ollama API with curl:

curl http://localhost:11434/api/chat -d '{
  "model": "deepseek-r1",
  "messages": [{ "role": "user", "content": "Solve: 25 * 25" }]

Note: curl is a command-line tool available on Linux and macOS that allows users to make HTTP requests directly from the terminal, making it useful for interacting with APIs.

Step 3: Accessing DeepSeek-R1 via Python

You can run Ollama in any integrated development environment (IDE) of your choice. First, install the Ollama Python package:

Once installed, use the following script to interact with the model:

import ollama

response = ollama.chat(
    model="deepseek-r1",
    messages=[
        {"role": "user", "content": "Explain Newton's second law of motion"},
    ],
)

print(response["message"]["content"]

The ollama.chat() function processes the user’s input as a conversational exchange with the model. The script then extracts and prints the model’s response.

Running DeepSeek-R1 Locally in VSCode

Running a Local Gradio App for RAG With DeepSeek-R1

Now, let's build a simple demo app using Gradio to query and analyze documents with DeepSeek-R1.

Step 1: Prerequisites

Before implementation, ensure you have the following tools and libraries installed:

• Python 3.8+

• LangChain – A framework for building LLM-powered applications, facilitating easy retrieval, reasoning, and tool integration.

• ChromaDB – A high-performance vector database for efficient similarity searches and embedding storage.

• Gradio – For creating a user-friendly web interface.

Install the necessary dependencies using:

Once installed, import the required libraries:

Step 2: Processing the Uploaded PDF

Now, let's process the uploaded PDF:

How it Works:

The process_pdf() function:
✔ Loads and prepares PDF content for retrieval-based answering.
✔ Extracts text using PyMuPDFLoader.
✔ Splits text into chunks using RecursiveCharacterTextSplitter.
✔ Generates vector embeddings using OllamaEmbeddings.
✔ Stores embeddings in a Chroma vector store for efficient retrieval.

Step 3: Combining Retrieved Document Chunks

After retrieving document chunks, we need to merge them for better readability:

Since retrieval-based models pull relevant excerpts rather than entire documents, this function ensures extracted content is properly formatted before being passed to DeepSeek-R1.

Step 4: Querying DeepSeek-R1 Using Ollama

Now, let’s set up DeepSeek-R1 for processing queries:

How it Works:

✔ Formats the user’s question and retrieved document context into a structured prompt.
✔ Sends the input to DeepSeek-R1 via ollama.chat().
✔ Processes the question in context and returns a relevant answer.
✔ Strips unnecessary thinking output using re.sub().

Step 5: The RAG Pipeline

Now, let’s build the full RAG pipeline:

How it Works:

✔ Searches the vector store using retriever.invoke(question).
✔ Retrieves and formats the most relevant document excerpts.
✔ Passes structured content to ollama_llm() for context-aware responses.

Step 6: Creating the Gradio Interface

Now, let's build a Gradio web interface to allow users to upload PDFs and ask questions:

def ask_question(pdf_bytes, question):
    text_splitter, vectorstore, retriever = process_pdf(pdf_bytes)

    if text_splitter is None:
        return None  # No PDF uploaded

    result = rag_chain(question, text_splitter, vectorstore, retriever)
    return {result}


interface = gr.Interface(
    fn=ask_question,
    inputs=[
        gr.File(label="Upload PDF (optional)"),
        gr.Textbox(label="Ask a question"),
    ]

How it Works:

✔ Checks if a PDF is uploaded.
✔ Processes the PDF using process_pdf() to extract text and generate embeddings.
✔ Retrieves relevant information using rag_chain().
✔ Sets up a Gradio interface with gr.Interface().
✔ Enables document-based Q&A in a web browser with interface.launch().