Building Your Own AI Agent: Deploying DeepSeek with Dify on a Local Machine

The world of AI is rapidly evolving, and the ability to create intelligent agents tailored to specific needs is becoming increasingly valuable. This article will guide you through the process of setting up your own AI agent using DeepSeek (a powerful language model) and Dify (an open-source AI application development platform), all on your local machine. This approach offers the benefits of data privacy, security, and customization.

Prerequisites: Hardware and Software Requirements

Before diving into the deployment process, ensure your system meets the following requirements:

Hardware:

• CPU: 2 Cores or more
• RAM: 16 GB or more (recommended)
• Disk Space: Sufficient space for models (DeepSeek-R1 7B requires around 4.7GB)

Software:

• Ollama: A cross-platform large language model management tool. It allows you to easily deploy and manage models like DeepSeek, Llama, and Mistral.
• Docker: A platform for containerization. Docker simplifies the deployment of Dify and other components through pre-built images.
• Git (Optional): Used for downloading the Dify codebase from GitHub.

Step 1: Deploying DeepSeek-R1 with Ollama

Ollama simplifies the processof running LLMs locally, allowing you to deploy DeepSeek with ease improving efficiency and speed.

1. Install Ollama:

   • Visit the Ollama website and download the appropriate version for your operating system (MacOS, Windows, or Linux).
   • Follow the installation instructions provided on the website.
   • Verify the installation by running the command ollama -v in your terminal. This should display the installed Ollama version.

2. Download and Run the DeepSeek Model:

   • Use the following command in your terminal to download and run the DeepSeek-R1 7B model:

   ollama run deepseek-r1:7b
   

   Alternative (Uncensored):

   ollama run huihui_ai/deepseek-r1-abliterated:7b
   

   *Note: The uncensored version might provide different outputs depending on the use case.

   • Choosing the Right Model Size: The 7b in the command refers to the model size (7 billion parameters). Choose a model size appropriate for your hardware. Smaller models require less resources but might have lower performance. Use the ollama website to search for different model sizes.

3. Test the Installation:

   • After the model is downloaded and running, you can directly input questions in the terminal to test it. DeepSeek should respond with generated text.

Step 2: Deploying Dify Community Edition

Dify provides a user-friendly interface for building and managing AI applications. Integrating DeepSeek with Dify empowers you to create more sophisticated agents with knowledge retrieval capabilities.

1. Download the Dify Code:

   • Using Git: Open your terminal and run the following command:

   git clone https://github.com/langgenius/dify.git
   

   • Using Web Download: Navigate to Dify's GitHub repository and download the code as a ZIP file. Extract the contents to a directory of your choice.

2. Navigate to the Docker Directory:

   • Open your terminal and navigate to the dify/docker directory within the downloaded Dify folder:

   cd dify/docker
   

3. Copy the Environment Configuration File:

   • Create a copy of the .env.example file and name it .env:

   cp .env.example .env
   

   • This file contains environment variables that configure the Dify deployment.

4. Configure Docker Proxy (Optional):

   • If you are behind a proxy, configure Docker to use it. This step is crucial for pulling Docker images successfully. Open the Docker configuration file (usually located at /etc/docker/daemon.json on Linux) and add the following content:

   {
     "registry-mirrors": [
       "https://dockerpull.org",
       "https://docker.1panel.dev",
       "https://docker.foreverlink.love",
       "https://docker.fxxk.dedyn.io",
       "https://docker.xn--6oq72ry9d5zx.cn",
       "https://docker.zhai.cm",
       "https://docker.5z5f.com",
       "https://a.ussh.net",
       "https://docker.cloudlayer.icu",
       "https://hub.littlediary.cn",
       "https://hub.crdz.gq",
       "https://docker.unsee.tech",
       "https://docker.kejilion.pro",
       "https://registry.dockermirror.com",
       "https://hub.rat.dev",
       "https://dhub.kubesre.xyz",
       "https://docker.nastool.de",
       "https://docker.udayun.com",
       "https://docker.rainbond.cc",
       "https://hub.geekery.cn",
       "https://docker.1panelproxy.com",
       "https://atomhub.openatom.cn",
       "https://docker.m.daocloud.io",
       "https://docker.1ms.run",
       "https://docker.linkedbus.com"
     ]
   }
   

   Restart Docker after making these changes.

5. Run Dify with Docker Compose:

   • Execute the following command in your terminal to start Dify using Docker Compose

   docker compose up -d
   

   • This command will download the necessary Docker images and start the Dify containers in detached mode.

Step 3: Connecting DeepSeek to Dify

Now that both DeepSeek and Dify are running, the next step is to connect them so you can use DeepSeek as the language model within your Dify applications.

1. Access Dify in Your Browser:

   • Open your web browser and navigate to http://127.0.0.1.

   • You should see the Dify initialization page. Create an administrator account by providing the required information.

2. Configure the Model Provider:

   • Click on your avatar in the top right corner and select "Settings".

   • Navigate to "Model Providers" and select "Ollama".

3. Enter DeepSeek Model Information:

   • Fill out the form with the following information:

   • Model Name: deepseek-r1:7b

   • Base URL: http://<your_machine_ip>:11434

     • Important: If Dify is deployed in a Docker container, use host.docker.internal or your machine's actual IP address instead of 127.0.0.1. Otherwise, Dify will not be able to access the Ollama service running on your host machine.

   • Max Tokens: 32768

4. Save your configuartion.

Step 4: Creating a Knowledge Base and AI Agent

With DeepSeek connected to Dify, you can now build your AI agent with a knowledge base.

1. Create a Knowledge Base:

   • In Dify, go to the "Knowledge" section.
   • Create a new knowledge base.
   • Upload your data by dragging and dropping files or by fetching content from web pages.

2. Create an Agent:

   • Go to the "Projects" section and create a new project.

   • Configure the project as an "Agent".

   • Connect the knowledge base you created to the agent.

   • Define the agent's role and instructions to tailor its behavior.

   • This is where you leverage the knowledge base and instruct the agent on how to use it.

Further Resources

For more information and troubleshooting, refer to these resources:

• Setting up Dify on a Local Machine: https://blog.csdn.net/piaoxue820/article/details/138191528
• Connecting to Locally Deployed Models with Ollama: https://docs.dify.ai/zh-hans/development/models-integration/ollama

Conclusion

By following these steps, you can establish a local environment for creating AI agents using DeepSeek and Dify, with local data processing. This setup allows you to experiment with AI, customize agents for specific tasks, and ensure data security within your local infrastructure, making it an excellent way to explore AI development.