Integrate ‘DeepSeek AI’ using OLLAMA API on your Azure Windows Server
In this project, we are going to run our own local copy of the most current and trending AI model DeepSeek . Since DeepSeek is OpenSource, getting it running on our PC should be relatively easy.
NOTE: You can also run the same steps on your own personal laptop (Windows or MAC) without actually deploying any cloud VMs.
— -
STEPS
Step 1: Set Up an Azure Windows Server
Step 2: Connect to the VM using Azure Bastion
Step 3: Download and install OLLAMA
Step 4: Download and install DeepSeek
Step 5: Create a GUI using ChatBox AI
Why is this important?
Deploying DeepSeek AI using Ollama API on an Azure Windows Server is an essential project for organizations seeking to harness the power of advanced AI while retaining full control over their data and costs. This approach aligns with modern trends in hybrid cloud infrastructure, enabling innovation, compliance, and scalability.
- On-Premises AI Deployment: By running DeepSeek locally (on a PC or Azure VM), organizations retain full control over their data and AI interactions, minimizing the risk of data leaks or breaches.
- Compliance with Regulations: Industries with strict compliance requirements (e.g., finance, healthcare) can meet GDPR, HIPAA, or other data security regulations by keeping sensitive data on local servers.
- API and GUI Support: By integrating Ollama’s API and ChatBox AI, organizations can create user-friendly interfaces and integrate AI capabilities into existing workflows or applications.
- Seamless Deployment: With a straightforward setup process, technical teams can quickly bring AI models into production environments.
Step 1: Set Up an Azure Windows VM
1. Create an Azure Account: > If you don’t have an Azure account, sign up at [Azure Portal](https://portal.azure.com/). > Ensure you have sufficient credits or a subscription to create and manage resources.
2. Create a Virtual Machine (VM): > Go to the Azure Portal and click on > Create a resource. > Select Virtual Machine > Choose windows server(e.g., Windows Server 2022 Datacenter) as the operating system. > Configure the VM settings: > Size: Select a VM size with sufficient CPU, RAM, and GPU (if required for DeepSeek). For example, choose an NC-series or NV-series VM if GPU acceleration is needed. > Authentication: Set up a username and password or use an SSH key for secure access. > Networking: Ensure the VM has a public IP address for remote access.
— Review and create the VM.
Step 2: Connect to the VM using Azure Bastion
Connect to the VM: Once the VM is provisioned, connect to it using Remote Desktop Protocol (RDP). > Download the RDP file from the Azure Portal and use the credentials you set up during VM creation.
For this test, I will be using the Azure Bastion to RDP into the VM from my local machine.
To do this click on Connect > Bastion > Deploy Bastion.
This could take a while to finish deploying, when it is complete, connect to it using the same credentials created when creating the Server.
Step 3: Download and install OLLAMA
Go to this webpage and download the right spec based on your machine.
https://ollama.com/download/windows
After the download, we will need to install it.
you may need to input the system variable path
Step 4: Download and install DeepSeek
Run the following command to get deepseek running. > check the version ollama --version
ollama run deepseek-r1:1.5b
Now, we have deepseek AI installed. We can run a specific task. For example: We can ask the AI to write us a python code that outputs the number 1–10.
Now, We have our own AI model running safely in our own PC.
Step 5: Create a GUI using ChatBox AI
To do this, go to the following url > download the windows
https://chatboxai.app/en
Choose a user , and then install it
After the installation, open the app to see the GUI.
We will need to configure it to use the Ollama API and Deepseek. To do this; click on, Settings > In Model Provider, click the dropdown menu and select ollama API > In API Host, leave as default > For Model > Click on the dropdown menu and select deepseek
Click save and run a prompt to test the model.
For the prompt, we will be asking the same question. Write a python script that outputs the numbers 1 — 10
Conclusion:
By following this guide, you should be able to successfully run a local DeepSeek AI model on an Azure Windows Server.
Conclusion:
By following this guide, you should be able to successfully run a local DeepSeek AI model on an Azure Windows Server.
