Integrating DeepSeek AI Assistant into IntelliJ IDEA: A Step‑by‑Step Guide

The article introduces DeepSeek, an AI‑driven code assistant, and explains why integrating it into IntelliJ IDEA can greatly improve Java development productivity.

1. Environment preparation – Use IntelliJ IDEA 2023.x or newer and install the CodeGPT plugin (a powerful alternative to GitHub Copilot). Ensure Python 3.7+ is installed and added to the system PATH.

Macbook:GithubCommitSpace Aion$ python --version
Python 3.12.8
Macbook:GithubCommitSpace Aion$

2. Install DeepSeek plugin – Open IDEA, go to Settings (Windows/Linux) or IntelliJ IDEA → Preferences (macOS), select Plugins → Marketplace , search for "CodeGPT", and install it. Restart IDEA after installation.

3. Configure DeepSeek

Create an API key on the DeepSeek website ( https://www.deepseek.com/ ).

In IDEA, open the CodeGPT settings, locate the "Custom OpenAI" section, and paste the API key.

Set the chat model URL to https://api.deepseek.com/chat/completions and model name to deepseek-chat .

For the reasoning model, enable code completions, parse response as chat completions, choose the FIM template "DeepSeek Coder", set the same URL, and use model deepseek-reasoner .

After configuring, click Apply and OK.

4. Using DeepSeek

Open the CodeGPT component, select "Custom OpenAI", and type a prompt such as "Write a Java method to filter numbers and provide a test".

The assistant generates the following code:

public class TestNumberFilter {

    public static String filterNumbers(String input) {
        // Use regex to keep only digits
        return input.replaceAll("[^0-9]", "");
    }

    public static void main(String[] args) {
        String testInput2 = "SSFSDFS结论萨芬122432的噶大都是4无人区是14认识832safsss";
        String result = filterNumbers(testInput2);
        System.out.println("Filtered numbers: " + result); // Output: 123456
    }
}

Code completion can be triggered with Ctrl + Space , and code explanation or question answering is available via right‑click context menu options "CodeGPT" and "CodeGPT: Ask Question".

5. Token usage statistics – The article shows screenshots of DeepSeek’s token consumption dashboard, illustrating daily usage, total tokens, and cost for the deepseek-chat model.

Conclusion – By following these steps, developers can successfully integrate DeepSeek into IntelliJ IDEA, leverage AI‑assisted coding features, and monitor usage, thereby enhancing development efficiency and experience.