Integrating DeepSeek into the Theia IDE and Theia AI

The past week has witnessed a surge of excitement surrounding DeepSeek, a Chinese AI startup that has made significant waves in the AI landscape. We were of course eager to explore the DeepSeek models in Theia AI and the AI-powered Theia IDE. However, due to the hype, it took us quite a while to gain access. Fortunately, DeepSeek’s models are open-source, allowing users to self-host them instead of relying on their hosted API.

When trying DeepSeek in the Theia IDE, we opted for DeepSeek Chat instead of DeepSeek Reasoner. The reason? DeepSeek Reasoner currently does not support system prompts or function calling, which are crucial for structured interactions in AI-powered development workflows. Additionally, some claiming that DeepSeek Chat is actually better at writing code, while DeepSeek Reasoner excels at understanding complex problems.

Integrating DeepSeek in Theia IDE

Since DeepSeek conveniently provides an OpenAI-compatible API, integrating it into Theia IDE is straightforward. Below is the configuration you can use to add DeepSeek Chat on the DeepSeek Platform as a custom OpenAI model. Of course, you can also host it yourself, e.g. via Ollama, see our documentation for more information.

"ai-features.openAiCustom.customOpenAiModels": [
    {
        "model": "deepseek-chat",
        "url": "https://api.deepseek.com",
        "id": "deepseek-chat",
        "apiKey": "YourAPIKey",
        "enableStreaming": true,
        "supportsDeveloperMessage": false
    }
]

Important: Set "supportsDeveloperMessage": false, as DeepSeek’s API does not yet support the newer “developer” message, which replaced the “system” role in most OpenAI models.

DeepSeek Chat in Action: Theia IDE’s Coder Agent

With Theia IDE 1.58, we introduced the Coder agent, a feature designed to assist developers in writing and understanding code efficiently. We can very easily switch this agent to use DeepSeek:

In the following video, we demonstrate how the Coder agent leverages DeepSeek Chat to generate code suggestions.

Because Theia IDE is based on Theia AI, the flexible framework for building AI-powered tools and IDEs, adding new models is as simple as adjusting a configuration—whether you’re working with Theia IDE or developing a custom AI-enhanced tool. You have the free choice to use any LLM they want—whether it’s DeepSeek, OpenAI, Anthropic or any self-hosted model. This freedom, of course, also comes with the burden of choosing the right model.

To help navigate this decision, we want to highlight this very interesting comparison of various locally hosted models used in Theia IDE:

🔗 Comparative Analysis of TypeScript Code Generation by LLM Models

Of course, every new model comes with its own characteristics, which may require tweaking prompts, refining workflows, or adjusting response handling to get the best results. Theia AI is designed to make this process as smooth as possible, giving users the freedom to experiment, adapt, and fine-tune their setup with ease.

If you’re interested in exploring what Theia AI can do or want to tailor your setup for specific LLMs, now is a great time to start. Whether you’re building AI-powered tools, extending Theia IDE, or contributing to the open-source ecosystem, Theia AI provides the flexibility and control you need to make AI work for you.

For more details, check out our blog post: Introducing AI Support in Theia IDE and Introducing Theia AI (the underlying framework for tool builders).

If you are interested in building custom AI-powered Tools, EclipseSource provides consulting and implementation services backed by our extensive experience with successful AI tool projects. We also specialize in web- and cloud-based tools and support for popular platforms like Eclipse Theia and VS Code.

Get in touch with us to learn more.