This is an ongoing project working with the DeepSeek API to enhance my skills in developing AI-based tools, prompt engineering and API integration. The goal is to develop a tool to automate the development of simple web applications. The user inputs a natural language prompt outlining their desired web application via a web UI, which is then fed with additional context to the DeepSeek API. The API returns a JSON file which is then used to generate all the necessary code files, which the user can then download and fine-tune.
A functional backend is fully implemented - a command line prompt can be passed to the DeepSeek API by the program, and the resulting JSON that is produced by the AI model is processed and converted to a fully functional web application.
Refinements that can be made to the backend in its current form include fine-tuning the context that is added to the DeepSeek prompt to ensure a high-quality response. Such improvements might include a system prompt in addition to the user prompt, alterations to the temperature (randomness) parameter, etc.
The next step is to develop a web UI for the application. This will be implemented using React - at present, I am completing a JavaScript and React course to gain the skills required for this part of the project.
© 2025 Fergal Riordan. All rights reserved.