Using Large Language Models to Rapidly Build Simple Frontend and Backend Test Tools

In many testing scenarios, colleagues from other departments need test data that cannot be created directly on a page; they may require API calls or database inserts, but lack the necessary skills or permissions. To address this, a lightweight internal testing tool is needed.

The tool must be internal‑only, require no authentication, provide a simple web interface for API calls or basic CRUD operations, and be developed quickly.

For pure‑frontend tools that call APIs, the author first tried low‑code platforms but found them cumbersome: generated code lacks comments, many features are paid, and unknown issues consume time. Instead, a large language model (LLM) was used to generate a complete HTML/CSS/JS page with comments, using Bootstrap components to keep the UI simple.

The LLM prompting strategy follows a role‑task‑instruction pattern: define the model’s role, describe the desired HTML page step‑by‑step, request Bootstrap styling, and add event handling for form submission to the target API. The generated code includes comments for easy adjustments and can be reviewed or regenerated if it does not meet expectations.

For backend tools (e.g., inserting data into a database), a minimal Flask endpoint can be generated by the same LLM. Prompts specify the endpoint’s request method, parameters, and response format, while emphasizing that sensitive information (database credentials, secrets) must not be fed to the model.

Deployment is kept simple: the frontend page can be served via Nginx, and the Flask service can run locally on Windows or Linux. After deployment, a curl command generated by the LLM can be used to verify the API works.

Overall, the practice demonstrates that leveraging an LLM such as 360 Zhì Nǎo can dramatically shorten the development cycle of temporary testing utilities, allowing engineers to focus on business logic rather than boilerplate code.