AI-powered code generation in the conzept 16 development environment. This article provides a detailed look at the current state of development.
Overview
As part of the research project AIDev, Vectorsoft AG is integrating artificial intelligence into yeet. In an initial prototype, AI support has now also been tested in conzept 16. In a survey conducted during AI Day at the end of January, the majority of participants ranked general interest in AI support as a top priority. Among other topics, the following were discussed:
- Providing Web-Based Online Help
- AI-Based Source Code Generation
With the release of version 5.9.07, an initial version of the online help was already published. During the implementation of the web-based help, special attention was paid to ensuring “good readability” for AI. There’s a reason for this. The help contains over 8,000 pages covering all areas of conzept 16 and is therefore a crucial source of data for AI analysis. The same applies to the CodeLibrary—a collection of examples on various development topics for conzept 16. Together, the online help and the CodeLibrary are made available for source code generation and are utilized by the AI plugin. The prototype was presented on March 25 during the AI Exchange.
Technical Background
In order for conzept 16 source code to be generated using AI, the AI requires high-quality examples and information on development with conzept 16. For example, if the AI is to generate source code for sending an email, it must know which commands conzept 16 provides for this purpose. The AI obtains this knowledge base in the form of vector databases through information from the online help and CodeLibrary. Using RAG (Retrieval-Augmented Generation), the AI model is enriched with reliable and targeted content for this purpose.
Technically speaking, the prototype is a conzept 16 plugin. The plugin interface allows the designer to be extended with functionality that it does not offer by default. Developing the prototype as a plugin has the advantage of ensuring independence from conzept 16 releases, meaning there is no need to wait for a specific release to be generated. A second advantage is that conzept 16 plugins are developed in conzept 16’s own programming language, making them easy for conzept 16 developers to get started with. If the plugin’s functionality is insufficient, it can be extended using the conzept 16 API, allowing external libraries written in C/C++ to be used as well.
For its AI functionality, the plugin uses the Google Gemini API. The necessary queries have been fully implemented in conzept 16 code, as has the rest of the plugin’s functionality. An external component is required only for the RAG system.
Application
The plugin prototype was implemented as a bot.
In the lower input field, you can enter instructions for the desired code generation (e.g., “generate a ‘hello world’ example”). The AI-generated conzept 16 code will then appear in the field above it. By clicking the button 
, the code can be copied to the clipboard and then pasted into the editor for further editing.
Outlook
From the last AI discussion, we received suggestions for further developing the plugin, including:
- Greater integration with the designer or code editor
- Standard prompts as options for frequently used AI commands
- Advanced features, such as code explanation and code analysis.
If we’ve piqued your interest, please let us know.
What features do you need? Would you like to test the plugin or help with its further development? If so, simply contact our support.