AI Model helps turn a description, a small requirements document or a diagram into a proposed data model. You describe the entities you need, inspect the result and decide whether to add it to your project. The assistant does not generate or run your application.
Open AI Model in the generator navigation. The conversation and input share one panel; the current model and proposed additions appear below it. You can inspect an entity's fields and relationships before accepting anything.
What you can ask it to do
| Task | Current behavior |
|---|---|
| Create new modules, persistent entities or virtual entities | The assistant proposes a model; the generator checks whether it is valid |
| Refine a new proposal | Continue the conversation before Apply to change its fields, names or relationships |
| Connect a new entity to a saved entity | Allowed when the relationship can be added without changing the saved side |
| Change a saved entity, field, relationship or module | Do this yourself in Content |
| Run Generate, migrations or deployment | These remain your separate actions |
Saved means present in the project model, even if you have not run Generate yet. After you apply new entities, they are saved too. The assistant cannot edit them in a later request. This restriction is checked by the application and the AI service; it is not just advice in the prompt.
For example, ask for RELATED_ENTITY_NAME and ENTITY_NAME together, then refine both while they are still a proposal. If RELATED_ENTITY_NAME is already saved and a relationship requires adding an inverse field to it, configure that part manually in Content.
What is sent and what stays local
An AI request includes the current Content model, necessary conversation context and the files you explicitly attach. Model review, validation and Apply happen in your local generator. Chat text is saved locally; the AI service does not receive permission to write project files or execute commands. File handling is described in Attachments and conversation history.
AI needs an internet connection and an account with AI assistant access and available allowance. Sign in through Account; you do not need your own OpenAI key or a local proxy. The model used for the request is configured by the Admingen service.
Your ordinary model editing and code generation remain available without AI. The separate account and project limits explain when an account is needed.