Choosing your AI model

The hardest task in your session is the one that should set your model choice. Your AI agent uses Novamira in two very different ways, and each puts a different demand on the model.

Match the model to the work

Novamira gives the agent two kinds of tools:

• Specialized abilities (mostly Novamira Pro): the agent calls a tool, the right thing happens. Editing an Elementor element, adding a Bricks widget, working with the integrated plugins. A smaller, faster model handles them just fine.
• Custom PHP code written on the fly (the heart of Novamira Free): the agent writes the code, runs it through Novamira, and reasons about the result. This is where model quality makes the difference.

For one-line PHP any model is fine. For tasks that require reasoning about WordPress internals or chaining several operations, pick the top-tier reasoning model your AI client offers (typically the largest or most capable in its lineup) and turn on extended thinking. Specific model names change every few months; the principle does not.

Try thinking before switching models

When a top model gets stuck, do not switch yet. Most reasoning models support an extended thinking mode: the response takes longer, but the model reasons through the problem before acting and often succeeds where the same model without thinking failed. Only if thinking still cannot get there, level up.

Switching models does not lose your work

You will not lose what matters. The chat history holds the narrative; the artifacts (code, database, drafts) live in the project. Switch model, give the new session a three-sentence recap (what you were doing, where you got to, last blocker), and let it rediscover the rest by reading the project.

One conversation per task

On models with limited context, do not pile multiple tasks into a single chat. Sooner or later the agent compacts the conversation, replacing the full history with a lossy summary, and quality drops from there on. Start a new conversation for each task. On a high-context model this matters less, but it never hurts.

Do not duplicate Pro's specialization with personal skills

If you use Novamira Pro, the plugin already gives the agent an opinionated strategy for Elementor, Bricks, and the third-party plugins it integrates with. It knows which ability to call, in what order, with what data shape.

Do not add your own skill, system prompt rule, or custom instruction on top of what Pro covers. Two strategies for the same job conflict: the agent does not know which one wins, and the result becomes inconsistent.

Personal skills are great for things neither Novamira Free nor Pro covers: your team's deployment workflow, your content guidelines, your design system, your project glossary. Not for things Pro is already specialized in.