Choose the model behind your PHP practice
PHP Trail works with three kinds of model: a cloud AI provider reached through a standard API endpoint, a model running locally on your PC, and the built-in model. The choice decides who receives your prompt text. See each option below.
Three ways to connect a model
A cloud AI provider
A cloud provider suits larger changes and longer explanations when your own PC is modest. You enter a standard API endpoint and your API key, and the key stays on your PC. Prompts leave the PC. The prompt text, including any file text the agent reads, goes to that provider, and any charges come from the provider rather than from this site.
A model on your own PC
A local model runs on your PC, so nothing leaves the machine: prompts, file text and replies all stay there. It suits private practice files, but the answers depend on the model you load and on your hardware. Review every diff. A smaller local model may need clearer, shorter prompts than a large cloud one.
The built-in model
The built-in model is the third option, alongside a cloud AI provider and a model running locally on your PC. As with every option, prompts go only to the model endpoint you pick, and the Test button can check that endpoint. Test your choice.
Test the connection first
The Test button checks an endpoint before you rely on it. It shows the response time and the context window, which is how much text the model can take in at once. Run it. After you enter or change an endpoint, a slow reply or a small window tells you what to expect from the next prompt.
Pick a model for the job
Use a bigger model for refactors and for logic that touches several files, and a smaller one for small edits such as a missing semicolon or a renamed variable. When one step is hard, the run_on_model tool sends only that step to a bigger model while the rest of the chat stays on your usual one. Switch per task.
Models differ. Read the diff, run your PHP command and keep Ask first on while you learn, because any model can be wrong and a result is not proof of correctness until you have run and tested it yourself.
