Pick the model behind your log reviews
SmartCode Relay works with a cloud AI provider through a standard API endpoint, a model running locally on your PC, or the built-in model. Three options exist. Your choice decides which text leaves the machine, so read the privacy note under each option.
Three ways to connect a model
A cloud AI provider
A cloud AI provider connects through a standard API endpoint and suits hard steps, such as reading a long set of server settings. The prompt text, including any file text the agent reads, goes to that provider, so remove visitor IP addresses from exported logs when you can. Check the provider's terms. Charges, if any, come from the provider.
A model on your PC
A model running locally on your PC suits logs and settings you would rather not send anywhere. Nothing leaves the machine. Response time and the usable context window depend on your hardware, so press Test to see both before you rely on the model for a long log.
The built-in model
The built-in model is the third choice, next to a cloud AI provider and a model running locally on your PC. Prompts go only to the model endpoint you pick, and the privacy page lists the optional network features that use the network. Read it first.
What the Test button shows
The Test button checks an endpoint before you start a review. It shows the response time and the context window, which tells you how much log text and settings fit into one request. If the test fails, check the endpoint details first. Test again after changes.
Choosing a model for the task
Use a bigger model for the hard parts, such as reading a long configuration with many rules, and a smaller one for small edits like a timeout value. run_on_model sends one hard step to a bigger model, and the rest of the chat carries on with the smaller one. Only that step moves.
Quality varies between models, so review every diff before you accept it. In Ask first mode the agent pauses before each change, and you test the result with your own test command on a copy. Models do make mistakes.
