Choose a model for your portfolio tracker work

Codeless Pilot talks to a model through an API endpoint. The choice has 3 options. The model you pick sets what the agent can handle and where the prompt text goes, which matters before you load real exports.

Three ways to connect a model

A cloud AI provider

A cloud provider suits long files and harder logic. The prompt text, including any file text the agent reads, goes to that provider through a standard API endpoint, and the provider may charge separately. Your API key stays on your PC. Start with placeholder data.

A model on your PC

A model running locally suits exports you would rather keep at home. Nothing leaves the machine. Quality depends on the model you load and the hardware you have, and a small model may need smaller steps, so ask for one change per request and read each diff.

The built-in model

The built-in model is the option to try when you have no endpoint or local model yet. Test it first. Read the privacy page, try placeholder data such as COINA and COINB, and send real exports only once you are comfortable with where they go.

Check the connection with Test

The Test button checks that your endpoint answers and shows the response time and the context window of the model. That window tells you how much of a long export fits in one request. Press it after every change. A failure means stop.

Pick a model for the job

Match size to task. A bigger model suits refactors and complex logic, such as merging 3 exports with different layouts into one table, while a smaller one is enough for small edits like renaming a column. When one step is hard, run_on_model sends just that step to a larger model.

Quality varies between models. Review each diff, keep Ask first on while you learn, and compare the totals with your statements, because a confident answer can still contain a wrong number.