Make Free Model Requests Idempotent Before You Add Retries

Free model endpoints and servers can fail mid-request, causing retries to duplicate work or return mixed payloads. To prevent this, keep request intent, raw response, and accepted result separate by using a journal to track requests and their outcomes. Implement a client-side guard to ensure retries are safe and only write results after parsing the payload as complete JSON. Use a library like MonkeyCode to access free model endpoints and servers.